Recent Developments and Applications In Social
Transcription
Recent Developments and Applications In Social
Recent Developments and Applications In Social Research Methodology Proceedings of Sixth International Conference on Logic and Methodology August 16 – 20, 2004 1 Recent Developments and Applications in Social Research Methodology August, 16-20, 2004, Amsterdam University, Amsterdam, The Netherlands In August 2004, the ISA Research Committee on Logic and Methodology, jointly with the Department of Methodology and Statistics Faculty of Utrecht University, the Netherlands Institute for the Social Sciences (SISWO) and the Netherlands Association for Methodological Research in the Social Sciences (NOSMO), organized the Sixth Conference in a series of four-yearly International Conferences on Social Science Methodology. Approximately five hundred researchers from different countries presented their papers in more than sixty sessions. Eight free tutorials in advanced methods offered participants a possibility to be introduced in the State of the Art and Science of Methodology. The organizers hope that the publication of papers in the proceedings will challenge further research in the field of the social science methodology. 2 General Information The registration fee includes: conference materials, refreshments, the welcome reception (in Hal E) on Tuesday evening, a copy of the CD-ROM with proceedings and this volume of abstracts. Registration hours: Tuesday, August 17 Wednesday, August 18 Thursday, August 19 09.00 - 17.00 09.00 – 17.00 09.00 – 13.00 Internet Facilities An Internet room has been reserved for the use of registered conference participants. The Internet Room is located in the H Building, third floor (take the elevator, close by the entrance of the E Building, Roeterstraat 11) Internet Opening Hours Tuesday Afternoon 13.00 – 18.00 Wednesday 09.00 - 18.00 Thursday 09.00 - 18.00 Friday 09.00 - 13.00 Presentation Facilities In each room of presentation a PC (Laptop), Beamer, Overheadprojector and a white/or/black Board is available. Paper presentators are advised to use an USB-stick or CD-ROM for their presentations. At request Internet can be made available and Video Facilities. Technical assistestance can be called on by the student assistent which is present in each room of presentation. 3 Scientific Executive Committee Nancy Andes Department of Sociology, University of Alaska Jörg Blasius University of Bonn, Seminar for Sociology Cor van Dijkum (chair) Department of Methodology and Statistics, Utrecht University Harry Ganzeboom Department of Sociology Free University Henk Kleijer, SISWO Nienke Siemonsma, Staff member, City of Noordwijk Document Editing Staff, Department of Methodology and Statistics Marieke Westeneng Herbert van Willenswaard Bram Vermeulen CD-ROM Mastering Branko van Hilten Printing Manager, Faculty of Social Sciences Marcel van der Linde Logistics Support, Department of Methodology and Statistics Roland Holdinga Maureen Postma Technical assistance during the conference Branko van Hilten Herbert van Willenswaard Website RC33 Design and Support Cor van Dijkum Henk Janssen Nienke Siemonsma Branko van Hilten Registration Agency Keynote Martien van Geene Laurette van Geene 4 5 6 Recent Developments and Applications In Social Research Methodology RC33 Sixth International Conference on Social Science Methodology Overview of Sessions and Programme August 16 – 20, 2004 University of Amsterdam The Netherlands Department of Methodology and Statistics Faculty of Social Science Utrecht University Heidelberglaan 2 3584 CS Utrecht The Netherlands 7 Overview of Sessions Topics, Starting-Times and Locations Building A Building E Plenary Speakers Lyn Richards: Qualitative Software and Methodological change: Myths, Mindsets and Methodological Mania Jan de Leeuw: Description, Prediction, Inference Computer Assisted Research Computer-Assisted Sociological Research Computer-Assisted Sociological Research Discoursive Etnography The Achievements and Challenges of Qualitative Computing Computer aided Content Analysis - Research, Methods and its Future ( Part I ) Computer aided Content Analysis - Research, Methods and its Future ( Part II ) Computer aided Content Analysis - Research, Methods and its Future ( Part III ) Computer aided Content Analysis - Research, Methods and its Future ( Part IV ) The Response Process in Establishment Surveys Qualitative Methods The Ethics and Social Relations of Research The Ethics and Social Relations of Research Studying Sensitive Topics in Qualitative Research Recent Developments in Ethnomethodological and Conversation-Analytic Research ( Part I ) Recent Developments in Ethnomethodological and Conversation-Analytic Research ( Part II ) Research Methodology and the Challenges of Social Diversity ( Part I ) Research Methodology and the Challenges of Social Diversity ( Part II ) Action Research Evaluating Cognitive Methods to Pre-Test Questionnaires Tuesday, 13:30 Bldg. A; Room A Thursday, 14:00 Bldg. A; Room A Tuesday, 15:00 Tuesday, 16:10 Wednesday, 11:30 Bldg. A; Room 306 Bldg. A; Room 306 Bldg. A; Room 108 Wednesday, 16:00 Bldg. A; Room 108 Thursday, 9:00 Bldg. A; Room 306 Thursday, 11:00 Bldg. A; Room 306 Thursday, 15:30 Bldg. A; Room 306 Friday, 11:30 Bldg. A; Room 306 Friday, 9:00 Bldg. A; Room 306 Tuesday, 15:00 Tuesday, 16:10 Bldg. A; Room 108 Bldg. A; Room 108 Wednesday, 9:00 Bldg. E; Room 010 Wednesday, 11:30 Bldg. E; Room 010 Wednesday, 16:00 Bldg. A; Room 305 Wednesday, 14:30 Bldg. E; Room 010 Thursday, 11:00 Bldg. A; Room 108 Wednesday, 16:00 Bldg. E; Room 010 Thursday, 9:00 Bldg. A; Room 108 8 Qualitative Methods Continued Secondary Analysis of Qualitative Data Rethinking Professional Ethics Feminism, Methodology and Methods (Part II ) Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data ( Part I ) Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data ( Part II ) Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data ( Part III ) Interrogating Conventionality in Qualitative Research: Bricolage and Ingenuity in Interdisciplinary Fieldwork Mixing Qualitative and Quantitative Methods Empirical Designs and their Theoretical Implications ( Part I ) Mixing Qualitative and Quantitative Methods Empirical Designs and their Theoretical Implications ( Part II ) Analyzing Interviews ( Part I ) Analyzing Interviews ( Part II ) Miscellaneous Survey Research Cognitive and Motivational Determinants for response Effects in Survey Research(Part I) Cognitive and Motivational Determinants for response Effects in Survey Research(Part 2) Approaches to Using Mixed Methods Interaction Analysis and Data Quality of Survey Interviews Estimating Informed Opinion: Advances in the Theory and Measurement ( Part I ) Estimating Informed Opinion: Advances in the Theory and Measurement ( Part II ) Telephone Survey Methodology in the 21st Century ( Part I ) Telephone Survey Methodology in the 21st Century ( Part II ) Telephone Survey Methodology in the 21st Century ( Part III ) Response Latencies in Survey Research: Theory, Measurement and Predictive Power Thursday, 15:30 Friday, 9:00 Friday, 11:30 Bldg. A; Room 108 Bldg. A; Room 108 Bldg. A; Room 108 Tuesday, 16:10 Bldg. E; Room 010 Wednesday, 9:00 Bldg. A; Room 305 Wednesday, 11:30 Bldg. A; Room 305 Wednesday, 14:30 Bldg. A; Room 305 Thursday, 11:00 Bldg. E; Room 010 Thursday, 15:30 Bldg. E; Room 010 Friday, 9:00 Friday, 11:30 Thursday, 9:00 Bldg. E; Room 010 Bldg. E; Room 010 Bldg. E; Room 010 Tuesday, 15:00 Bldg. A; Room 102 Tuesday, 16:10 Bldg. A; Room 102 Wednesday, 9:00 Bldg. A; Room 102 Wednesday, 11:30 Bldg. A; Room 102 Wednesday, 14:30 Bldg. A; Room 102 Wednesday, 16:00 Bldg. A; Room 102 Thursday, 9:00 Bldg. A; Room 102 Thursday, 11:00 Bldg. A; Room 102 Thursday, 15:30 Bldg. A; Room 102 Friday, 9:00 Bldg. A; Room 102 9 Complex and Longitudinal Data Analysis Classification and Structure: Recent Developments in Latent Class and Latent Trajectory Analysis and New Developments in Cluster Analysis ( Part I ) Classification and Structure: Recent Developments in Latent Class and Latent Trajectory Analysis and New Developments in Cluster Analysis ( Part II ) Multilevel Analysis ( Part I ) Multilevel Analysis ( Part II ) Measurement (In)Variance in Longitudinal Research ( Part I ) Measurement (In)Variance in Longitudinal Research ( Part II ) Methods of Social Network Analysis (Part I) Methods of Social Network Analysis(Part II) Panel Data Analysis Combining Data from Different Sources (Part I) Combining Data from Different Sources (Part II) Information and Simulation Systems Non Linear Modelling ( Part I ) Non Linear Modelling ( Part II ) Non Linear Modelling ( Part III ) The Applicability of Agent-Based Social Simulation ( Part I ) The Applicability of Agent-Based Social Simulation ( Part II ) The Applicability of Agent-Based Social Simulation ( Part III ) Informetric and Text Based Procedures for Information Retrieval and Analysis, e.g. in the Context of Digital Libraries ( Part I ) Informetric and Text Based Procedures for Information Retrieval and Analysis, e.g. in the Context of Digital Libraries ( Part II ) Informetric and Text Based Procedures for Information Retrieval and Analysis, e.g. in the Context of Digital Libraries ( Part III ) Time Series Analysis The Statistical Estimation of Linear and NonLinear Stochastic Differential Equations Tuesday, 15:00 Bldg. E; Room 150 Tuesday, 16:10 Bldg. E; Room 150 Wednesday, 9:00 Wednesday, 11:00 Bldg. E; Room 150 Bldg. E; Room 150 Wednesday, 14:30 Bldg. E; Room 150 Wednesday, 16:00 Bldg. E; Room 150 Thursday, 9:00 Friday, 11:30 Thursday, 11:00 Bldg. E; Room 150 Bldg. E; Room 150 Bldg. E; Room 150 Thursday, 15:30 Bldg. E; Room 150 Friday, 9:00 Bldg. E; Room 150 Tuesday, 15:00 Thursday, 11:00 Friday, 11:30 Bldg. E; Room 007 Bldg. E; Room 007 Bldg. E; Room 007 Tuesday, 16:10 Bldg. E; Room 007 Wednesday, 9:00 Bldg. E; Room 007 Wednesday, 11:30 Bldg. E; Room 007 Wednesday, 14:30 Bldg. E; Room 007 Wednesday, 16:00 Bldg. E; Room 007 Thursday, 9:00 Bldg. E; Room 007 Thursday, 15:30 Bldg. E; Room 007 Friday, 9:00 Bldg. E; Room 007 10 General Statistics Assesment of Model Fit ( Part I ) Assesment of Model Fit ( Part II ) New Directions in Political (and Social) Methodology ( Part I ) New Directions in Political (and Social) Methodology ( Part II ) Recent Developments in Nonparametric Item Response Theory ( Part I ) Recent Developments in Nonparametric Item Response Theory ( Part II ) Session by Nanny Wermuth ( Part I ) Session by Nanny Wermuth ( Part II ) The Nets: between Neural Networks and Network Analysis. Fuzzy Logic, Complexity and/or Network Analysis Social Theory Miscellaneous DataCollection, Sampling and Non-Response Improving the Validity of Fraud Research Using Instruments from the Social Sciences Data Nonresponse ( Part I ) Nonresponse ( Part II ) Nonresponse ( Part III ) Nonresponse ( Part IV ) Aided Recall Techniques in Survey Interviews Data Collection Sampling Methods ( Part I ) Sampling Methods ( Part II ) Latent Variables Modelling Latent Growth Curve Structural Equation Models, Autoregressive and Hybrid Models Multidimensional Scaling and Facet Theory Correspondence Analysis ( Part I ) Correspondence Analysis ( Part II ) Correspondence Analysis ( Part III ) Correspondence Analysis ( Part IV ) Classification and Visualisation Improving Survey Methodology: The European Social Survey ( Part III ) Tuesday, 15:00 Tuesday, 16:10 Bldg. E; Room 003 Bldg. E; Room 003 Wednesday, 9:00 Bldg. E; Room 003 Wednesday, 11:30 Bldg. E; Room 003 Wednesday, 14:30 Bldg. E; Room 003 Wednesday, 16:00 Bldg. E; Room 003 Thursday, 9:00 Thursday, 11:00 Bldg. E; Room 003 Bldg. E; Room 003 Friday, 9:00 Bldg. E; Room 003 Thursday, 15:30 Bldg. E; Room 003 Tuesday, 15:00 Bldg. A; Room 305 Tuesday, 16:10 Wednesday, 9:00 Thursday, 15:30 Friday, 9:00 Friday, 11:30 Bldg. A; Room 305 Bldg. A; Room 306 Bldg. A; Room 305 Bldg. A; Room 305 Bldg. A; Room 305 Wednesday, 11:30 Bldg. A; Room 306 Wednesday, 14:30 Thursday, 9:00 Thursday, 11:00 Bldg. A; Room 306 Bldg. A; Room 305 Bldg. A; Room 305 Wednesday, 9:00 Bldg. E: Room 008 Wednesday, 11;30 Wednesday, 14:30 Thursday, 15:30 Friday, 9:00 Friday, 11:30 Wednesday, 16:00 Bldg. E: Room 008 Bldg. E: Room 008 Bldg. E: Room 008 Bldg. E: Room 008 Bldg. E: Room 008 Bldg. E: Room 008 Thursday, 9:00 Bldg. E: Room 008 11 International Comparative Research Improving Survey Methodology: The European Social Survey ( Part I ) Improving Survey Methodology: The European Social Survey ( Part II ) Comparable Data across Countries Methodological Issues in Designing and Implementing Cross-Cultural Surveys (Part I) Methodological Issues in Designing and Implementing Cross-Cultural Surveys (Part II) Methodological Issues in Designing and Implementing Cross-Cultural Surveys (Part III) Methods of Cross-National Comparative Analysis Cross National Harmonization of SocioDemographic Variables ( Part I ) Cross National Harmonization of SocioDemographic Variables ( Part II ) Cross-National Comparison of Daily Activity Patterns: Issues in the Collection and Analysis of Cross-National Time Data ( Part I ) Cross-National Comparison of Daily Activity Patterns: Issues in the Collection and Analysis of Cross-National Time Data ( Part II ) Data Collection & Internet Surveys Visual Design and Layout in Internet and Mail Surveys ( Part I ) Visual Design and Layout in Internet and Mail Surveys ( Part II ) The Internet and Social Research: New Opportunities, New Challenges ( Part I ) The Internet and Social Research: New Opportunities, New Challenges ( Part II ) Internet Surveys ( Part I ) Internet Surveys ( Part II ) Internet Surveys ( Part III ) Internet Surveys ( Part IV ) Internet Surveys ( Part V ) Miscellaneous Tuesday, 15:00 Bldg. E; Room 151 Tuesday, 16:10 Bldg. E; Room 151 Wednesday, 9:00 Bldg. E; Room 151 Wednesday, 11:30 Bldg. E; Room 151 Wednesday, 14:30 Bldg. E; Room 151 Thursday, 15:30 Bldg. E; Room 151 Wednesday, 16:00 Bldg. E; Room 151 Thursday, 9:00 Bldg. E; Room 151 Thursday, 11:00 Bldg. E; Room 151 Friday, 9:00 Bldg. E; Room 151 Friday, 11:30 Bldg. E; Room 151 Tuesday, 15:00 Bldg. A; Room A Tuesday, 16:10 Bldg. A; Room A Wednesday, 14:30 Bldg. A; Room 303 Wednesday, 16:00 Bldg. A; Room 303 Thursday, 9:00 Thursday, 11:00 Thursday, 15:30 Friday, 9:00 Friday, 11:30 Wednesday, 11:30 Bldg. A; Room A Bldg. A; Room A Bldg. A; Room A Bldg. A; Room 303 Bldg. A; Room 303 Bldg. A; Room 303 Tuesday, 15:00 Bldg. E; Room 004 General Methodology Is There a Fundamental Change in the Individual Shaping of the Life Course in Modernity? Or How Can We Do Research on Fundamental Transformations in Modernization? 12 General Methodology continued Methodological Foundations of Sociological Analysis of Social Sphere ( Part I ) Methodological Foundations of Sociological Analysis of Social Sphere ( Part II ) Methodological Foundations of Sociological Analysis of Social Sphere ( Part III ) Methodological Foundations of Sociological Analysis of Social Sphere ( Part IV ) Feminism, Methodology and Methods (Part I ) Complex Societal Problems ( Part I ) Complex Societal Problems ( Part II ) Miscellaneous Wednesday, 9:00 Bldg. E; Room 004 Thursday, 15:30 Bldg. E; Room 004 Friday, 9:00 Bldg. E; Room 004 Friday, 11:30 Bldg. E; Room 004 Wednesday, 11:30 Thursday, 9:00 Thursday, 11:00 Wednesday, 14:30 Bldg. E; Room 004 Bldg. E; Room 004 Bldg. E; Room 004 Bldg. E; Room 004 13 FREE TUTORIALS IN ADVANCED AND RECENT METHODS These Tutorials take place before and after the conference and are free of charge. Monday, 16th August: Simulation for the Social Scientist Klaus G. Troitzsch & Nigel Gilbert Time: 14.00 - 17.45 Place: SISWO, Plantage Muidergracht 4, 1018 CV Amsterdam (5 minutes to walk from the Conference Venue) Introduction to Bayesian Statistics Herbert Hoijtink & Irene Klugkist Time: 14.00 - 17.00 Place: Amsterdam University, Roeterstraat 11, Room E-007 (HAL E) Choosing qualitative software: why does QSR have two packages (N6 and NVivo) and how to choose between them? Tom and Lyn Richards, QSR International, Australia., http://www.qsrinternational.com Time: 14.00 - 18.00 Place: Amsterdam University, Roeterstraat 11, Internet Room 4 (Third floor, HAL H, Take the Elevator left of the Entrance) Tuesday, 17th August: Computer aided Text Analysis with TextQuest Harald Klein, Developer of TextQuest, Social Science Consulting , Rudolstadt, Germany Time: 09.00 - 12.30 Place: Amsterdam University, Roeterstraat 11, Room E-007 (HAL E) Introduction to QCA: approach and applications Alain Gottcheiner, Laboratory of Research in Mathematics and Social Science, Free University of Brussels (Université Libre de Bruxelles) Sakura Yamasaki, Centre for Comparative Politics, University of Louvain (Universite catholique de Louvain) Time: 9.00-12.30 Place: Amsterdam University, Roeterstraat 11, E003 (HAL E) Everything you ever wanted to know about preparing qualitative data ...but were afraid to ask Louise Corti and John Southall, UK Data Archive, University of Essex Time: 9.00-12.30 Place: Amsterdam University, Roeterstraat 11, Internet Room 4 (Third floor, HAL H, Take the Elevator left of the Entrance) 14 Friday, 20th August: Overview of Software for Qualitative Data Analysis. Sharlene Hesse-Biber, Boston College, Developer of Software Package, HyperRESEARCH Raymond Maietta , ResearchTalk Inc., Bohemia, Long Island Time: 14.00 - 17.00 Place: Amsterdam University, Roeterstraat 11, E003 (Hal E) Modern Multidimensional Scaling Patrick J.F. Groenen, Econometric Institute, Erasmus University Rotterdam Time: 14.00 - 17.00 Place: Amsterdam University, Roeterstraat 11, Room E-007 (HAL E) 15 Tuesday, 17th August 2004 General Program 9:00 13:00 Registration Bldg. A; Hall 13:00 13:30 Welcome Bldg. A; Room A 13:30 - Plenary Speech: Lyn Richards, Australia Subject: Qualitative Software and Methodological Change Myths, Mindsets and Methodological Mania 14:30 Bldg. A; Room A 14:30 15:00 15:00 16:10 16:10 17:45 Tea 18:00 21:00 Welcome Reception Bldg. E; Hall Session Session 16 Tuesday 15:00 - Computer-Assisted Sociological Research Organizer: Karl M. van Meter, France 17:45 Assessing Attitudes toward Muslims in the Netherlands: Does asking both Open and Closed Questions Provide more Insight? Christine L. Carabain, The Netherlands Methodological Issues in Designing and Implementing Cross-Cultural Surveys Assessing cross-national construct equivalence in the ESS Jaak Billiet; Jerry Welkenhuysen-Gybels, Belgium Bldg. A Rm. 306 Combining Test Interviews with Experimental Surveys to Compare Question Formats in a Postal Survey on Alcohol Consumption Harrie Jansen; Viviënne Lahaut, The Netherlands Elites at Home: A Multimethod Community Case Study of a Wealthy Chicago Suburb Al Hunter, USA A Computer Assisted Mixed Method and Mixed Model Analysis of Police Decision Making Jennifer L. Schulenberg, Canada When Quantitative is Qualitative: An Integrated Analysis Rod Gutierrez; Pat Bazeley, Australia The Combined Use of Qualitative and Quantitative Methods in Educational Research K. Niglas, Estonia 17 Tuesday 15:00 - The Ethics and Social Relations of Research Organizers: Carole Truman, UK ; Anne Ryen, Norway 17:45 The Best and Worst of Research Relations David Chavis, USA On Vulnerabilities and Responsibilities in the Qualitative Interview Anne Halvorsen, Norway Informed Consent, Gatekeepers and Go-Betweens Bldg. A Rm. 108 Sue Heath; Vikki Charles; Graham Crow; Rose Wiles, UK Wading through "The Jungle" with my Main Informant A critical discussion of ethical practice in cross-cultural ethnography Anne Ryen, Norway Emotional Implications in Informant-Researcher Relations Trond Stalsberg Mydland, Norway The Regulation of Research Ethics Some Challenges of Social Diversity Carole Truman, UK Ethics and Power in Community-Campus Partnerships for Research Susan Boser, USA 18 Tuesday 16:10 17:45 Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data ( Part I ) Organizers: Ruth Sautu, Argentina ; Silvana Figueroa; Jochen Dreher, Germany Transcendental Realism in a Particular Methodological Secuency Bldg. E Rm. 010 Beatriz Irene Wehle, Argentina Post-communist Systemic Transitions, National Identity and Mass Media: Theoretical and Methodological Aspects of Analysis Izabella Zandberg, USA Representations of Death: Theoretical and Methodological Approaches Christine Matter, Germany Tracing symbols. The Hermeneutic Analysis of Paradoxical Constructions within Argentine Collective Symbolism Silvana K. Figueroa; Jochen Dreher, Germany 19 Tuesday 15:00 17:45 Cognitive and Motivational Determinants for Response Effects in Survey Research Organizer: Volker Stocké, Germany National Identity, Nationalism, or Patriotism: Do we Really Measure what we Assume? Bldg. A Rm. 102 Horst-Alfred Heinrich; Karsten Stephan, Germany A Cognitive Model for Survey Response Effects Ákos Münnic, Hungary ; Jaap Murre; Willem Saris, The Netherlands Direction of Response Scales, Change of Numerical Values and Respondents' Behavior Bettina Langfeldt, Dagmar Krebs, Jürgen H.P. Hoffmeyer-Zlotnik, Germany Cognitive and Motivational Determinants for the Strength and Direction of Social Desirability Bias in Racial Attitude Surveys Volker Stocké, Germany 20 Tuesday 15:00 17:45 Classification and Structure: Recent Developments in Latent Class and Latent Trajectory Analysis and New Developments in Cluster Analysis Organizers: Johann Bacher; Jost Reinecke; Christian Tarnai, Germany Part I 15:00 16:10 Methods for Comparing Latent Class Cluster Solutions Michael D. Larsen, USA L? -consensus for Dendrograms : An Alternative to the Average Consensus Procedure Bldg. E Rm. 150 Guy Cucumel, Canada Classifying Adolescents Using Value Orientations - A Comparison of Cluster Analysis and Latent Class Analysis Andreas Pöge, Germany Detecting Aberrant Paterns of Change for the Dimensions of a "Quality of Life" Questionnaire by means of Linear Mixed Models Herbert Matschinger, Germany Part II 16:10 17:45 SPSS TwoStep Clustering - A First Evaluation Johann Bacher; Knut Wenzig; Melanie Vogler, Germany Clustering Messy Social Data with ClustanGraphics David Wishart, UK Cluster Analysis Software: Proposals of Tools for Validation Michael Wiedenbeck; Cornelia Züll, Germany Empirical Validation of "Whisky Classified" David Wishart, UK 21 Tuesday 15:00 - Non Linear Modelling ( Part I ) Organizer: Cor van Dijkum, The Netherlands 16:10 Dynamics of Functional Connectivity in Ongoing, No-task EEG. Evidence for Scale-Free Properties and Explanation in Terms of Coupled Limit Cycle Oscillators Bldg. E Rm. 007 C.J. Stam; E.A. de Bruin; T. Montez; K. Linkenkaer-Hansen; M. Breakspear; B.W. van Dijk, The Netherlands Dynamical Complexity. A measure of Critical Instability During Human Change Processes S. Weihrauch; G.Schiepek, The Netherlands Evolutionary Economics and Complexity Carl Henning Reschke, Germany The Applicability of Agent-Based Social Simulations ( Part I) 16:10 17:45 Organizers: Klaus G. Troitzsch, Germany ; Nigel Gilbert, UK Tracking the Invisible Hand: An A.B.M. Approach to CDA Design with Evolutionary Behaviour M. Posada Calvo; Cesaro Hernández Iglesias; Adolfo López Paredes, Spain Development of a Hybrid Multi-Agent Model for Modelling Petrol Price Markets Alison J. Heppenstall; Andrew J. Evans; Mark H. Birkin, UK Dynamics of International Negotiations. A Simulation of EU Intergovernmental Conferences Nicole J. Saam; Paul W. Thurner; Frank Arndt, Germany On a Theorem in the Theory of Guerilla Warfare Jim Doran, UK 22 Tuesday 15:00 - Assessment of Model Fit Organizer: Tamas Rudas, Hungary 17:45 Analysis of Deviance with Penalized Likelihoods David Firth, UK Bldg. E Rm. 003 Examining the Scalability of Intimacy Permissiveness in Taiwan Pei-shan Liao; Su-hao Tu, Taiwan Additive and Multiplicative Effects in a Fixed 2 x 2 Design using ANOVA can be difficult to Differentiate: Demonstration and Mathematical Reasons Johannes A. Landsheer; Godfried van den Wittenboer; Gerard H. Maassen, The Netherlands How to Evaluate the Fit of Linear and Generalized Multilevel Models Wolfgang Langer, Germany Collinearity involving Ordered and Unordered Categorical Variables John Hendrickx; Ben Helzer; Manfred te Grotenhuis; Jan Lammers, The Netherlands Testing the Fit of Conditional Independence with a Continuous Control Variable Wicher Bergsma, The Netherlands Assessment of Model Fit using Incomplete Data Tamas Rudas, Hungary 23 Tuesday 15:00 16:10 Improving the Validity of Fraud Research Using Instruments from the Social Sciences organizers: Gerty Lensvelt-Mulders; Peter van der Heijden, The Netherlands Evaluating the Validity of (Vicarious) Self-Reports of Deviant Behavior Using Vignette Analyses Bldg. A Rm. 305 Stefanie Eifler, Germany Statistical and Technical Methods for the Detection and Prevention on Interviewer Fraud in Data Collection Joe Eyerman; Paul Biemer ; Craiig Hill; Joe Murphy; Chris Ellis, UK Research Designs for Fraud Studies: How to Chose the Optimal Research Design Gerty Lensvelt-Mulders; Peter van der Heijden, The Netherlands 16:10 17:45 Preventing, Diagnosing and Analyzing Missing Data organizers: Joop J. Hox; Edith D. de Leeuw, The Netherlands Mixed Missing: Mode Elemental Structures Antonio Alaminos, Spain Social Distance, Interviewers Perception of Rapport and the Types of Invalid Response Su-hao Tu; Pei-Shan Liao, Taiwan Socio-economic Status of Survey Respondents with Missing Household Income Data Michael Wood, USA Statistical Editing and Imputation for Social Data Frank van den Eijkhof, The Netherlands 24 Tuesday 15:00 - Improving Survey Methodology: The European Social Survey Organizer: Ineke Stoop, The Netherlands 17:45 Part I Influence of Media Reported Events on Attitudes and Opinions 15:00 - Jurjen Iedema; Ineke Stoop, The Netherlands 16:10 Bldg. E Rm. 151 Part II 16:10 17:45 Modeling Events for ESS: Toward the Creation of an Autonomous Tool for Survey Research Theoni Stathopoulou, Greece Measuring Attitudes towards Immigration across Countries: Potential Problems of Functional Equivalence in the ESS Nina Rother, Germany Improving Survey Methodology: the ESS. Index and Measurement: The Cultural Borders of Meaning Antonio Alaminos, Spain Some Methods for Investigating Validity and Reliability of Questions in the European Social Survey A. J. Kutylowski, Norway 25 Tuesday 15:00 - Visual Design and Layout in Internet and Mail Surveys Organizer: Don A. Dillman, USA 17:45 A Comparison of Sliding Scales with Other Scale Types in Online Services Leonard R. Bayer; Randall K. Thomas, USA Bldg. A Rm. A Online Surveys - Does One Size Fit All? Dan Coates, USA Web Survey Design: Effect of Layout on Measurement Error Andy Peytchev, USA An Experimental Testing of Format Changes to Reduce Missing Data and Increase Cooperation in the Nielsen TV Diary Kenneth W. Steve; Mildred . Bennett, Paul J. Lavrakas; Jennie W. Lai, USA Achieving Usability in Establishment Surveys through the Application of Visual Design Principles Don A. Dillman; Arina Gersteva, USA 26 Tuesday 15:00 16:10 Is There a Fundamental Change in the Individual Shaping of the Life Course in Modernity? Or How Can We Do Research on Fundamental Transformations in Modernization? Organizer: Jens Zinn, UK Using Biographical Material to Research University Students' Identity: An Avenue for Research on Social Change Lupita Rodriguez Bulnes, UK Bldg. E Rm. 004 Experiential Glances at Transformations and Changes Occurring in Reflexive Modernization Life Ada Cattaneo, Italy Reflexive Modernization and Biotechnology: The Conflict of Biotechnology as a Conflict between the Logic of Industrial Society and the Logic of Reflexive Modernity Piet Sellke, Germany Achieving Usability in Establishment Surveys through the Application of Visual Design Principles Don A. Dillman; Arina Gersteva, USA 27 Wednesday 18th August 2004 General Program 9:00 11:00 11:00 11:30 11:30 13:00 13:00 14:30 14:30 15:40 15:40 16:00 16:00 17:30 Session 18:00 19:00 Business Meeting SISWO Coffee/Tea Session Lunch Session Tea Session 28 Wed. 9:00 - Studying Sensitive Topics in Qualitative Research Organizer: Hennie R. Boeije, The Netherlands 11:00 Minerva's Web: The Internet and Methodologies of Sensitive Subjects Indhu Rajagopal, Canada Bldg. E Rm. 010 Positioning Myself as a Researcher in a Politically Sensitive and Volatile Situation Chaya Possick, Israel Studying Decision-Making Issues at Home: The Nature of the Relation between Researcher and Participants María Elena Ramos Tovar, Mexico Some Functions of Probing Tactics in Interviews on Friendship Gerben Moerman, The Netherlands Processes in Research into Sensitive Topics Pat Cox, UK 29 Wed. 9:00 11:00 Bldg. A Rm. 305 Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data ( Part II ) Organizers: Ruth Sautu, Argentina ; Silvana Figueroa; Jochen Dreher, Germany People's Representations on Corrupt Practices and the Construction of a New Moral Order: In-Depth Interviews to Middle-Class Residents of Buenos Aires Ruth Sautu; Paula Boniolo; Ignacia Perugorría, Argentina Social Mimesis: a Study about Cultural Performances Vincenzo Romania, Italy Impact of Interviewer Attitudes and Socio-Economic Status in an Inter-Ethnic Survey Peter Pillók; Dávid Simon, Hungary Video-based Analysis of Social Situations: Comparing the Analytic Range of Ethnomethodology and Genre Analysis Dirk vom Lehn, UK ; Bernt Schnettler, Germany Giving Visibility to Hidden Rural Localities: Contributions of an Ethnographic Research Fernando Ilídio Ferreira, Portugal The Emergence of New Identities in Argentine Armed Forces: a Life Course Research. Alejandra Navarro, Argentina 30 Wed. 9:00 - Approaches to Using Mixed Methods Organizers: Angela Dale; Lynn Jamieson, UK 11:00 Integrating Quantitative and Qualitative Research: How is it done? Alan Bryman, UK Bldg. A Rm. 102 Multi-Methods, Triangulation and Integration - Different Goals or Just Different Terms? Jo Moran-Ellis; Victoria Alexander; Ann Cronin; Mary Dickinson; Jane Fielding; Judith Sleney; Hilary Thomas, UK Qualitative Methodology as a Parenthesis for Quantitative Methodology Mihai Pascaru, Romania Multi-Semiotic Ethnography Bella Dicks, UK A Delicate Dialectic: Mixing Qualitative and Quantitative Methods in Two Studies for US Federal Government Agencies Susan G. Berkowitz; Cynthia S. Robins, USA 31 Wed. 9:00 - Multilevel Analysis ( Part I ) Organizer: Tom A. B. Snijders, The Netherlands 11:00 Frequentist MCMC Estimation Methods for Multilevel Logistic Regression Carlos Coimbra; Tom A.B.Snijders, The Netherlands Bldg. E Rm. 150 The Use of Internal Pilot Studies to Derive Powerful and CostEfficient Designs for Studies with Nested Data M. Moerbeek, The Netherlands Performance of Likelihood-Based Estimation Methods for Multilevel Binary Regression Models Marc Callens and Christophe Croux, Belgium Outliers and Multilevel Models John F. Bell; Eva Malacova, UK On the Relative Efficiency of Unequal Cluster Sizes in Multilevel Intervention Studies L. Kotova; G.J.P. van Breukelen; M.J.J.M. Candel; M.P.F. Berger, The Netherlands 32 Wed. 9:00 - The Applicability of Agent-Based Social Simulations ( Part II ) Organizers: Klaus G. Troitzsch, Germany ; Nigel Gilbert, UK 11:00 Using Multi Agent Systems and Role-Playing Games to Simulate Water Management in Peri-urban Catchments Diana F. Adamatti; Jaime S. Sichman, Brazil Bldg. E Rm. 007 Lagoon, Agents and Kava. A Companion Modelling Experience in the Pacific P. Perez; C. Lepage; P. D'Aquino, France ; A. Dray; I. White, Australia Agent-Based Simulations bacing Use of Role-Playiing Games as Dialogue Support Tools: Teaching from Experience William's Daré; Olivier Barreteau, France Actualities of Social Representation: Simulation on Diffusion Processes of SARS Representation Kazuhiko Shibuya, Japan 33 Wed. 9:00 - New Directions in Political (and Social) Methodology ( Part I ) Organizer: Nathaniel Beck, USA 11:00 Time-Series--Cross-Section Data Nathaniel Beck; Jonathan Katz, USA Bldg. E Rm. 003 New Directions in Spatial Analysis: Space is more than Geography Nathaniel Beck, USA ; Kristian Gleditsch, Norway American Public Opinion in the 1930s and 1940s: the Analysis of Quota-Controlled Sample Survey Data Adam Berinsky, USA Estimating Spatial Autoregressive Coefficients in Spatial Lag Models Robert J. Franzese Jr.; Jude C. Hays, USA 34 Wed. 9:00 - Nonresponse ( Part I ) Organizers: Claire Durand; John Goyder, Canada 11:00 Recruitment Procedures and Response Rate: A Swiss Experiment N. Schoebi; D. Joye, Switzerland Bldg. A Rm. 306 Efficacy of Incentives in Increasing Response Rates M. Fahimi; R. Whitmore; J. Chromy; M. Calahan; L. Zimbler, USA New Evidence on Amount of Payment on Prepaid Incentives on a Mailed Questionnaire J. Goyder; G. Martinelli; S. Brown, Canada The Drop-off Pick-up Method: An Alternative Approach to Reducing Nonresponse in Mail Surveys A. Ross-Davis; S. R. Broussard, USA A Plea for the Tailored Use of Respondent Incentives G. Nicolaas; N. Stratford, UK 35 Wed. 9:00 11:00 Bldg. E Rm. 008 Latent Growth Curve Structural Equation Models, Autoregressive and Hybrid Models Organizers: Jost Reinecke; Peter Schmidt, Germany Combining Autoregressive and Semiparametric Mixture Models: Investigating the Selection-Facilitation Effect of Peers on Antisocial Behaviors Eric Lacourse; Richard E. Tremblay, Canada ; Daniel Nagin, USA Pay off and Impact of Varying the Intercept in Latent Growth Curve Modeling Manuel C. Voelkle, Germany Latent Growth Curve and Interaction Modeling: The Case of Travel Mode Choice Fan Yang Wallentin; Peter Schmidt; Eldad Davidov Anomia, Left Right Orientation and Group Related Enmity Steffen Kühnel; Peter Schmidt; Elmar Schlüter 36 Wed. 9:00 - Comparable Data across Countries Organizers: Willem Saris; Ineke Stoop, The Netherlands 11:00 Pursuing Equivalence in Cross-National Surveys Roger Jowell Bldg. E Rm. 151 Comparability across Countries of Responses in the ESS Willem E. Saris, The Netherlands Trapped in Translation? ESS Translation Protocols Provide a Key Janet Harkness, Germany Sampling for the European Social Survey Seppo Laaksonen, Finland ; Siegfried Gabler; Sabine Häder, Germany Peter Lynn, UK Data Quality Assessment in ESS Round 1: Between Wishes and Reality Jaak Billiet; Michel Philippens, Belgium Context, Events and Attitudes Jurjen Iedema; Ineke Stoop, The Netherlands 37 Wed. 9:00 11:00 Methodological Foundations of Sociological Analysis of Social Sphere Part I Organizers: Vassiliy Zhukov; Yelena Meshkova; Galina Ossadchaya, Russia Designing of Disablement in the Institutional Field Vladimir N. Petrov; Inna B. Kantemirova, Russia Bldg. E Rm. 004 Theories and Methods in Cross-National Youth Research1 Ken Roberts, UK Tasks of Historiography of Social Sphere Sociology Viktor Gorodyanenko, Ukrain Methodology and Methods of Comparative social research Vassili I. Joukov, Russia 38 Wed. 11:30 - Discoursive Ethnography Organizers: GiamPetro Gobo, Italy ; Anne Ryen, Norway 13:00 Using Conversation Analysis to Test Questionnaires R. J. W. M. Vis-Visschers, The Netherlands Bldg. A Rm. 108 For "she who knows who she is…" Combining Ethnography and Conversation Analysis in the Study of Internet Newsgroups A. Vayreda; F. Nuñez; E. Ardevol, Spain ; Ch. Antaki, UK Talk of Stakeholders, Talk of Players. Goffmans Contributions to the Use of an RPG as a Social Investigation Tool William's Daré, France The Methodological Discourse about the Complex Discourse of "The Childs best Interest" Turid Midjo, Norway 39 Wed. 11:30 13:00 Recent Developments in Ethnomethodological and ConversationAnalytic Research ( Part I ) Organizer: Paul ten Have, The Netherlands Technomethodology Andy Crabtree, UK Bldg. E Rm. 010 Exploring the Possibilities for the Use of Ethnographic, Contextual Information in Ethnomethodological/ Conversation of Analytic Studies Bregje C. de Kok, UK Shifting the Focus: The Impact of Audiovisual Technology on Empirical Research Antonia L. Krummheuer, Germany Order in Disorder: Rule-Governed Actions as Sociological Phenomena Alain Bovet, Switzerland ; Philippe Sormani, UK 40 Wed. 11:30 13:00 Bldg. A Rm. 305 Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data Organizers: Ruth Sautu, Argentina ; Silvana Figueroa; Jochen Dreher, Germany People's Perceptions on the Construction of Collective Goals: Interviews in Street Meetings of Collective Protest Luciano Brom; Pablo Dalle; Rodolfo Elbert, Argentina Images of the Argentine Ruling Class: Focus Groups using Mass Media Photographs as Stimulus Valeria Maidana; Gabriela Medin; María Alejandra Otamendi, Argentina Evaluation of the Internet use - A Call for Mixing Method Approach Boris Kragelj, Slovenia 41 Wed. 11:30 - Interaction Analysis and Data Quality of Survey Interviews Organizers: J. van der Zouwen; J. H. Smit, The Netherlands 13:00 Conversational and Formal Questions in Survey Interviews Yfke Ongena; Wil Dijkstra; Stasja Draisma, The Netherlands Bldg. A Rm. 102 The Effectiveness of Repair Strategies of Interviewers in Survey Interviews J. van der Zouwen; J. H. Smit, The Netherlands Opening the Black Box in Survey Research. Data Collection as an Ongoing Process of Total Quality Management. John Lievens; Els Bundervoet; Hans Waege, Belgium 42 Wed. 11:30 - Multilevel Analysis ( Part II ) Organizer: Tom A. B. Snijders, The Netherlands 13:00 Multilevel Multiprocess Modelling of Partnership Transitions and Fertility in Britain Bldg. E Fiona Steele; Constantinos Kallis; Harvey Goldstein; Heather Joshi, UK Rm. 150 The Concept of 'Social Level' and how to assess it Pieter van den Eeden, The Netherlands The Hausman Test of Random Effects Specifications A. Fielding, UK The Problem of Time Dependent Explanatory Variables at the Context Level in Discrete Time Multilevel Event History Analysis. Michael Windzio, Germany 43 Wed. 11:30 - The Applicability of Agent-Based Social Simulation ( Part III ) Organizers: Klaus G. Troitzsch, Germany ; Nigel Gilbert, UK 13:00 MASON: A JAVA Multi-Agent Simulation Library S. Luke; G. C. Balan; L. Panait; C. Cioffi-Revilla; S. Paus, USA Bldg. E Rm. 007 Simma: A Toolkit for Participatory Business Simulations X. Noria ,G. Peffer, Spain ; Nigel Gilbert, UK The Political Actor Simulator (PAS) - The Simulation of a Foreign Policy Crisis during the Peacekeeping Mission in Bosnia 1995 Markus Bresinsky, Germany Using GIS and Multi-agent Systems for Interactive Design of Policy in Natural Resource Management Rhodora M. Gonzalez, The Philippines Multiple Equilibria Regulation Model in Cellular Automata Topology Ioannis D. Katerelos; Andreas Koulouris, Greece 44 Wed. 11:30 - New Directions in Political (and Social) Methodology ( Part II ) Organizer: Nathaniel Beck, USA 13:00 Biases in the Effects of Parental Background on Voting Jannes de Vries; Nan Dirk de Graaf; Rob Eisinga, The Netherlands Bldg. E Rm. 003 Event Data Based Network Analysis (EDNA) Thomas Widmer; Vera Troeger, Switzerland Matching Methods for the Causal Analysis of Observational Data Markus Gangl; Thomas A. DiPrete, Germany 45 Wed. 11:30 - Aided Recall Techniques in Survey Interviews Organizer: Wander van der Vaart, The Netherlands 13:00 Improving the Quality of Retrospective Reports: Calendar Interviewing Methodologies Robert F. Belli, USA Bldg. A Rm. 306 Collecting Event History Data about Work Episodes Retrospectively: What Recall Errors Occur and how can we Prevent them? Maike Reimer, Germany Influence Dating Formats on the Telescoping Effect and the Distribution of Autobiographical Memory S. M. J. Janssen; A. G. Chessa; J. M. J. Murre, The Netherlands Timeline Methods and Related Aided Recall Applications in Social Sciences and Health Studies: a Review. Wander van der Vaart, The Netherlands 46 Wed. 11:30 - Multidimensional Scaling and Facet Theory Organizers: Patrick Groenen, The Netherlands ; Ingwer Borg, Germany 13:00 What is not what in Facet Theory Wijbrandt H. van Schuur, The Netherlands Bldg. E Rm. 008 Questionnaire Design and Data Analysis using Facet Approach: The Case of the International Census on Attitudes toward Languages Kazufumi Manabe, Japan New Type of Psychodiagnostical Test: Situational Operational Grid Anna S. Naumenko, Russia Explaining the Structure of Values of an Ideal Organization by Schwartz's Universal Value Theory Ingwer Borg; Wolfgang Bilsky, Germany ; Patrick J.F. Groenen; Karen A. Jehn, The Netherlands 47 Wed. 11:30 13:00 Methodological Issues in Designing and Implementing CrossCultural Surveys ( Part I ) Organizers: Janet Harkness, Germany ; Edith D. de Leeuw, The Netherlands A Process Quality Approach to the Testing and Evaluation of Cross-Cultural Surveys Bldg. E Rm. 151 Johnny Blair; Linda Piccini, USA Language, Usage, and Cultural Context: Challenges to Functional Equivalence in Questions Michael Braun; Janet Harkness, Germany Developing a Low-Cost Technique for Parallel Cross-Cultural Instrument Development: The Question Appraisal System (QAS-04) Elizabeth Dean; Rachel Caspar; Georgina McAvinchey; Rosanna Quiroz; Leticia Reed, USA Test of Anchoring Vignettes in a Low Income Population Patricia M. Gallagher; Floyd J. Fowler, USA 48 Wed. 11:30 - Miscellaneous Organizer: Henk Kleijer 13:00 Using E-mail as a Site for Constructing Narratives through Interview Nalita James; Hugh Busher, UK Bldg. A Rm. 303 Getting it Right: Designing an Evaluation of an INGO led HIV/AIDS Prevention Programme in Resource Poor Settings Fiona Samuels; Sam McPherson; Pepukai Chikukwa, UK Cyberethnography: New Fields of Inquiry or are we simply Re-Cycling Old Methodology? Denise Carter, UK 49 Wed. 11:30 - Feminism, Methodology and Methods ( Part I ) Organizer: Sharlene Hesse-Biber, USA 13:00 Critical Rationalism as a Tool for Feminist Comparative Historical Sociology Judith Buber Agassi, Israel Bldg. E Rm. 004 The Clear Feminist Rejection of an Obscure Positivism Maria Ignez Silveira Paulilo, Brazil Entering Sites; Entering Ethnography Anastasia Norton, USA Power Tools: A Feminist Approach to Large Sample Quantitative Data Analysis Allison J. Tracy, USA 50 Wed. 14:30 15:40 Research Methodology and the Challenges of Social Diversity ( Part I ) Organizers: Carole Truman, UK ; Donna Mertens, USA The Hermit's Knowledge Isacco Turina, Italy Bldg. E Rm. 010 Philosophic Sagacity: Another Way of Doing Research Bagele Chilisa, Botswana Irrationality and Violence in the Postmodernity. Proposal of a Model of Analysis of the Social Reality under the Perspective of a Transdisciplinary Method Augusto Renato Pérez Mayo; Maricela Guzmán Cáceres, Mexico 51 Wed. 14:30 15:40 Interrogating Conventionality in Qualitative Research. Bricolage and Ingenuity in Interdisciplinary Fieldwork Organizer: Marsha Henry, UK The Tape Recorder as a Major Player in the In-Depth Interview Tom Slater, UK Bldg. A Rm. 305 If the Shoe Fits: Authority, Authenticity and Agency in Feminist Diasporic Research Marsha Henry, UK Colonial Contexts and Post-colonial Understandings: Researching on Women in the Indian Nationalist Movement Suruchi Thapar-Bjorkert, UK Insider/Outsider Debates: Reflexivity in Military Research Paul Higate, UK 52 Wed. 14:30 15:40 Estimating Informed Opinion: Advances in the Theory and Measurement Part I Organizer: Patrick Sturgis, UK CAPI-based Information Intervention (CIi): A New Way of Estimating Informed Opinion Bldg. A Rm. 102 Patrick Sturgis; Helen Cooper, UK Toward Informed Opinions by the Means of Deliberation Kasper M. Hansen, Denmark Gauging Informed Opinion amongst Young Citizens Marcus Longley; Christian Thomas; Rachel Iredale; Anita Shaw; Angela Burgess, UK 53 Wed. 14:30 - Measurement (In)Variance in Longitudinal Research ( Part I ) Organizers: Han Oud; Reinoud Stoel, The Netherlands 15:40 Practical Issues in the Assessment of Measurement Invariance and why Invariance does not Imply Measurement Reinoud D. Stoel, The Netherlands Bldg. E Rm. 150 Measurement (In)Variance in SEM State Space Modelling of Panel Data Johan H.L. Oud, The Netherlands Types of Change in Self-Report Data: Definition, Interpretation and Operationalization Frans J. Oort, The Netherlands 54 Wed. 14:30 15:40 Informetric and Text Based Procedures for Information Retrieval and Analysis, e.g. in the Context of Digital Libraries ( Part I ) Organizers: H. Peter Ohly; Maximilian Stempfhuber, Germany Bibliometric Mining Peter Ohly, Germany Bldg. E Rm. 007 Automating the Analysis of Co-Words in Contexts: Measuring the Meaning of "Stem-Cells" and "Frankenfoods" in Different Domains Iina Hellsten; Loet Leydesdorff, The Netherlands Mining Document Contents in Order to Analyse a Scientific Domain Josiane Mothe; Bernard Dousset; Désiré Kompaore, France 55 Wed. 14:30 - Recent Developments in Nonparametric Item Response Theory ( Part I ) Organizers: Klaas Sijtsma; L. Andries van der Ark, The Netherlands 15:40 Detecting Discontinuity On a Cognitive Developmental Ability: A Comparison of the Binomial Mixture Model and the Latent Class Model. Bldg. E Rm. 003 Samantha Bouwmeester, The Netherlands A Markov Chain Approach for Ranking Individuals From Item Responses Matthew S. Johnson, USA A Bayesian Framework For the Monotone Homogeneity Model of Non-Parametric Item Response Theory George Karabatsos, USA 56 Wed. 14:30 - Data Collection Organizers: Marek Fuchs, Katja Lozar Manfreda, Slovenia 15:40 Patterns of mobile phone usage and their impact on participation in mobile phone surveys Vasja Vehovar Bldg. A Rm. 306 Comparing anonymous and non-anonymous research methods asking sensitive and socially desirable questions Hans-Ullrich Mühlenfeld Improving Questionnaire Design through the provision of electronic access to large scale survey questionnaires in the UK Martin Bulmer; Harshad Keval; Julie Lamb Effects of Survey Sponsorship and Mode of Administration on Respondents’ Answers about their Racial Attitudes Volker Stocké 57 Wed. 14:30 - Correspondence Analysis ( Part I ) Organizer: Jörg Blasius, Germany 15:40 Nordic Business Researchers' Use of Research Methodology: An Assessment of Researchers in Denmark, Finland, Norway and Sweden Bldg. E Rm. 008 Tor Korneliussen, Norway Using Combined Methods to Develop Sociological Concepts. An Exploration of the Concept of Social Capital using Correspondence Analysis and Case Studies Research Francesca Odella, Italy The Analysis of Lifestyle Groups and Social Milieus in Germany Using Cluster and Correspondence Analysis. Ingvill C. Mochmann; Yasmin El-Menouar, Germany International Comparison with Correspondence Analysis. Attitudes towards Same Sex Partnership as an Empirical Example Maria Rohlinger, Germany 58 Wed. 14:30 15:40 Methodological Issues in Designing and Implementing CrossCultural Surveys ( Part II ) Organizers: Janet Harkness, Germany ; Edith D. de Leeuw, The Netherlands Disambiguating Culture: The Cultural Element of Survey Response in Perspective Bldg. E Rm. 151 Eleanor Gerber, USA Question Design Features and Cultural Variability in Comprehension Timothy Johnson; Young Ik Cho; Allyson Holbrook; Diane O’Rourke; Richard Warnecke, USA Socio-Cultural Factors in the Question-Response Process: A Comparative Coding Analysis of Latino and Anglo Cognitive Interviews Kristen Miller; Gordon Willis; Concepcion Trevino Eason, USA What are we talking about ?Interviewing and interpretation for a biographical survey in Sub-Saharan Africa. Cross-cultural/cross-national survey research session Martine Quaglia, France 59 Wed. 14:30 15:40 The Internet and Social Research: New Opportunities, New Challenges Part I Organizers: Jonathan Elford; Dick Wiggins, UK Methodological and Ethical Problems in Analysis of Web Site Authors Gregor Petrič, Slovenia Bldg. A Rm. 303 Gender Specific Analysis of Virtual Communication in Panel Online Focus Groups Brigitte Salfinger; Verena Paul, Austria If you Build it, they will come - or will they? A Discussion on the Acceptability of the Web as a Means to Respond to Surveys Zoë Dowling; Martin Bulmer, UK 60 Wed. 14:30 - Miscellaneous Organizer: Harry Ganzeboom 15:40 Counting, Measuring, Meaning and Reality: Issues in Sociological Research Anne Seitz, Australia Bldg. E Rm. 004 Socio-Cultural Foundations of Gender Parity Yelena G. Meshkova, Russia 61 Wed. 16:00 - The Achievements and Challenges of Qualitative Computing Organizer: Lyn Richards, Australia 17:30 Qualitative Computing and Theory Building: A Methodological Debate César A. Cisneros Puebla, Mexico Bldg. A Rm. 108 Why Qualitative Computing is radically changing Qualitative Research Tom Richards, Australia Adopting and Adapting to Qualitative Software: Social And Methodological Considerations from the Periphery Diógenes Carvajal, Colombia A Context-Oriented Approach to Design of Computer-Assisted Qualitative Data Analysis Software Gennady Kanyguine, Russia 62 Wed. 16:00 - Action Research Organizer: Nancy Andes 17:30 The Catalyse Methodology and the Restoration of the Inquiry Results in the Action Research. Several Recent Experiments in the Rural Romanian Areas Bldg. E Rm. 010 Mihai Pascaru, Romania Russian Drug-Addiction - NGO - Action Research : Meeting on the Crossroad Galina Saganenko, Russia Research on Childhood and Children as Researchers: The Participatory Methodologies on Social Worlds of Children Natália Fernandes Soares; Manuel Jacinto Sarmento; Catarina Tomás, Portugal 63 Wed. 16:00 17:30 Recent Developments in Ethnomethodological and ConversationAnalytic Research ( Part II ) Organizer: Paul ten Have, The Netherlands Tracking Psychotherapeutic Agendas with Conversation Analysis Charles Antaki; Ivan van Leudar; Rebecca Barnes, UK Bldg. A Rm. 305 The Production of a live TV-interview through Mediated Interaction Mathias Broth, Sweden Talk about Talk: The Methodological Status of Teacher Interviews on Classroom Interaction Tom Koole, The Netherlands What's in a Name?: Revealing Social Identity Erica Lea Zimmerman, USA 64 Wed. 16:00 17:30 Bldg. A Rm. 102 Estimating Informed Opinion: Advances in the Theory and Measurement Part II Organizer: Patrick Sturgis, UK Do People really care about GM Food Risks? A latent StateTrait Model to differentiate 'Attitudes' from 'Opinions' amongst the UK Public Nick Allum, UK Self-reported Familiarity and Coherence of Opinions on Biotechnology Sally Stares, UK Measuring Nonattitudes through Response Latencies Piet Sellke, Germany 65 Wed. 16:00 17:30 Measurement (In)Variance in Longitudinal Research ( Part II ) Organizers: Han Oud; Reinoud Stoel, The Netherlands Impact of Direction of Answering Categories on Response Behavior Dagmar Krebs, Germany Bldg. E Rm. 150 Testing Measurement Invariance with Respect to Time, Group, and Latent Gitta Lubke, USA Individual Differences in Factor Loadings and the Fit of the Standard Fac Henk Kelderman, Peter C. M. Molenaar, The Netherlands 66 Wed. 16:00 17:30 Informetric and Text Based Procedures for Information Retrieval and Analysis, e.g. in the Context of Digital Libraries ( Part II ) Organizers: H. Peter Ohly; Maximilian Stempfhuber, Germany Comparing a Knowledge-Based Economy: In the Case of South Korea and the Netherlands Bldg. E Rm. 007 Han Woo Park; Heung Deug Hong; Sung Jo Hong, South Korea Loet Leydesdorff, The Netherlands Graph-Based Representations of Motion Pictures: An Attempt to Discover Statistical Features of Aesthetic Merit Gregory H. Leazer; Jonathan Furner; Rachel Napper, USA Using Search Engines and Web Crawlers in Social Science Research Mike Thelwall, UK 67 Wed. 16:00 17:30 Recent Developments in Nonparametric Item Response Theory ( Part II ) Organizers: Klaas Sijtsma; L. Andries van der Ark, The Netherlands Multiple Imputation of Item Scores in Test and Questionnaire Data, and the Influence on Psychometric Results Joost van Ginkel, The Netherlands Bldg. E Rm. 003 A Nonparametric IRT Model for the Circumplex: Generalization to Polytomous Data Wijbrandt H. van Schuur, The Netherlands On Ordering Properties of Classical Optimal Scaling Matthijs J. Warrens, The Netherlands 68 Wed. 16:00 17:30 Classification and Visualisation Organizer: Simona Balbi, Italy Cluster Based Conjoint Analysis Carlo Natale Lauro; Germana Scepi; Giuseppe Giordano, Italy Bldg. E Rm. 008 The Typological Non Symmetrical Correspondence Analysis: Theory and Application for Social Research Marilena Fucili and Danilo Leone, Italy Some Different Typologies of Employment in Italy M. G. Grassia; F. Pintaldi; L. Quattrociocchi, Italy 69 Wed. 16:00 17:30 Issues of Cross-National Comparative Research Organizer: Harry B. G. Ganzeboom, The Netherlands How to Grasp Nations in International Survey Comparisons Markus Hadler, Austria Bldg. E Rm. 151 (P)raising Cross-National Comparability: Digital Divide Benchmarks Based on Time Distance Method Vesna Dolničar; Vasja Vehovar; Pavle Sicherl, Slovenia Measuring Attitudes towards Immigration across Countries: Potential Problems of Functional Equivalence in the ESS Nina Rother, Germany An Empirical Enquiry into Students' Socio-Economic Background, Costs and Financing of Studies, and Future Career Aspirations in Higher Education in China: Issues of Methodology Lihong Huang, Sweden 70 Wed. 16:00 17:30 The Internet and Social Research: New Opportunities, New Challenges Part II Organizers: Jonathan Elford; Dick Wiggins, UK Dropping Out of an Online Sex Survey: when and why? Bldg. A Alison Evans; Dick Wiggins; Graham Bolding; Mark Davis; Jonathan Elford, Uk Rm. 303 Validity of Data and Representativeness of Sample in Internet Survey: University Surveys in the United States Hisako Matsuo; Kevin McIntyre; Terry Tomazic; Barry Katz, USA Reinoud Bosch, Italy Actors Invisibility: About the Necessity to Re-Think Internet Research Maren Lübcke; Rasco Perschke, Germany 71 Thursday 19th August 2004 General Program 9:00 10:30 10:30 11:00 11:00 12:30 12:30 14:00 Session 14:00 15:00 Plenary Speech: Jan de Leeuw, The Netherlands Subject: Description, Prediction, Inference Bldg. A; Room A 15:00 15:30 15:30 17:30 Tea 19:00 23:00 Conference Dinner Coffee/Tea Session Lunch Session 72 Thu. 9:00 10:30 Computer Aided Content Analysis - Research, Methods and its Future Part I Organizer: Harald Klein, Germany Content Analysis of Short, Structured Texts: The Need for Multifaceted Strategies Bldg. A Rm. 306 Harold W. Bashor, France Extraction of Semantic Information: New Models and old Thesauri Jan Kleinnijenhuis; Wouter H. van Atteveldt; Ivar Vermeulen, The Netherlands Why People Reject Super-fair Offers. Evaluating Video Experiments on Ultimatum Bargaining Run in PR China Heike Hennig-Schmidt; Chaoliang Yang, Germany ; Zhu-Yu Li, China 73 Thu. 9:00 10:30 Evaluating Cognitive Methods to the Pre-test Questionnaires Organizers: Ger Snijkers, The Netherlands ; Gordon Willis, USA Examining Expert Reviews as a Pretest Method Terry DeMaio; Ashley Landreth, USA Bldg. A Rm. 108 Examining Think Aloud as a Pretest Method: The State of the Art of think aloud in Cognitive Psychology Tony Hak, The Netherlands Toward the Accumulation of Cognitive Interview Findings: Considerations in the Design of a Tested-Question Database Paul Beatty; Kristen Miller, USA Looking for Trouble and Finding It, Sometimes: Exploring Relationships between Pre-survey and Post-survey Evaluation Data James L. Esposito, USA 74 Thu. 9:00 10:30 Miscellaneous Organizer: Henk Kleijer Funding of Social Research in Pakistan S. Hamidullah, Pakistan Bldg. E Rm. 010 Children as Respondents: Developing Questionnaires for Children Edith de Leeuw; Natacha Borgers; Astrid Smits, The Netherlands Feminist Method and Methodology and the Struggle to define Social Inquiry and Epistemology. The Logic behind the Measure. Ronald Marshall, Trinidad and Tobago Disambiguating Culture: The Cultural Element of Survey Response in Perspective Eleanor Gerber, USA 75 Thu. Telephone Survey Methodology in the 21st Century ( Part I ) 9:00 - Organizer: Edith D. de Leeuw, The Netherlands 10:30 Finding (and Listing) the Unlisted: A Strategy for Achieving a Listed Sample's Cost Savings without Sacrificing Coverage Anthony Salvanto; Kathleen Frankovic, USA Bldg. A Rm. 102 Where Can I Call You? The "Mobile Revolution" and its Impact on Survey Research and Coverage Error: discussing the Italian Case Mario Callegaro, USA ; Teresio Poggio, Italy Mobile Phone Surveys in Hong Kong: Methodological Issues and Comparisons with Conventional Phone Surveys Liam K P Lau, Hong Kong Occurrence and Consequences of Multiple Telephone Lines, Answering Machines, and Voice Mail Chris Barnes; Alicia Burrington, USA Calculating Response Rates for Mobile Phone Surveys. A Proposal of a Modified AAPOR Standard and its Application to three Case Studies Mario Callegaro; Trent D. Buskirk; Linda Piekarski; Charlotte Steeh, USA Vesa Kuusela, Finland ; Vasja Vehovar, Slovenia The Influence of Advance Contacts on Response in Telephone Surveys: A Meta- Analysis Edith De Leeuw; Joop Hox; Elly Korendijk; Gerty Lensvelt-Mulders, The Netherlands 76 Thu. 9:00 10:30 Methods of Social Network Analysis ( Part I ) Organizers: Peter J. Carrington, Canada ; Anuska Ferligoj, Slovenia Playing God on a Budget: Understanding Large-Scale Social Structure from Sample Data and Random Graph Models of Interdependent Local Processes Bldg. E Rm. 150 Rick Grannis, USA New Specifications for Exponential Random Graph Models Tom A.B. Snijders, The Netherlands ; Mark S. Handcock, USA Philippa E. Pattison; Garry L. Robins, Australia New Developments in Estimation Methods for the P2 Mode Bonne J.H. Zijlstra; Marijtje A.J. van Duijn, The Netherlands 77 Thu. 9:00 10:30 Informetric and Text Based Procedures for Information Retrieval and Analysis, e.g. in the Context of Digital Libraries ( Part III ) Organizers: H. Peter Ohly; Maximilian Stempfhuber, Germany Digital Libraries and Heterogeneity Maximilian Stempfhuber, Germany Bldg. E Rm. 007 Searching and Browsing Multiple Subject Gateways in the Renardus Service Michael Day, UK Ontology Switching for the Social Sciences. Methods for the Underlying Corpus Analysis Robert Strötgen, Germany Bilateral Transfer Modules for the Conceptual Integration of Heterogeneous Document Bases in Content Analysis Xueying Zhang, Hong Kong 78 Thu. 9:00 - Graphical Modelling ( Part I ) Organizer: Nanny Wermuth 10:30 Some Aspects of the Analysis of Graphical Markov Models D.R. Cox, UK Bldg. E Rm. 003 Computational Aspects of Chain Graph Models with Mixed Response Variables Vanessa Didelez, UK Checking for a Confounder in Models Generated over Directed Acyclic Graphs Elena Stanghellini; Giovanni M. Marchetti, Italy 79 Thu. 9:00 - Sampling Methods ( Part I ) Organizer: Siegfried Gabler, Germany 10:30 Comparing Sampling Frames Renáta Németh; Ildikó Zakariás, Hungary Bldg. A Rm. 305 Methods for Achieving Equivalence of Samples in CrossNational Surveys Peter Lynn, UK ; Siegfried Gabler; Sabine Häder, Germany ; Seppo Laaksonen, Finland Design Effects in the Analysis of Longitudinal Survey Data Chris Skinner; Marcel de Toledo Vieira, UK Impact of the Geographical Size of Clusters on the Precision of Survey Estimates Kevin Pickering; Sarah Tipping 80 Thu. 9:00 - Improving Survey Methodology: The European Social Survey ( Part III ) Organizer: Ineke Stoop, The Netherlands 10:30 Quality Of Life in Europe: Comparing States using a Composite Index Jeroen Boelhouwer, The Netherlands Bldg. E Rm. 008 Analysis and Relation between Social Variables and Economic Variables in the European Social Survey and World Bank and UNDP Reports: the Social Capital and Economic Capital at the European Countries Juan Sebastián Fernández Prados; Jaime Andreu Abela; Antonia Lozano Díaz, Spain Geographical Variations in Electoral Participation: Evidence from a Multilevel Analysis of the European Social Survey Ed Fieldhouse; Mark Tranmer, UK European Value Map. Based on ESS Data Hans Bay, Denmark 81 Thu. 9:00 - Cross-National Harmonization of Socio-Demographic Variables ( Part I ) Organizers: Jürgen H. P. Hoffmeyer-Zlotnik; Christof Wolf, Germany 10:30 Obtaining Harmonized Demographic Information from European Surveys -Experiences of the German Federal Statistical Office form Input to Output Harmonization Bldg. E Rm. 151 Thomas Körner; Iris Meyer, Germany Harmonizing Background Variables in International Surveys Kirstine Kolsrud; Knut Kalgraff Skjåk, Norway Harmonization of Survey Data in the "International Social Survey Programme" (ISSP) Evi Scholz, Germany Measuring Religious Membership and Religiosity across Europe Christof Wolf, Germany 82 Thu. 9:00 - Internet Surveys ( Part I ) Organizers: Mick P. Couper; Vasja Vehovar 10:30 A Comparison of Ranking versus Rating in Online Surveys Randall K. Thomas Bldg. A Rm. A Behavioral Intention Measurement: International Findings Randall K. Thomas; George Terhanian; and Leonard R. Bayer Evaluating the Effectiveness of Visual Analog Scales: A Web Experiment Mick Couper; Eleanor Singer; Roger Tourangeau; Fred Conrad Don't Know and No Opinion Responses in Web Surveys Reg Baker; Fred Conrad; Mick Couper; Roger Tourangeau Design Issues in Collecting Data on Ego-centered Social Networks on the Web Gasper Koren; Valentina Hlebec; Vasja Vehovar; Katja Lozar Manfreda A Comparison of Two Web-Based Surveys: Static versus Dynamic Versions of the NAMCS Questionnaire Mirta Galesic; Jimmie Givens; Mick P. Couper; Roger Tourangeau 83 Thu. 9:00 - Complex Societal Problems ( Part I ) Organizer: Dorien J. DeTombe, The Netherlands 10:30 Statistical Evidence: Responsibilities And Burden Of Proof Frans A.J. Birrer, The Netherlands Bldg. E Rm. 004 The Geography of Decision-Making: Linking Attitudes, Knowledge, Management Activities and Forestland Characteristics in a Spatial Environment Shorna R. Broussard; Karen Cox, USA Causality in Complexity Dorien J. DeTombe, The Netherlands Interactive Mapping of Complex Policy Networks: A Quick Scan for Strategy Making Bert Enserink, The Netherlands 84 Thu. 11:00 12:30 Computer Aided Content Analysis - Research, Methods and its Future Part II Organizer: Harald Klein, Germany Measurement Instrument Choice for Tracking Email Communication Marguerite Kolar, Australia Bldg. A Rm. 306 Inductive versus Deductive Approach: Computer-assisted Content Analysis without Dictionary? Cornelia Zuell; Juliane Landmann, Germany Overview on Standardised Category Systems in Computer aided Content Analysis Harald Klein, Germany 85 Thu. 11:00 - Research Methodology and the Challenges of Social Diversity ( Part II ) Organizers: Carole Truman; Donna Mertens, USA 12:30 Researching Civic Participation within A Diverse Context: Conceptual and Methodological Opportunities and Challenges Kien Shan Lee, USA Bldg. A Rm. 108 Transformative Research and Dimensions of Diversity Donna M. Mertens, USA Culturally Engaged Research - A Specific Application of an Emerging Methodology Anthony J. Alberta; Bonny Beach; John Whiteshirt, USA 86 Thu. 11:00 12:30 Mixing Qualitative and Quantitative Methods - Empirical Designs and their Theoretical Implications ( Part I ) Organizer: Udo Kelle, Germany Qualitative and Quantitative Methods between Paradigm and Pragmatics - Mixing, Integration or Triangulation? Bldg. E Uwe Flick; Alice Salomon, Germany Rm. 010 Mixed Methods Sampling Considerations and Designs in Social Science Research Anthony J. Onwuegbuzie; Kathleen M. T. Collins, USA Why it is Necessary to Combine Qualitative and Quantitative Data and Methods to Examine the Thesis of Individualization? Jens Zinn, UK Stratification and Social Mechanisms: an empirical point of view Flavio Ceravolo; Laura Accornero, Italy Coherency Analysis for an Updating of Sociological Thought Mascia Ferri, Italy 87 Thu. 11:00 12:30 Telephone Survey Methodology in the 21st Century ( Part II ) Organizer: Edith D. de Leeuw, The Netherlands Interviewer Training: Responding to Changes in Telephone Survey Methodology Lisa R. Carley-Baxter, USA Bldg. A Rm. 102 An Experiment in Cognitive Training of Telephone Survey Interviewers Claire Durand; Marie-Eve Gagnon; Christine Doucet, Canada Sociolinguistic Parameters of Cross-cultural Variation in Telephone Surveys: Brian Kleiner; Yuling Pan, USA Data Quality of Surveys of Minority-Majority Attitudes in Divided Societies: A Comparison between Telephone and Face-to-Face Surveys among Arabs in Israel Galit Gordoni, Israel Can Multimode Approaches Salvage RDD for Public Health Surveillance? Michael W. Link; Ali Mokdad, USA Using Interactive Voice Response Technology to Capture Census Data Sarah E. Brady; Robin A. Pennington; James B. Treat, USA 88 Thu. 11:00 12:30 Panel Data Analysis Organizer: Uwe Engel, Germany Estimation of a two Equations Panel Model with Mixed Continuous and Ordered Categorical Response Variables in the Presence of Missing Data Bldg. E Rm. 150 Martin Spiess, Germany Comparing Fixed Effects and Covariance Structure Estimators Mette Ejrnaes; Anders Holm, Denmark Analyzing Individual and Collective Change of Attitudes Using Panel Data. The Example of Attitudes to Social Security in Germany Nina Baur; Christian Lahusen, Germany 89 Thu. 11:00 12:30 Non Linear Modelling ( Part II ) Organizer: Cor van Dijkum, The Netherlands The Dynamics of Bilingual Language Development in Immigrant Children Annemarie Peltzer-Karpf; Marion Griessler, Austria Bldg. E Rm. 007 Methodological Issues of Labour Market Analysis for Countries in Transition Mikhail Mikhalevich; Ludmilla Koshla, Ukrain Anticipation and Codification in Communication Systems: Towards a Simulation Model for Luhmann's Sociological Theory of Communication Loet Leydesdorff, The Netherlands Catastrophe Theory and Social Policy Making: A Nonlinear Model of the Early Pension Insurance Legislation in Western Europe Georg P. Mueller, Switzerland 90 Thu. 11:00 12:30 Graphical Modelling ( Part II ) Organizer: Nanny Wermuth An Application of Marginal Log-linear Models to Examine Changes in Social Mobility in Hungary During the Transition Period Renáta Németh, Hungary Bldg. E Rm. 003 Marginal Models for Multivariate Dependencies Tamas Rudas, Hungary ; Wicher Bergsma, The Netherlands The Use of Bayesian Networks for Imputation Paola Vicard; Marco Di Zio; Mauro Scanu, Italy 91 Thu. 11:00 12:30 Sampling Methods ( Part II) Organizer: Siegfried Gabler, Germany GenD - An Evolutionary Algorithm for Re-sampling in Survey Research Daria Croce; Cinzia Meraviglia, Italy Bldg. A Rm. 305 Sampling Process-Generated Data in order to Trace Social Change Nina Baur; Christian Lahusen, Germany Estimating Domestic Tourism Expenditure in Developing Economies: Lessons from India Rajesh Shukla; Pradeep Srivastava, India Sampling Undocumented Immigrants in Brussels via Snowball Sampling: Theoretical and Practical Considerations Mila Paspalanova, Belgium Using Adaptive Sampling Methods in the Analysis of Social Phenomena on the WWW Robert Ackland, Australia 92 Thu. 11:00 12:30 Improving Survey Methodology: The European Social Survey ( Part IV ) Organizer: Ineke Stoop, The Netherlands Improving Survey Methodology: the ESS. Index and Measurement: The Cultural Borders of Meaning Joop Hox; Edith de Leeuw, The Netherlands Bldg. E Rm. 008 Response Propensity and Data Max Bergman; Nicole Schoebi; Dominique Joye, Switzerland Modelling Interviewer-Effects in the European Social Survey Geert Loosveldt; Michel Philippens, Belgium 93 Thu. 11:00 12:30 Cross-National Harmonization of Socio-Demographic Variables ( Part II ) Organizers: Jürgen H. P. Hoffmeyer-Zlotnik; Christof Wolf, Germany Is there Another Way to View Race? Listening to Subjectivity in the Categorization of Ethnicity and Race among Hispanics in the United States. Bldg. E Rm. 151 Christine Horak; Ricardo Costa Gazel, USA Title: Ethnicity and the Comparative Analysis of Contemporary Survey Data Paul S. Lambert, UK Measuring Ethno-Poverty: Methodological Challenges Lilia Dimova, Bulgaria Measuring "Income" in Cross-National Research Uwe Warner, Luxembourg ; Juergen H.P. Hoffmeyer-Zlotnik, Germany 94 Thu. 11:00 12:30 Internet Surveys ( Part II ) Organizers: Mick P. Couper; Vasja Vehovar Computing methods and their impact on drop out Lars Kaczmirek; Wolfgang Neubarth; Michael Bosnjak; Wolfgang Bandilla, Germany Bldg. A Rm. A Effectiveness of Progress Indicators in Web Surveys Fred Conrad; Mick Couper; Roger Tourangeau; Andy Peytchev, USA Item Non Response and Dropouts in Web Surveys Yasemin El-Menouar; Jörg Blasius, Germany Effects of personalization on web survey response rates and data quality Dirk Heerwegh; Tim Vanhove; Geert Loosveldt; Koen Matthijs, Belgium 95 Thu. 11:00 12:30 Complex Societal Problems ( Part II ) Organizer: Dorien J. DeTombe, The Netherlands Application of Fuzzy - based Methods for Urbanization Problem Ludmilla Koshlai; Mikhail Mikhalevich, Ukrain Bldg. E Rm. 004 Area Studies Approach: A Critique and Search for Different Methods Karori Singh, India Multi-Criterion Decision Modeling Within Rural Societal Policy The Case Of South - East Europe Sofija Adžić; Jasminka Adzic, Serbia & Montenegro Methodological Challenges Concerning a Comparison of the Complex Social Problem Marie Valentova, Luxembourg 96 Thu. 15:30 17:30 Computer Aided Content Analysis - Research, Methods and its Future Part III Organizer: Harald Klein, Germany TCA: General Purpose Software for Semantic and Thematic Text Analysis Carl W. Roberts, USA Bldg. A Rm. 306 Routinizing Frame Analysis through the Use of CAQDAS Thomas König, UK Qualitative Software 2004 and Beyond Raymond C. Maietta; Sharlene Hesse-Biber, USA 97 Thu. 15:30 17:30 Secondary Analysis of Qualitative Data Organizer: Louise Corti, UK Experiences with Secondary Use of Qualitative Data Material a Feasibility Study with German Qualitative Social Researchers Diane Opitz; Reiner Mauer, Germany Bldg. A Rm. 108 Secondary analysis of panel data: Establishing and Using a text databank Irena Medjedović; Andreas Witzel, Germany The Anonymization Process for Secondary Use of Qualitative Data D. Thomson; L. Bzdel; K. Golden-Biddle; T. Reay; C. Estabrooks, Canada Revisiting 'Classic' Qualitative Sociology Mike Savage, UK Strategies in Teaching Secondary Analysis of Qualitative Data Louise Corti; Libby Bishop, UK Reanalyzing Qualitative Interviews from Different Angles: The Merits of Sharing Qualitative Data Harry van den Berg, The Netherlands 98 Thu. 15:30 17:30 Mixing Qualitative and Quantitative Methods - Empirical Designs and their Theoretical Implications ( Part II ) Organizer: Udo Kelle, Germany Danger: Data Overload! Mixed methods in a cross-national study of care home staff Bldg. E Rm. 010 Ingrid Eyers, UK The German Research Foundation Project "Mobilzeit" as example for the Integration of Quantitative and Qualitative Methods Andrea Tina Booh; Siegrid Wieczorek, Germany Do Media Discourses Influence Opinions? Connecting Discourse Analysis of Newspaper Data with Time Series Analysis of Survey Data Nina Baur; Christian Lahusen, Germany The Conceptual Net of Internal Consistency and the Es of Praxiology in Human Inquiry Arne Collen, USA 99 Thu. 15:30 17:30 Combining Data from Different Sources ( Part I ) Organizer: Susanne Rässler, Germany Estimation with Population-Level Constraints on Two Survey Datasets: An Application to first Birth Probabilities by Education in Italy Bldg. E Rm. 150 Alessandra DeRose; Paola DiGiulio; Filomena Racioppi, Italy Mark S. Handcock; Michael S. Rendall, USA Combining Information from Multiple Surveys in the Analysis of Coarsened Longitudinal Data T. E. Raghunathan, USA A Comparison Record Linkage and Statistical Matching and Their Impact on Linear Regression Analysis Michael D. Larsen, USA Combining Separate Datasets in the OPUS Project J. Polak, UK ; Ch.D.R. Lindveld, The Netherlands Modelling the Construction of a Social Accounting Matrix in the Context of Statistical Matching Marcello D’Orazio; Marco Di Zio; Mauro Scanu, Italy Integrating Data for Innovative Analyses: The North Carolina Education Research Data Center Elizabeth Glennie, USA 100 Thu. 15:30 17:30 Time Series Analysis Organizer: Rainer Metz, Germany Comparison of Seasonal Variation of Births and Deaths in Urban and Rural Areas of Silesia (Poland) Zofia Mielecka-Kubien, Poland Bldg. E Rm. 007 Population at Risk (PAR) Rates and other Denominators used to Measure Unemployment Rates Ray Thomas; Hasan Al-Madfai, UK Auto-regressive Time Series Cross Section Analysis (TSCS) for Proportions Data with Random Coefficients. A Flexible Model and its Application in Regional Science Michael Windzio, Germany 101 Thu. 15:30 17:30 Miscellaneous Organizer: Karl van Meter A Strategy for the Selection of Weighting Variables Barry Schouten, The Netherlands Bldg. E Rm. 003 The Measurement of Duration of Studies in Higher Education Stephan Boes, Germany 102 Thu. 15:30 17:30 Nonresponse ( Part II ) Organizers: Claire Durand; John Goyder, Canada Measuring the Respondent's Attitude towards Surveys. Geert Loosveldt; Vicky Storms, Belgium Bldg. A Rm. 305 Survey Experience, Generalized Attitudes Towards Surveys and Respondents' Willingness to Answer Questions in Subsequent Surveys Volker Stocké, Germany Task Dedication, Task Alienation and What Data Collection Instruments Mean to Respondents Patricia A. Gwartney; Vikas Kumar Gumbhir; Anthony Leiserowitz, USA Does the Performance Matter ? A Multilevel Analysis of the Effects of Individual and Performance Characteristics on Response in Audience Research Henk Roose; John Lievens; Hans Waege, Belgium 103 Thu. 15:30 17:30 Correspondence Analysis ( Part II ) Organizer: Jörg Blasius, Germany The Use of Correspondence Analysis for Stratification Renato Salvatore, Italy Bldg. E Rm. 008 Measure versus Variable Duality in Correspondence Analysis Henry Rouanet; Brigitte Le Roux, France The Space of Elite Opinions. The Case of Norway Johs. Hjellbrekke; Olav Korsnes, Norway Processes of Social Differentiation in an Economic "Social Field" Lennart Rosenlund and Odd Einar Olsen, Norway 104 Thu. 15:30 17:30 Methodological Issues in Designing and Implementing CrossCultural Surveys ( Part III ) Organizers: Janet Harkness, Germany ; Edith D. de Leeuw, The Netherlands Challenges in Implementing a Multi-National Tobacco Use Survey Bldg. E Jutta Thornberry; Michele Bloch; Robert Goldenberg; Norman Goco; Don Jackson; Nancy Moss, USA Rm. 151 Attitude Scale Development in Countries which send Illegal Immigrants Kees van der Veer, The Netherlands ; Reidar Ommundsen, Norway ; Hao Van Le, Vietnam ; Krum Krumov, Bulgaria ; Knud S. Larsen, USA Behavior Coding of Multilingual Survey Interviews: What Can We Tell by Systematically Listening? Gordon Willis; Elaine Zahnd; Sherm Edwards; Stephanie Fry; David Grant; Nicole Lordi, USA Comparability of Attitude Measures towards Social Inequality. The Level of Equivalence in the ISSP 1999, its Implications on Further Analysis and the Presentation of Results Mag. Vlasta Zucha, Austria 105 Thu. 15:30 17:30 Internet Surveys ( Part III ) Organizers: Mick P. Couper; Vasja Vehovar Feasibility of a random sample Wolfgang Neubarth; Lars Kaczmirek; Michael Bosnjak; Wolfgang Bandilla, Germany Bldg. A Rm. A Self-selection as a Sampling Method in Web Surveys Jan B. Steffensen, Denmark Noncoverage and Nonresponse in a Web Survey Mick P. Couper; Arie Kapteyn; Matthias Schonlau; Joachim Winter Utilization of an Internet survey of work-family conflict among home-based teleworkers Dorit Ben-Baruch; Itzhak Harpaz, Israel Investigating the imagined online community Dávid Simon; Klára Benda, Hungary 106 Thu. 15:30 17:30 Methodological Foundations of Sociological Analysis of Social Sphere Part II Organizers: Vassiliy Zhukov; Yelena Meshkova; Galina Ossadchaya, Russia Cross-Cultural Dialogue: Methodology of Research Z.Kara, Turkey ; H.Meshkova, Russia Bldg. E Rm. 004 The Peculiarity of Treatment the Problem of Trafficking in Woman in Mass Media Ludmila Ykusheva, Belarussia Sociological Measurement of Social Protection Morozova Helen, Russia 107 Friday 20th August 2004 General Program 9:00 11:00 11:00 11:30 11:30 13:00 13:00 13:30 Session 13:30 End of Conference Coffee/Tea Session Closing Session Bldg. A; Room A 108 Friday 9:00 11:00 The Response Process in Establishment Surveys Organizer: Tony Hak, The Netherlands The Evaluation of Establishment Survey Questionnaires: Experiences from combining Quantitative and Qualitative Methods Deirdre Giesen, The Netherlands Bldg. A Rm. 306 Redesigning Statistics Canada's Annual Survey of Manufactures Overview of Respondent Testing using Cognitive Interview Techniques Kevin Roberts, Canada The Use Of Qualitative Interviewing To Determine The Acceptability And Take Up Of Web Data Collection Instruments for Mandatory Business Surveys Zoë Dowling, UK The Contact and Response Process in Business Surveys: Lessons from a Multimode Survey of Employers in the UK Peter Lynn; Emanuela Sala, UK Using Respondent Site Visits to Improve the Design of Establishment Surveys Stanley R. Freedman; Robert Rutchik, USA 109 Friday 9:00 11:00 Rethinking Professional Ethics Organizers: Liora Gvion; Diana Luzzatto, Israel Doing Ethics' In The Field: The Art And Politics Of Covert Research Dr David Calvey, UK Bldg. A Rm. 108 Methodological Issues Experienced in a Comparative Study of Children's Free Time Activities Sidsel Hadler-Olsen, Norway ; Doug Springate, UK The Hybridity of Borders in Ethnographic Field Work:The Case of Open-Ended Interview Suhad Haj-Yahia, OTR Making Private Issues Public: Ethnic dilemmas Versus Moral Obligations Liora Gvion; Diana Luzzatto, Israel 110 Friday 9:00 11:00 Analyzing Interviews ( Part I ) Organizers: Manfred Max Bergman; Veronique Mottier, Switzerland Strategies for Analysing In-depth Interviews of People with Disabilities in Northers Aegean, Greece D. Papageorgiou; T. Iosifides; M. Sidiropoulou, Greece Bldg. E Rm. 010 The Interpretation of Qualitative Data from a Critical Theoretical Framework Claire Wagner; David JF Maree, RSA Contradictories in the Interview: Public and Personal Olga B. Savinskaya, Russia The Interview as a Text: Analyzing In-depth-interviews in Social Science with the Conceptual Framework of Linguistics Veronika Sieglin, Germany Analysis of Interviews with Russian Entrepreneurs using Critical Discourse Analysis. Tatyana Kosyaeva, Czech Republic Analysis of Semi-Structured, In-depth Interview Data Kathleen W. Piercy, USA 111 Friday 9:00 11:00 Bldg. A Rm. 102 Response Latencies in Survey Research: Theory, Measurement and Predictive Power Organizers: Volker Stocké; Jochen Mayerl, Germany ; Stasja Draisma, The Netherlands Response Latency as an Indicator of Optimizing. A Study Comparing Job Applicants and Job Incumbents' Response Time on a Web Survey Mario Callegaro; Yongwei Yang; Dennison S. Bhola; Don A. Dillman, USA Question Wording Effects and Processes of Question Answering. A Different Way to Measure Response Latencies Bregje Holleman, The Netherlands Measuring Information Accessibility and Predicting Response Effects: The Relative Validity of Differently Transformed Response Latencies Volker Stocké, Germany Controlling the Baseline Speed of Respondents: An Empirical Evaluation of Data Treatment Methods of Response Latencies Jochen Mayerl, Germany Expressed Uncertainty in Question Answering: Response Latencies and (Para)Linguistic Expressions in Factual and Attitude Questions Stasja Draisma, The Netherlands 112 Friday 9:00 11:00 Combining Data from Different Sources ( Part II ) Organizer: Susanne Rässler, Germany Linking Job Episodes from Survey and Register Data: Specific Challenges and Empirical Results Maike Reimer; Ralf Künster, Germany Bldg. E Rm. 150 Non-Exact vs. Exact Matching with Applications to Wages Statistics Seppo Laaksonen, Finland Combining Datasets to Create a Synthetic Population: Sample Enumeration. Ch.D.R. Lindveld, The Netherlands ; J. Polak, UK Robustifying Imputation Model Estimation Florian Koller, Germany Linking public health data in Georgia, USA Mohamed G. Qayad; Emily Kahn, USA Two Stories about Youth Unemployment? - Combining Register Data and Self-Reported Data Hans Dietrich, Germany 113 Friday 9:00 11:00 The Statistical Estimation of Linear and Non-Linear Stochastic Differential Equations Organizer: Hermann Singer, Germany A Survey of Estimation Methods for Stochastic Differential Equations Hermann Singer, Germany Bldg. E Rm. 007 A Comparison of Different Procedures to Estimate the Damped Linear Occilator for Panel Data J.H.L. Oud, The Netherlands Multivariate Multilevel Models of Dynamical Systems using Latent Differential Equations Steven M. Boker and Stacey S. Poponak, USA Continuous-time Modeling of Bi-directional Influences between Family Relationships and Adolescent Problem Behavior Marc J. M. H. Delsing Estimating Dynamical Systems Disturbed by Noise Lutz-M. Alisch; Alexander Robitzsch, Germany Moment Equations and Hermite Expansion for Nonlinear Stochastic Differential Equations with Application to Stock Price Models Hermann Singer, Germany 114 Friday 9:00 11:00 The Nets: between Neural Networks and Network Analysis: Fuzzy Logic, Complexity and/or Network Analysis Social Theory Organizers: M. Ampola; D. Givigliano, Italy Social Network Analysis, Social Theory and Research Design: Experiences from an Italian Contexts Bldg. E Rm. 003 A. Salvini, Italy The Complex Dispositional Space of the Social Relation and the Nets: Between Social Networks and Neural Networks. The Problem of the Language. A. Givigliano, Italy The Aim of a Neural Social Researcher S. Gabbriellini, Italy Social Capital: Theoretical Considerations S. Milella, Italy 115 Friday 9:00 11:00 Nonresponse ( Part III ) Organizers: Claire Durand; John Goyder, Canada The Influence of Advance Contacts on Response in Telephone Surveys: A Meta Analysis E. De Leeuw; J. Hox; E. Korendijk; G. Lensvelt Mulders, The Netherlands Bldg. A Rm. 305 Determinants of Mail Survey Response Rates over a 30 Year Period, 1970-2000. An Analysis of Community Studies at the German Institute of Urban Affairs Volker Hüfken, Germany Nonresponse trends in the United States National Health Interview Survey, 1997-2003 Howard Riddick, USA Unit Nonresponse among Foreigners Caused by Non Willingness to Participate in Survey Research. An International Comparison of Strategies Dealing with the Problem Remco Feskens; Joop Hox; Hans Schmeets; Gerty Lensvelt-Mulders, The Netherlands 116 Friday 9:00 11:00 Correspondence Analysis ( Part III ) Organizer: Jörg Blasius, Germany Specific MCA with Special Reference to Questionnaires Brigitte Le Roux; Jean Chiche, France Bldg. E Rm. 008 The Space of Central Bankers in the World Frédéric Lebaron, France Visualizing Complex Change Ken Reed, Australia The (Missing) Importance of Dimension One in Factor Analytic Approaches Jörg Blasius, Germany ; Victor Thiessen, Canada 117 Friday 9:00 11:00 Cross-National Comparison of Daily Activity Patterns: Issues in the Collection and Analysis of Cross-National Time Use Data ( Part I ) Organizers: Kimberly Fisher, UK ; Elsa Fontainha, Portugal Issues in the Collection and Analysis of Cross National Time Use Data with Reference to Developing Countries Bldg. E Rm. 151 Indira Hirway Cross-National Analysis of Adolescent Time Use Data: Methodological Problems of Comparability Jiri Zuzanek; Roger Mannell, Canada ; Hannu Paakkonen, Finland Measuring Activity Patterns of Young Children - a Focus on Exercise in Early Childhood in Australia Kimberly Fisher, UK ; Michael Bittman; Melissa Wake; Ann Sanson; Robert Johnstone, Australia Improving Time Use Data Quality: Recent Experiences in Italy R. Camporese; M.C. Romanom, Italy 118 Friday 9:00 - Internet Surveys ( Part IV) Organizers: Mick P. Couper; Vasja Vehovar 11:00 Mixed mode effects in online course evaluations Roman Soucek; Anja S. Göritz; Johann Bacher, Germany Bldg. A Rm. 303 Some basic experiments on presentation of questions and response options in Web questionnaires Katja Lozar Manfreda; Gašper Koren; Valentina Hlebec, Slovenia An Experimental Comparison of Four Modes of Data Collection Within the Eurobarometer Measurement Domain Emilia Peytcheva; Robert M. Groves; Robert Manchin; Robert Tortora Mode effects on data quality in mixed mode self-administered paper, on-line and e-mail panel Anna Stangl, Germany Sample Adjustment Weights for Internet Surveys: Restricted Access versus Voluntary Participation Oztas Ayhan, Turkey The State of Online Surveys - Evidence from Germany Thorsten Faas, Germany 119 Friday 9:00 11:00 Bldg. E Rm. 004 Methodological Foundations of Sociological Analysis of Social Sphere Part III Organizers: Vassiliy Zhukov; Yelena Meshkova; Galina Ossadchaya, Russia Approaches for the Construction of a Theoretical Model in the University of the Knowledge's Age. The Configuration of Complex Cognitive Systems. Augusto Renato Pérez Mayo; Maricela Guzmán Cáceres, Mexico Enrique Erosa Solana, UK Situation on Lithuanian Labour Market and EU Accession Liongina Beinoraviciene, Lithuania Methodology of the Sociological Research. Svetlana Burova, Belarussia 120 Friday 11:30 13:00 Computer Aided Content Analysis - Research, Methods and its Future Part IV Organizer: Harald Klein, Germany Comparative Computerized Content Analysis of Party Manifestos: Methodological Considerations and Empirical Results Bldg. A Paul Pennings, The Netherlands Rm. 306 Concepts of Animal Kingdom in Brazilian Television Advertise: a Comparative Content Analysis Ana Laura Arruda; Claudia Rosa Acevedo, Brazil Instrumentally Rational Approaches to Health Care Quality in the U.S. Health Care System: From Professional Education to Practice Guidelines Amit Nigam, USA 121 Friday 11:30 - Feminism, Methodology and Methods ( Part II ) Organizer: Sharlene Hesse-Biber, USA 13:00 Gender Imago: Searching for New Feminist Methodologies Niza Yanay; Nitza Berkovitch, Israel Bldg. A Rm. 108 The Problem of Methods on Feminist Epistemologies: Discussing their Ideological and Political Background as Opposed to Solutions of Concrete Scientific Research Subjects Maricela Guzmán Cáceres; Augusto Renato Pérez, Mexico The Feminist Practice of Oral History and the Merging of Personal and Social Problems Patricia Leavy, USA Womanhood at the Margins: Notes of Researching and Resolving the Stigmatization of "Undesirable" Women" Yu-Wen Fan, USA 122 Friday 11:30 - Analyzing Interviews ( Part II ) Organizers: Manfred Max Bergman, Switzerland ; Anthony Coxon, UK 13:00 From Interview to Results: The Processes of Analysis Barbara B. Kawulich, USA Bldg. E Rm. 010 Intersubjectivity and Beyond: the use of flexible content analysis in sorting qualitative data Margrit Schreier, Germany Towards Rigorous Practice in Qualitative Research Jacques de Wet; Zimitri Erasmus, RSA Analysing Relationships in Qualitative Data in the Context of Applied Policy Research Josie Dixon, UK 123 Friday 11:30 - Methods of Social Network Analysis ( Part II ) Organizers: Peter J. Carrington, Canada ; Anuska Ferligoj, Slovenia 13:00 Actors Invisibility: About the Necessity to Re-Think Internet Research Maren Lübcke; Rasco Perschke, Germany Bldg. E Rm. 150 Using Social Networks Methods to Measure Changes in Value Systems in Cross-Cultural Perspective: the Case of Nordic-Baltic Relationship in 1991-2002 Indrek Tart, Estonia Assessing Influence and Selection in Network-Behavioral Co-evolution, with an Application to Collective Action and Informal Sanctioning Christian Steglich; Tom Snijders; Michael Schweinberger, The Netherlands Software for Statistical Analyses of Social Networks Mark Huisman; Marijtje A.J. van Duijn, The Netherlands 124 Friday 11:30 - Non Linear Modelling ( Part III ) Organizer: Cor van Dijkum, The Netherlands 13:00 Application of Fuzzy - based Methods for Urbanization Problem Ludmilla Koshlai; Mikhail Mikhalevich, Ukrain Bldg. E Rm. 007 Low Dimensional Nonlinear Dynamical Systems as a Tool for Socio-Economic Modelling Yuri Yegorov 125 Friday 11:30 - Nonresponse ( Part IV ) Organizers: Claire Durand; John Goyder, Canada 13:00 Mood and Socio-Economic Status Bias in Survey Nonresponse: Results from an 11-Wave Panel Femke De Keulenaer; Katia Levecque, Belgium Bldg. A Rm. 305 The Long-Term Effectiveness of Procedures for Minimising Attrition on Longitudinal Surveys Jon Burton; Heather Laurie; Peter Lynn, UK Survey Attrition and Estimation Strategies in the Medical Expenditure Panel Survey Trena M. Ezzati-Rice and David Kashihara, USA Understanding why People take part in Surveys: The Case of the Families and Children (panel) Survey Miranda Philips, UK 126 Friday 11:30 - Correspondence Analysis ( Part IV ) Organizer: Jörg Blasius, Germany 13:00 A Measure of Asymmetries in Brand-Attribute Dominance Relationships: Correspondence Analysis of Matched Matrices Anna Torres-Lacomba, Spain Bldg. E Rm. 008 Applying Correspondence Analysis and Logistic Regression in a Survey on Self Healing Methods Zerrin Asan; Levent Terlemez; Sevil Şentürk, Turkey Comparing Nonlinear Principal Component Analysis with Factor Analysis and Ordered Logistic Regression with an Application to Country Data Dicle Taspinar; Ozlem Deniz, Turkey Power Analysis and Sample Size Determination in the Context of Correspondence Analysis George Menexes; Iannis Papadimitriou, Greece 127 Friday 11:30 13:00 Bldg. E Rm. 151 Cross-National Comparison of Daily Activity Patterns: Issues in the Collection and Analysis of Cross-National Time Use Data ( Part II ) Organizers: Kimberly Fisher, UK ; Elsa Fontainha, Portugal The Expected and the Unexpected: Declines in Details, in Non Response, in Sociability, and in Explaining Power in Time Use Surveys Koen Breedveld; Jurjen Idema, The Netherlands The Use of Cluster Analysis to Derive Time Use Profiles from Time Use Diary Data Ken Reed, Australia ; Guy Cucumel, Canada Time Use - a Literature Review on Methods based on Economic Journals and Publications Elsa Fontainha, Portugal 128 Friday 11:30 13:00 Methodological Foundations of Sociological Analysis of Social Sphere Part IV Organizers: Vassiliy Zhukov; Yelena Meshkova; Galina Ossadchaya, Russia Actors and Interpreters: Models of Legitimization of New SocioCultural Practices in Contemporary Russia Bldg. E Rm. 004 Alexander V. Kachkin, Russia Monitoring of the Social Sphere: Methodology and Methods Galina I. Osadchaya, Russia Methodology and Methods of Migration Sociological Analysis T.N. Yudina, Russia 129 Contents INVITED LECTURES Qualitative Software and Methodological Change – myths, mindsets and methodological mania. Description, prediction, inference 145 145 145 STREAM: COMPUTER ASSISTED RESEARCH 145 Session: Computer-Assisted Sociological Research 145 Assessing Attitudes toward Muslims in the Netherlands: Does asking both Open and Closed Questions Provide more Insight? 146 Methodological Issues in Designing and Implementing Cross-Cultural Surveys. Assessing crossnational construct equivalence in the ESS 146 Combining Test Interviews with Experimental Surveys to Compare Question Formats in a Postal Survey on Alcohol Consumption 146 Elites at Home: A Multimethod Community Case Study of a Wealthy Chicago Suburb 147 A Computer-Assisted Mixed Method and Mixed Model Analysis of Police Decision-Making 147 When Quantitative is Qualitative: An Integrated Analysis 148 The Combined Use of Qualitative and Quantitative Methods in Educational Research 148 Session: Discoursive Ethnography 149 Using Conversation Analysis to Test Questionnaires 150 For “she who knows who she is…” Combining Ethnography and Conversation Analysis in the Study of Internet Newsgroups 150 Talk of Stakeholders, Talk of Players. Goffman’s Contributions to the Use of a RPG as a Social Investigation Tool 151 The Methodological Discource about the Complex Discourse of “The Child’s best Interest” 151 Session: The Achievements and Challenges of Qualitative Computing 152 Qualitative Computing and Theory Building: A Methodological Debate 152 Why Qualitative Computing is radically changing Qualitative Research 153 Adopting and Adapting to Qualitative Software: Social And Methodological Considerations from the Periphery 153 A Context-Oriented Approach to Design of Computer-Assisted Qualitative Data Analysis Software 154 Session: Computer Aided Content Analysis: Research, Methods and its Future 155 Content Analysis of Short, Structured Texts: The Need for Multifaceted Strategies 155 Extraction of Semantic Information: New Models and old Thesauri 155 Why People Reject Super-fair Offers. Evaluating Video Experiments on Ultimatum Bargaining Run in PR China 156 Measurement Instrument Choice for Tracking Email Communication 157 Inductive versus Deductive Approach: Computer-assisted Content Analysis without Dictionary? 157 Overview on Standardised Category Systems in Computer aided Content Analysis 157 TCA: General Purpose Software for Semantic and Thematic Text Analysis 158 Routinizing Frame Analysis through the Use of CAQDAS 158 Qualitative Software 2004 and Beyond 159 Comparative Computerized Content Analysis of Party Manifestos: Methodological Considerations and Empirical Results 159 Concepts of Animal Kingdom in Brazilian Television Advertise: a Comparative Content Analysis 160 Instrumentally Rational Approaches to Health Care Quality in the U.S. Health Care System: From Professional Education to Practice Guidelines 160 130 Session: The Response Process in Establishment Surveys 161 The Evaluation of Establishment Survey Questionnaires: Experiences from combining Quantitative and Qualitative Methods 161 Redesigning Statistics Canada’s Annual Survey of Manufactures – Overview of Respondent Testing using Cognitive Interview Techniques 162 The Use Of Qualitative Interviewing To Determine The Acceptability And Take Up Of Web Data Collection Instruments for Mandatory Business Surveys 162 The Contact and Response Process in Business Surveys: Lessons from a Multimode Survey of Employers in the UK 163 Using Respondent Site Visits to Improve the Design of Establishment Surveys 164 STREAM: QUALITATIVE METHODS (1) 164 Session: The Ethics and Social Relations of Research The Best and Worst of Research Relations On Vulnerabilities and Responsibilities in the Qualitative Interview Informed Consent, Gatekeepers and Go-Betweens Wading through “The Jungle” with my Main Informant. A critical discussion of ethical practice in cross-cultural ethnography. Emotional Implications in Informant-Researcher Relations The Regulation of Research Ethics – Some Challenges of Social Diversity Ethics and Power in Community-Campus Partnerships for Research 164 165 165 166 Session: Studying Sensitive Topics in Qualitative Research Minerva’s Web: The Internet and Methodologies of Sensitive Subjects Positioning Myself as a Researcher in a Politically Sensitive and Volatile Situation Studying Decision-Making Issues at Home: The Nature of the Relation between Researcher and Participants Some Functions of Probing Tactics in Interviews on Friendship Processes in Research into Sensitive Topics 169 169 169 166 167 167 168 170 171 171 Session: Recent Developments in Ethno-Methodological and Conversation Analytic Research172 Technomethodology 172 Exploring the Possibilities for the Use of Ethnographic, Contextual Information in Ethnomethodological/ Conversation Analytic Studies. 173 Shifting the Focus: The Impact of Audiovisual Technology on Empirical Research 174 Order in Disorder: Rule-Governed Actions as Sociological Phenomena 174 Tracking Psychotherapeutic Agendas with Conversation Analysis 175 The Production of a live TV-interview through Mediated Interaction 175 Talk about Talk: The Methodological Status of Teacher Interviews on Classroom Interaction 176 What’s in a Name?: Revealing Social Identity 176 Session: Research Methodology and the Challenges of Social Diversity 177 The Hermit’s Knowledge 177 Philosophic Sagacity: Another Way of Doing Research 178 Irrationality and Violence in the Postmodernity. Proposal of a Model of Analysis of the Social Reality under the Perspective of a Transdisciplinary Method 179 Researching Civic Participation within a Diverse Context: Conceptual and Methodological Opportunities and Challenges 179 Transformative Research and Dimensions of Diversity 180 Culturally Engaged Research – A Specific Application of an Emerging Methodology 181 Session: Action Research 181 The Catalyse Methodology and the Restoration of the Inquiry Results in the Action Research. Several Recent Experiments in the Rural Romanian Areas 182 131 Russian Drug-Addiction – NGO – Action-Research: Meeting on Crossroad 182 Research on Childhood and Children as Researchers: The Participatory Methodologies on Social Worlds of Children 183 Session: Evaluating Cognitive Methods to Pre-Test Questionnaires 184 Examining Expert Reviews as a Pretest Method 184 Examining Think Aloud as a Pretest Method: The State of the Art of think aloud in Cognitive Psychology 185 Toward the Accumulation of Cognitive Interview Findings: Considerations in the Design of a TestedQuestion Database 185 Looking for Trouble and Finding It, Sometimes: Exploring Relationships between Pre-survey and Post-survey Evaluation Data 186 Session: Secondary Analysis of Qualitative Data 187 Experiences with Secondary Use of Qualitative Data Material – a Feasibility Study with German Qualitative Social Researchers 187 Secondary Analysis of Panel Data: Establishing and Using a Text Databank 187 The Anonymization Process for Secondary Use of Qualitative Data 188 Revisiting ‘Classic’ Qualitative Sociology 188 Strategies in Teaching Secondary Analysis of Qualitative Data 189 Reanalyzing Qualitative Interviews from Different Angles: The Merits of Sharing Qualitative Data 189 Session: Rethinking Professional Ethics: The Hidden Researcher and the Uninformed Informant. 'Doing Ethics’ In The Field: The Art And Politics Of Covert Research Methodological Issues Experienced in a Comparative Study of Children’s Free Time Activities Making Private Issues Public: Ethnic Dilemmas Versus Moral Obligations 190 190 191 191 Session: Feminism, Methodology and Methods 192 Critical Rationalism as a Tool for Feminist Comparative Historical Sociology 192 The Clear Feminist Rejection of an Obscure Positivism 193 Entering Sites; Entering Ethnography 193 Power Tools: A Feminist Approach to Large Sample Quantitative Data Analysis 194 Gender Imago: Searching for New Feminist Methodologies 194 The Problem of Methods on Feminist Epistemologies: Discussing their Ideological and Political Background as Opposed to Solutions of Concrete Scientific Research Subjects 195 The Feminist Practice of Oral History and the Merging of Personal and Social Problems 195 Womanhood at the Margins: Notes of Researching and Resolving the Stigmatization of “Undesirable” Women” 195 STREAM: QUALITATIVE METHODS (2) 196 Session: Theoretical Approaches and Methodological Strategies to Produce and Analyze Qualitative Empirical Data 196 Transcendental Realism in a Particular Methodological Secuency 197 Post-communist Systemic Transitions, National Identity and Mass Media: Theoretical and Methodological Aspects of Analysis 197 “Representations” of Death: Theoretical and Methodological Approaches 198 Tracing Symbols. The Hermeneutic Analysis of Paradoxical Constructions within Argentine Collective Symbolism 199 People’s Representations on Corrupt Practices and the Construction of a New Moral Order: In-Depth Interviews to Middle-Class Residents of Buenos Aires 199 Social Mimesis: A Study about Cultural Performances 200 Impact of Interviewer Attitudes and Socio-Economic Status in an Inter-Ethnic Survey 200 132 Video-based Analysis of Social Situations: Comparing the Analytic Range of Ethnomethodology and Genre Analysis 201 Giving Visibility to Hidden Rural Localities: Contributions of an Ethnographic Research 201 The Emergence of New Identities in Argentine Armed Forces: A Life Course Research. 202 People’s Perceptions on the Construction of Collective Goals: Interviews in Street Meetings of Collective Protest 202 Images of the Argentine Ruling Class: Focus Groups using Mass Media Photographs as Stimulus 203 Evaluation of the Internet Use: A Call for Mixing Method Approach 204 Session: Interrogating Conventionality in Qualitative Research. Bricolage and Ingenuity in Interdisciplinary Fieldwork The Tape Recorder as a Major Player in the In-Depth Interview If the Shoe Fits: Authority, Authenticity and Agency in Feminist Diasporic Research Colonial Contexts and Post-Colonial Understandings: Researching on Women in the Indian Nationalist Movement Insider/Outsider Debates: Reflexivity in Military Research 204 205 205 206 206 Session: Miscellaneous 206 Funding of Social Research in Pakistan 206 Children as Respondents: Developing Questionnaires for Children 207 Feminist Method and Methodology and the Struggle to define Social Inquiry and Epistemology. The Logic behind the Measure. 207 Disambiguating Culture: The Cultural Element of Survey Response in Perspective 208 Session: Mixing Qualitative and Quantitative Methods: Empirical Designs and their Implications 208 Qualitative and Quantitative Methods between Paradigm and Pragmatics: Mixing, Integration or Triangulation? 209 Mixed Methods Sampling Considerations and Designs in Social Science Research 209 Why it is Necessary to Combine Qualitative and Quantitative Data and Methods to Examine the Thesis of Individualization? 210 Stratification and Social Mechanisms: An Empirical Point of View 211 Coherency Analysis for an Updating of Sociological Thought 211 Danger: Data Overload! Mixed Methods in a Cross-National Study of Care Home Staff 212 The German Research Foundation Project “Mobilzeit” as Example for the Integration of Quantitative and Qualitative Methods 212 Do Media Discourses Influence Opinions? Connecting Discourse Analysis of Newspaper Data with Time Series Analysis of Survey Data 213 The Conceptual Net of Internal Consistency and the Es of Praxiology in Human Inquiry 214 Session: Analyzing Interviews (Part 1): Qualitative Analysis 214 Strategies for Analysing In-depth Interviews of People with Disabilities in Northers Aegean, Greece 214 The Interpretation of Qualitative Data from a Critical Theoretical Framework 215 Contradictories in the Interview: Public and Personal 216 The Interview as a Text: Analyzing In-depth Interviews in Social Science with the Conceptual Framework of Linguistics 216 Analysis of Interviews with Russian Entrepreneurs using Critical Discourse Analysis. 217 Analysis of Semi-Structured, In-depth Interview Data 217 From Interview to Results: The Processes of Analysis 218 Intersubjectivity and Beyond: The Use of Flexible Content Analysis in Sorting Qualitative Data 218 Towards Rigorous Practice in Qualitative Research 219 Analysing Relationships in Qualitative Data in the Context of Applied Policy Research 220 133 STREAM: SURVEY RESEARCH 221 Session: Cognitive and Motivational Determinants for Response Effects in Survey Research 221 National Identity, Nationalism, or Patriotism: Do we Really Measure what we Assume? 221 A Cognitive Model for Survey Response Effects 222 Direction of Response Scales, Change of Numerical Values and Respondents’ Behavior 222 Cognitive and Motivational Determinants for the Strength and Direction of Social Desirability Bias in Racial Attitude Surveys 223 Session: Interaction Analysis and Data Quality of Survey Interviews 223 Conversational and Formal Questions in Survey Interviews 224 The Effectiveness of Repair Strategies of Interviewers in Survey Interviews 224 Opening the Black Box in Survey Research. Data Collection as an Ongoing Process of Total Quality Management. 225 Session: Approaches to Using Mixed Methods 226 Integrating Quantitative and Qualitative Research: How is it done? 226 Multi-Methods, Triangulation, and Integration – Different Goals or Just Different Terms? 227 Qualitative Methodology as a Parenthesis for Quantitative Methodology 228 Multi-Semiotic Ethnography, 228 A Delicate Dialectic: Mixing Qualitative and Quantitative Methods in Two Studies for US Federal Government Agencies 229 Session: Estimating Informed Opinion: Advances in the Theory and Measurement 229 CAPI-based Information Intervention (CIi): A New Way of Estimating Informed Opinion 230 Toward Informed Opinions by the Means of Deliberation 230 Gauging Informed Opinion amongst Young Citizens 231 Do People really care about GM Food Risks? A latent State-Trait Model to differentiate ‘Attitudes’ from ‘Opinions’ amongst the UK Public 232 Self-reported Familiarity and Coherence of Opinions on Biotechnology 232 Measuring Nonattitudes through Response Latencies 233 Session: Telephone Survey Methodology in the 21st Century 234 Finding (and Listing) the Unlisted: A Strategy for Achieving a Listed Sample’s Cost Savings without Sacrificing Coverage 234 Where Can I Call You? The “Mobile Revolution” and its Impact on Survey Research and Coverage Error: discussing the Italian Case 235 Mobile Phone Surveys in Hong Kong: Methodological Issues and Comparisons with Conventional Phone Surveys 236 Occurrence and Consequences of Multiple Telephone Lines, Answering Machines, and Voice Mail 236 Calculating Response Rates for Mobile Phone Surveys. A Proposal of a Modified AAPOR Standard and its Application to three Case Studies 237 The Influence of Advance Contacts on Response in Telephone Surveys: A Meta- Analysis 237 Interviewer Training: Responding to Changes in Telephone Survey Methodology 238 An Experiment in Cognitive Training of Telephone Survey Interviewers 238 Sociolinguistic Parameters of Cross-cultural Variation in Telephone Surveys: 239 Data Quality of Surveys of Minority-Majority Attitudes in Divided Societies: A Comparison between Telephone and Face-to-Face Surveys among Arabs in Israel 240 Can Multimode Approaches Salvage RDD for Public Health Surveillance? 240 Using Interactive Voice Response Technology to Capture Census Data 241 134 Session: Response Latencies in Survey Research 241 Response Latency as an Indicator of Optimizing. A Study Comparing Job Applicants and Job Incumbents’ Response Time on a Web Survey 242 Question Wording Effects and Processes of Question Answering. A Different Way to Measure Response Latencies 243 Measuring Information Accessibility and Predicting Response Effects: The Relative Validity of Differently Transformed Response Latencies 243 Controlling the Baseline Speed of Respondents: An Empirical Evaluation of Data Treatment Methods of Response Latencies 244 Expressed Uncertainty in Question Answering: Response Latencies and (Para)Linguistic Expressions in Factual and Attitude Questions 245 STREAM: COMPLEX AND LONGITUDINAL DATA ANALYSIS 246 Session: Classification and Structure: Recent Developments in Latent Class and Latent Trajectory Analysis and New Developments in Cluster Analysis 246 Methods for Comparing Latent Class Cluster Solutions 246 L∞-consensus for Dendrograms : an Alternative to the Average Consensus Procedure 247 Classifying Adolescents Using Value Orientations – a Comparison of Cluster Analysis and Latent Class Analysis 247 Detecting Aberrant Patterns of Change for the Dimensions of a “Quality of Life” Questionnaire by means of Linear Mixed Models 248 SPSS TwoStep Clustering – A First Evaluation 248 Clustering Messy Social Data with ClustanGraphics 249 Cluster Analysis Software: Proposals of Tools for Validation 249 Empirical Validation of “Whisky Classified” 249 Session: Multilevel Analysis Frequentist MCMC Estimation Methods for Multilevel Logistic Regression The Use of Internal Pilot Studies to Derive Powerful and Cost-Efficient Designs for Studies with Nested Data Performance of Likelihood-Based Estimation Methods for Multilevel Binary Regression Models Outliers and Multilevel Models On the Relative Efficiency of Unequal Cluster Sizes in Multilevel Intervention Studies Multilevel Multiprocess Modelling of Partnership Transitions and Fertility in Britain The Concept of ‘Social Level’ and how to assess it The Hausman Test of Random Effects Specifications The Problem of Time Dependent Explanatory Variables at the Context Level in Discrete Time Multilevel Event History Analysis. 250 250 251 251 251 252 252 253 253 254 Session: Measurement (In)Variance in Longitudinal Research 254 Practical Issues in the Assessment of Measurement Invariance and why Invariance does not Imply Measurement 255 Measurement (In)Variance in SEM State Space Modelling of Panel Data 255 Types of Change in Self-Report Data: Definition, Interpretation and Operationalization 256 Impact of Direction of Answering Categories on Response Behavior 256 Testing Measurement Invariance with Respect to Time, Group, and Latent 257 Individual Differences in Factor Loadings and the Fit of the Standard Factor Model: Consequences for Quantitative Behavioral Genetics Research 257 Session: Methods of Social Network Analysis 258 Playing God on a Budget: Understanding Large-Scale Social Structure from Sample Data and Random Graph Models of Interdependent Local Processes 258 New Specifications for Exponential Random Graph Models 259 135 259 New Developments in Estimation Methods for the P2 Mode Actors Invisibility: About the Necessity to Re-Think Internet Research 260 Using Social Networks Methods to Measure Changes in Value Systems in Cross-Cultural Perspective: the Case of Nordic-Baltic Relationship in 1991-2002 261 Assessing Influence and Selection in Network-Behavioral Co-evolution, with an Application to Collective Action and Informal Sanctioning 262 Software for Statistical Analyses of Social Networks 262 Session: Panal Data Analysis 263 Estimation of a two Equations Panel Model with Mixed Continuous and Ordered Categorical Response Variables in the Presence of Missing Data 263 Comparing Fixed Effects and Covariance Structure Estimators 263 Analyzing Individual and Collective Change of Attitudes Using Panel Data. The Example of Attitudes to Social Security in Germany 264 Session: Combining Data from Different Source 265 Estimation with Population-Level Constraints on Two Survey Datasets: An Application to first Birth Probabilities by Education in Italy 266 Combining Information from Multiple Surveys in the Analysis of Coarsened Longitudinal Data 266 A Comparison Record Linkage and Statistical Matching and Their Impact on Linear Regression Analysis 267 Combining Separate Datasets in the OPUS Project 267 Modelling the Construction of a Social Accounting Matrix in the Context of Statistical Matching 268 Integrating Data for Innovative Analyses: The North Carolina Education Research Data Center 268 Linking Job Episodes from Survey and Register Data: Specific Challenges and Empirical Results 269 Non-Exact vs. Exact Matching with Applications to Wages Statistics 270 Combining Datasets to Create a Synthetic Population: Sample Enumeration. 270 Robustifying Imputation Model Estimation 271 Linking Public Health Data in Georgia, USA 271 Two Stories about Youth Unemployment? Combining Register Data and Self-Reported Data 272 STREAM: INFORMATION AND SIMULATION SYSTEMS 272 Session: Non Linear Modelling 272 Dynamics of Functional Connectivity in Ongoing, No-task EEG. Evidence for Scale-Free Properties and Explanation in Terms of Coupled Limit Cycle Oscillators 273 Dynamical Complexity. A Measure of Critical Instability During Human Change Processes 274 Evolutionary Economics and Complexity 274 The Dynamics of Bilingual Language Development in Immigrant Children 275 Methodological Issues of Labour Market Analysis for Countries in Transition 275 Anticipation and Codification in Communication Systems: Towards a Simulation Model for Luhmann’s Sociological Theory of Communication 276 Catastrophe Theory and Social Policy Making: A Nonlinear Model of the Early Pension Insurance Legislation in Western Europe 277 Application of Fuzzy - based Methods for Urbanization Problem 277 Low-Dimensional Nonliear Dynamical Systems as a Tool for Socio-Economic Modelling 278 Session: Tools and Techniques for Agent-Based Social Simulation 279 Tracking the Invisible Hand: An A.B.M. Approach to CDA Design with Evolutionary Behaviour 279 Development of a Hybrid Multi-Agent Model for Modelling Petrol Price Markets 280 Dynamics of International Negotiations: A Simulation of EU Intergovernmental Conferences 280 On a Theorem in the Theory of Guerrilla Warfare 281 Using Multi Agent Systems and Role-Playing Games to Simulate Water Management in Peri-urban Catchments 282 Lagoon, Agents and Kava: A Companion Modelling Experience in the Pacific 282 136 Agent-Based Simulations backing Use of Role-Playing Games as Dialogue Support Tools: Teaching from Experiments 283 Actualities of Social Representation: Simulation on Diffusion Processes of SARS Representation 283 MASON: A JAVA Multi-Agent Simulation Library 284 Simma: A Toolkit for Participatory Business Simulations 284 The Political Actor Simulator (PAS): The Simulation of a Foreign Policy Crisis during the Peacekeeping Mission in Bosnia 1995 285 Using GIS and Multi-agent Systems for Interactive Design of Policy in Natural Resource Management 285 Multiple Equilibria Regulation Model in Cellular Automata Topology 286 Session: Informetric and Text Based Procedures for Information Retrieval and Analysis Bibliometric Mining Automating the Analysis of Co-Words in Contexts: Measuring the Meaning of “Stem-Cells” and “Frankenfoods” in Different Domains Mining Document Contents in Order to Analyse a Scientific Domain Comparing a Knowledge-Based Economy: In the Case of South Korea and the Netherlands Graph-Based Representations of Motion Pictures: An Attempt to Discover Statistical Features of Aesthetic Merit Using Search Engines and Web Crawlers in Social Science Research Digital Libraries and Heterogeneity Searching and Browsing Multiple Subject Gateways in the Renardus Service Ontology Switching for the Social Sciences. Methods for the Underlying Corpus Analysis Bilateral Transfer Modules for the Conceptual Integration of Heterogeneous Document Bases in Content Analysis 287 287 287 288 289 289 290 291 291 291 292 Session: Time Series Analysis 293 Comparison of Seasonal Variation of Births and Deaths in Urban and Rural Areas of Silesia (Poland) 293 Population at Risk (PAR) Rates and other Denominators used to Measure Unemployment Rates 293 Auto-regressive Time Series Cross Section Analysis (TSCS) for Proportions Data with Random Coefficients. A Flexible Model and its Application in Regional Science 294 Session: The Statistical Estimation of Linear and Non-Linear Stochastic Differential Equations 295 A Survey of Estimation Methods for Stochastic Differential Equations 295 A Comparison of Different Procedures to Estimate the Damped Linear Occilator for Panel Data 295 Multivariate Multilevel Models of Dynamical Systems using Latent Differential Equations 296 Continuous-time Modeling of Bi-directional Influences between Family Relationships and Adolescent Problem Behavior 296 Estimating Dynamical Systems Disturbed by Noise 297 Moment Equations and Hermite Expansion for Nonlinear Stochastic Differential Equations with Application to Stock Price Models 297 STREAM: GENERAL STATISTICS 297 Session: Assessment of Model Fit Analysis of Deviance with Penalized Likelihoods Examining the Scalability of Intimacy Permissiveness in Taiwan Additive and Multiplicative Effects in a Fixed 2 x 2 Design using ANOVA can be difficult to Differentiate: Demonstration and Mathematical Reasons How to Evaluate the Fit of Linear and Generalized Multilevel Models Collinearity involving Ordered and Unordered Categorical Variables Testing the Fit of Conditional Independence with a Continuous Control Variable Assessment of Model Fit using Incomplete Data 297 298 298 137 299 299 299 300 300 Session: New Methods in Political (and Social) Methodology 300 Time-Series--Cross-Section Data 300 New Directions in Spatial Analysis: Space is more than Geography 301 American Public Opinion in the 1930s and 1940s: the Analysis of Quota-Controlled Sample Survey Data 301 Estimating Spatial Autoregressive Coefficients in Spatial Lag Models 302 Biases in the Effects of Parental Background on Voting 303 Event Data Based Network Analysis (EDNA) 303 Matching Methods for the Causal Analysis of Observational Data 304 Session: Recent Developments in Nonparametric Item Response Theory Detecting Discontinuity On a Cognitive Developmental Ability: A Comparison of the Binomial Mixture Model and the Latent Class Model. A Markov Chain Approach for Ranking Individuals from Item Responses A Bayesian Framework for the Monotone Homogeneity Model of Non-Parametric Item Response Theory Multiple Imputation of Item Scores in Test and Questionnaire Data, and the Influence on Psychometric Results A Nonparametric IRT Model for the Circumplex: Generalization to Polytomous Data On Ordering Properties of Classical Optimal Scaling 304 305 306 306 307 307 308 Session: Graphical Modelling 309 Some Aspects of the Analysis of Graphical Markov Models 309 Computational Aspects of Chain Graph Models with Mixed Response Variables 309 Checking for a Confounder in Models Generated over Directed Acyclic Graphs 309 An Application of Marginal Log-linear Models to Examine Changes in Social Mobility in Hungary During the Transition Period 310 Marginal Models for Multivariate Dependencies 310 The Use of Bayesian Networks for Imputation 310 Session: The Nets between Neural Networks and Network Analysis. Fuzzy Logic, Complexity and/ or Network Analysis Social Theory 311 Social Network Analysis, Social Theory and Research Design: Experiences from an Italian Contexts 311 The Complex Dispositional Space of the Social Relation and the Nets between Social Networks and Neural Networks. The Problem of the Language. 312 The Aim of a Neural Social Researcher 312 Social Capital: Theoretical Considerations 313 Session: Miscellaneous A Strategy for the Selection of Weighting Variables The Measurement of Duration of Studies in Higher Education 314 314 314 STREAM: DATA COLLECTION, SAMPLING AND NON-RESPONSE 315 Session: Improving the Validity of Fraud Research using Instruments from the Social Sciences 315 Evaluating the Validity of (Vicarious) Self-Reports of Deviant Behavior Using Vignette Analyses 315 Statistical and Technical Methods for the Detection and Prevention on Interviewer Fraud in Data Collection 316 Research Designs for Fraud Studies: How to Chose the Optimal Research Design? 316 Session: Preventing, Diagnosing and Analyzing Missing Data Mixed Missing: Mode Elemental Structures Social Distance, Interviewer’s Perception of Rapport and the Types of Invalid Response 138 317 317 318 Socioeconomic Status of Survey Respondents with Missing Household Income Data Statistial Editing and Imputation for Social Data 318 318 Session: Nonresponse 319 The Influence of Advance Contacts on Response in Telephone Surveys: A Meta Analysis 319 Efficacy of Incentives in Increasing Response Rates 320 New Evidence on Amount of Payment on Prepaid Incentives on a Mailed Questionnaire 320 A Plea for the Tailored Use of Respondent Incentives 321 The Drop-off Pick-up Method: An Alternative Approach to Reducing Nonrespone in Mail Surveys 321 Recruitment Procedures and Response Rate: A Swiss Experiment 321 The Long-Term Effectiveness of Procedures for Minimising Attrition on Longitudinal Surveys 322 Mood and Socio-Economic Status Bias in Survey Nonresponse: Results from an 11-Wave Panel 323 Survey Attrition and Estimation Strategies in the Medical Expenditure Panel Survey 324 Determinants of Mail Survey Response Rates over a 30 Year Period, 1970-2000. An Analysis of Community Studies at the German Institute of Urban Affairs 324 Understanding why People take part in Surveys: The Case of the Families and Children (panel) Survey 324 Nonresponse trends in the United States National Health Interview Survey, 1997-2003 325 Understanding Nonresponse of the Twelfth Graders in the National Assessment of Education Progress: Social Isolation Theoretical Approach 326 Unit Nonresponse among Foreigners Caused by Non Willingness to Participate in Survey Research. An International Comparison of Strategies Dealing with the Problem 326 Task Dedication, Task Alienation and What Data Collection Instruments Mean to Respondents 327 Measuring the Respondent’s Attitude towards Surveys. 327 Does the Performance Matter? A Multilevel Analysis of the Effects of Individual and Performance Characteristics on Response in Audience Research 328 Survey Experience, Generalized Attitudes Towards Surveys and Respondents’ Willingness to Answer Questions in Subsequent Surveys 328 Session: Aided Recall Techniques in Survey Interviews 329 Improving the Quality of Retrospective Reports: Calendar Interviewing Methodologies 330 Collecting Event History Data about Work Episodes Retrospectively: What Recall Errors Occur and how can we Prevent them? 330 Influence Dating Formats on the Telescoping Effect and the Distribution of Autobiographical Memory 331 Timeline Methods and Related Aided Recall Applications in Social Sciences and Health Studies: a Review. 331 Session: Data Collection 332 Patterns of Mobile Phone Usage and their Impact on Participation in Mobile Phone Surveys 332 Comparing Anonymous and Non-Anonymous Research Methods Asking Sensitive and Socially Desirable Questions. 333 Improving Questionnaire Design through the Provision of Electronic Access to Large Scale Survey Questionnaires in the UK 333 Effects of Survey Sponsorship and Mode of Administration on Respondents’ Answers about their Racial Attitudes 334 Session: Sampling Methods Comparing Sampling Frames Methods for Achieving Equivalence of Samples in Cross-National Surveys Design Effects in the Analysis of Longitudinal Survey Data Impact of the Geographical Size of Clusters on the Precision of Survey Estimates GenD: An Evolutionary Algorithm for Re-sampling in Survey Research Sampling Process-Generated Data in order to Trace Social Change Estimating Domestic Tourism Expenditure in Developing Economies: Lessons from India 139 334 335 335 336 336 337 338 339 Sampling Undocumented Immigrants in Brussels via Snowball Sampling: Theoretical and Practical Considerations 340 Using Adaptive Sampling Methods in the Analysis of Social Phenomena on the WWW 341 STREAM: LATENT VARIABLES MODELLING Session: Latent Growth Curve Structurual Equation Models, Autoregressive and Hybrid Models Combining Autoregressive and Semiparametric Mixture Models: Investigating the SelectionFacilitation Effect of Peers on Antisocial Behaviors Pay Off and Impact of Varying the Intercept in Latent Growth Curve Modeling Latent Growth Curve and Interaction Modeling: The Case of Travel Mode Choice Anomia, Left right Orientation and Group Related Enmity. 341 341 342 342 343 343 Session: Multidimensional Scaling and Facet Theory 343 What is not what in Facet Theory 344 Questionnaire Design and Data Analysis using Facet Approach: The Case of the International Census on Attitudes toward Languages 344 New Type of Psychodiagnostical Test: Situational Operational Grid 345 Explaining the Structure of Values of an Ideal Organization by Schwartz‘s Universal Value Theory 345 Session: Correspondence Analysis 346 Nordic Business Researchers’ Use of Research Methodology: An Assessment of Researchers in Denmark, Finland, Norway and Sweden 347 Using Combined Methods to Develop Sociological Concepts. An Exploration of the Concept of Social Capital using Correspondence Analysis and Case Studies Research 347 The Analysis of Lifestyle Groups and Social Milieus in Germany Using Cluster and Correspondence Analysis 348 International Comparison with Correspondence Analysis. Attitudes towards Same Sex Partnership as an Empirical Example 348 The Use of Correspondence Analysis for Stratification 348 Measure versus Variable Duality in Correspondence Analysis 349 The Space of Elite Opinions. The Case of Norway 350 Processes of Social Differentiation in an Economic “Social Field” 350 Specific MCA with Special Reference to Questionnaires 351 The Space of Central Bankers in the World 352 Visualizing Complex Change 352 The (Missing) Importance of Dimension One in Factor Analytic Approaches 353 A Measure of Asymmetries in Brand-Attribute Dominance Relationships: Correspondence Analysis of Matched Matrices 353 Applying Correspondence Analysis and Logistic Regression in a Survey on Self Healing Methods 354 Comparing Nonlinear Principal Component Analysis with Factor Analysis and Ordered Logistic Regression with an Application to Country Data 354 Power Analysis and Sample Size Determination in the Context of Correspondence Analysis 355 Session: Classification and Visualisation Cluster Based Conjoint Analysis The Typological Non Symmetrical Correspondence Analysis: Theory and Application for Social Research Some Different Typologies of Employment in Italy Session: Improving Survey Methodology: The European Social Survey Measuring Attitudes towards Immigration across Countries: Potential Problems of Functional Equivalence in the ESS 140 356 356 357 357 358 359 Quality Of Life in Europe: Comparing States using a Composite Index 359 Development of a Standardized Interviewer Questionnaire for International Research into Interviewer Effects on Nonresponse and Data Quality 360 Improving Survey Methodology: The ESS. Index and Measurement: The Cultural Borders of Meaning 360 Response Propensity and Data 361 Geographical Variations in Electoral Participation: Evidence from a Multilevel Analysis of the European Social Survey 361 European Value Map. Based on ESS Data 362 Analysis and Relation between Social Variables and Economic Variables in the European Social Survey and World Bank and UNDP Reports: The Social Capital and Economic Capital at the European Countries 362 Influence of Media Reported Events on Attitudes and Opinions 363 Some Methods for Investigating Validity and Reliability of Questions in the European Social Survey 363 Modelling Interviewer-Effects in the European Social Survey 364 International Benchmarking: Are we Comparing Apples to Oranges? 364 Modeling Events for ESS: Toward the Creation of an Autonomous Tool for Survey Research 365 STREAM: INTERNATIONAL COMPARATIVE RESEARCH 366 Session: Comparable Data across Countries Pursuing Equivalence in Cross-National Surveys Comparability across Countries of Responses in the ESS Trapped in Translation? ESS Translation Protocols provide a Key Sampling for the European Social Survey Data Quality Assessment in ESS Round 1: Between Wishes and Reality Context, Events and Attitudes 366 366 366 367 367 368 368 Session: Methodological Issues in Designing and Implementing Cross-Culturel Surveys 369 A Process Quality Approach to the Testing and Evaluation of Cross-Cultural Surveys 369 Language, Usage, and Cultural Context: Challenges to “Functional Equivalence” in Questions 370 Developing a Low-Cost Technique for Parallel Cross-Cultural Instrument Development: The Question Appraisal System (QAS-04) 370 Test of Anchoring Vignettes in a Low Income Population 371 Disambiguating Culture: The Cultural Element of Survey Response in Perspective 372 Question Design Features and Cultural Variability in Comprehension 372 Socio-Cultural Factors in the Question-Response Process: A Comparative Coding Analysis of Latino and Anglo Cognitive Interviews 373 What are we talking about ?Interviewing and Interpretation for a Biographical Survey in Sub-Saharan Africa. Cross-cultural/ Cross-National Survey Research Session 373 Challenges in Implementing a Multi-National Tobacco Use Survey 374 Attitude Scale Development in Countries which send Illegal Immigrants 375 Behavior Coding of Multilingual Survey Interviews: What Can We Tell by Systematically Listening? 375 Comparability of Attitude Measures towards Social Inequality. The Level of Equivalence in the ISSP 1999, its Implications on Further Analysis and the Presentation of Results 376 Session: Issues of Cross-National Comparative Research 377 How to Grasp Nations in International Survey Comparisons 377 (P)raising Cross-National Comparability: Digital Divide Benchmarks Based on Time Distance Method 378 An Empirical Enquiry into Students’ Socio-Economic Background, Costs and Financing of Studies, and Future Career Aspirations in Higher Education in China: Issues of Methodology 379 141 Values in Europe: A Multiple Group Comparison with 20 Countries Using the European Social Survey 2003. 379 Session: Cross-National Harmonization of Socio-Demographic Variables 379 Obtaining Harmonized Demographic Information from European Surveys -Experiences of the German Federal Statistical Office form Input to Output Harmonization 380 Harmonizing Background Variables in International Surveys 380 Harmonization of Survey Data in the "International Social Survey Programme" (ISSP) 380 Measuring Religious Membership and Religiosity across Europe 381 Is there Another Way to View Race? Listening to Subjectivity in the Categorization of Ethnicity and Race among Hispanics in the United States. 381 Ethnicity and the Comparative Analysis of Contemporary Survey Data 382 Measuring Ethno-Poverty: Methodological Challenges 383 Measuring "Income" in Cross-National Research 383 Session: Cross-National Comparison of Daily Activity Patterns: Issues in the Collection and Analysis of Cross-National Time Use Data 384 Issues in the Collection and Analysis of Cross National Time Use Data with Reference to Developing Countries 384 Cross-National Analysis of Adolescent Time Use Data: Methodological Problems of Comparability 385 Measuring Activity Patterns of Young Children – a Focus on Exercise in Early Childhood in Australia 386 Improving Time Use Data Quality: Recent Experiences in Italy 386 The Expected and the Unexpected: Declines in Details, in Non Response, in Sociability, and in Explaining Power in Time Use Surveys 387 The Use of Cluster Analysis to Derive Time Use Profiles from Time Use Diary Data 387 Time Use: A Literature Review on Methods based on Economic Journals and Publications 388 STREAM: DATA COLLECTION AND INTERNET SURVEYS 388 Session: Visual Design and Layout in Internet and Mail Surveys 388 A Comparison of Sliding Scales with Other Scale Types in Online Surveys 389 Online Surveys-Does One Size Fit All? 389 Web Survey Design: Effect of Layout on Measurement Error 390 An Experimental Testing of Format Changes to Reduce Missing Data and Increase Cooperation in the Nielsen TV Diary 391 Achieving Usability in Establishment Surveys through the Application of Visual Design Principles 391 Session: The Internet and Social Research: New Opportunities, New Challenges 392 Methodological and Ethical Problems in Analysis of Web Site Authors 392 Gender Specific Analysis of Virtual Communication in Panel Online Focus Groups 393 “If you Build it, they will come” – or will they? A Discussion on the Acceptability of the Web as a Means to Respond to Surveys 393 Dropping Out of an Online Sex Survey: when and why? 394 Validity of Data and Representativeness of Sample in Internet Survey: University Surveys in the United States 394 Actors Invisibility: About the Necessity to Re-Think Internet Research 395 Session: Internet Surveys A Comparison of Ranking versus Rating in Online Surveys Behavioral Intention Measurement: International Findings Evaluating the Effectiveness of Visual Analog Scales: A Web Experiment Don’t Know and No Opinion Responses in Web Surveys Design Issues in Collecting Data on Ego-centered Social Networks on the Web 142 395 396 396 397 397 398 A Comparison of Two Web-Based Surveys: Static versus Dynamic Versions of the NAMCS Questionnaire 398 Computing Methods and their Impact on Drop Out 399 Effectiveness of Progress Indicators in Web Surveys 400 Item Non Response and Dropouts in Web Surveys 400 Effects of Personalization on Web Survey Response Rates and Data Quality 401 Feasibility of a Random Sample 401 Self-Selection as a Sampling Method in Web Surveys 402 Noncoverage and Nonresponse in a Web Survey 402 Utilization of an Internet Survey of Work-Family Conflict among Home-Based Teleworkers 403 Investigating the Imagined Online Community 403 Mixed Mode Effects in Online Course Evaluations 404 Some Basic Experiments on Presentation of Questions and Response Options in Web Questionnaires 404 An Experimental Comparison of Four Modes of Data Collection Within the Eurobarometer Measurement Domain 405 Mode Effects on Data Quality in Mixed Mode Self-Administered Paper, On-line and E-mail Panel 406 Sample Adjustment Weights for Internet Surveys: Restricted Access versus Voluntary Participation 406 The State of Online Surveys: Evidence from Germany 407 Session: Miscellaneous Using Email as a Site for Constructing Narratives through Interview Getting it Right: Designing an Evaluation of an INGO Led HIV/AIDS Prevention Programme in Resource Poor Settings Cyberethnography: New Fields of Inquiry or are we simply Re-Cycling Old Methodology? 408 408 408 409 STREAM: GENERAL METHODOLOGY 410 Session: Is There a Fundamental Change in the Individual Shaping of the Life Course in Modernity? Or how can we do Research on Fundamental Transformations in Modernization? 410 Using Biographical Material to Research University Students’ Identity: An Avenue for Research on Social Change 410 Experiential Glances at Transformations and Changes Occurring in Reflexive Modernization Life. 411 Reflexive Modernization and Biotechnology: The Conflict of Biotechnology as a Conflict between the Logic of Industrial Society and the Logic of Reflexive Modernity 412 Session: Methodological Foundations of Sociological Analysis of Social Sphere 413 Designing of Disablement in the Institutional Field 413 Theories and Methods in Cross-National Youth Research 414 Tasks of Historiography of Social Sphere Sociology 414 Methodology and Methods of Comparative social research 415 Cross-Cultural Dialogue: Methodology of Research 415 The Peculiarity of Treatment the Problem of Trafficking in Woman in Mass Media 416 Sociological Measurement of Social Protection 416 Approaches for the Construction of a Theoretical Model in the University of the Knowledge’s Age. The Configuration of Complex Cognitive Systems. 416 Situation on Lithuanian Labour Market and EU Accession 417 Methodology of the Sociological Research. Level of Public Awareness on Domestic Violence against Women/ Girls and Sexual Harassment at the Work/ Study Place; its Scale, Types, Forms and Consequences 418 Actors and Interpreters: Models of Legitimization of New Socio-Cultural Practices in Contemporary Russia 418 Monitoring of the Social Sphere: Methodology and Methods 419 143 Methodology and Methods of Migration Sociological Analysis 419 Session: Complex Societal Problems 420 Statistical Evidence: Responsibilities and Burden of Proof 420 The Geography of Decision-Making: Linking Attitudes, Knowledge, Management Activities and Forestland Characteristics in a Spatial Environment 421 Causality in Complexity 421 Interactive Mapping of Complex Policy Networks: A Quick Scan for Strategy Making 422 Application of Fuzzy - based Methods for Urbanization Problem 422 Area Studies Approach: A Critique and Search for Different Methods 423 Multi-Criterion Decision Modeling Within Rural Societal Policy - The Case Of South - East Europe 423 Methodological Challenges Concerning a Comparison of the Complex Social Problem. Methodological Reflection upon the Empirical Study Comparing how People of the EU Countries Manage to Balance Family Life and Job Career 424 Session: Miscellaneous Counting, Measuring, Meaning and Reality: Issues in Sociological Research Socio-Cultural Foundations of Gender Parity 144 425 425 425 INVITED LECTURES Qualitative Software and Methodological Change – myths, mindsets and methodological mania. Lyn Richards, Research Services, QSR International, Australia Innocent tool or deus ex machina? At its advent, software for qualitative research was frequently greeted with fear that it would preemptively determine method. Two (contradictory) arguments dominated - that qualitative computer programs are skewed to support for grounded theory and that hierarchical index systems forced top-down thinking. Neither has been seriously debated, and it seems that this debate may now not take place. As the use of computer programs becomes taken for granted the implications for methodological choice are rarely discussed. This keynote argues the question should be addressed, not allowed to fade with familiarity. Description, prediction, inference Jan de Leeuw, Department of Statistics, UCLA, USA We analyze the uses and misuses of multiple regression and related techniques in the social, behavioral, and educational sciences. Its main uses are description of multivariate distributions, prediction of missing or future events, and causal inference, and statistical inference from sample to population. We'll argue that description is always legitimate and often useful. Prediction is very useful indeed, always legitimate, but seldom practiced, except in very artificial settings. And inference, of both the causal and the statistical type, is almost always problematical. References Berk, R. (2003). Regression Analysis. Sage. Wang, C. (1993). Sense and Nonsense of Statistical Inference. Dekker. STREAM: COMPUTER ASSISTED RESEARCH SESSION: COMPUTER-ASSISTED SOCIOLOGICAL RESEARCH Organizors: Karl M. van Meter Contact: National Center of Scientific Research, 59 rue Pouchet, 75017 Paris, France, tel/fax 33 (0)1 40 51 85 19, [email protected] Abstract With the widespread use of computers in all aspects of sociological research, few empirical sociologists still make a distinction between "qualitative" and "quantitative" research. These two terms have come to mean different styles or different emphasis in doing research. We invite submissions which explore this complementarity and, in particular, research projects which "cross the borders" by going from data-intensive "qualitative" approaches to formalized "quantitative" results and, vice versa, from formalized "quantitative" survey approaches and data "back to the field" to confront results with "qualitative" data-intensive case studies. 145 Assessing Attitudes toward Muslims in the Netherlands: Does asking both Open and Closed Questions Provide more Insight? Christine L. Carabain, Vrije Universiteit, Faculty of Social Sciences, Department of Social Research Methodology, De Boelelaan 1081, 1081 HV Amsterdam, [email protected] In attitude research, it is common practice to assess attitudes by presenting a series of statements and asking respondents to what extent they agree or disagree with each of those statements. Reactions to statements are generally gauged via a five-point Likert-scale. This paper provides a systematical comparison between the answers of such statements and open questions, in order to make inferences about the meaning of choosing moderate answer categories, such as ‘partly agree’, ‘don’t agree and don’t disagree’, and ‘partly disagree’. In this study, Dutch students were interviewed on their opinions about Muslims in the Netherlands. When they were confronted with statements on the topic, almost 80%, of their answers were scored within the moderate answer categories. A systematical comparison between those answers and the answers of the respondents on comparable open questions shows that open questions trigger the expression of more differentiated opinions. Respondents while answering open questions 1) constrained their answer with a condition; 2) differentiated the group they were referring to, and 3) differentiated the qualifications they assigned to a group. A striking result of this study is that when answering statements, the choice between the different moderate answer categories seemed rather arbitrary. Respondents who made comparable distinctions while answering open questions did often chose different answer categories while answering the statements and vice versa. This study shows the risks of using solely closed question to assess attitudes and the possibilities of using ‘mixed methods’ to assess them. Methodological Issues in Designing and Implementing Cross-Cultural Surveys. Assessing cross-national construct equivalence in the ESS Jaak Billiet & Jerry Welkenhuysen-Gybels, K.U. Leuven, Belgium, [email protected] One of the prime objectives of cross-national survey research is to compare concepts across countries or cultures. It is therefore important that these concepts are measured adequately in all of the countries involved in the survey. Moreover; in order that country-scores on items or scales can be compared in a valid way, concepts have to be measured in a sufficiently equivalent way. This paper tries to assess this equivalence by checking whether the indicators of a number of latent traits such as religious commitment, and several dimensions of the attitude toward immigration measure the same construct or trait in all of the countries. This will be done by testing for the factorial invariance of factor loadings across countries. Two different methods are used. Whenever possible, response style effects are taken into account in this analysis. The analysis shows that it is possible to obtain comparable measurement elements for a number of concepts that are measured in ESS round 1 Combining Test Interviews with Experimental Surveys to Compare Question Formats in a Postal Survey on Alcohol Consumption Harrie Jansen1 and Viviënne Lahaut2 1 IVO, Addiction Research Institute Rotterdam, The Netherlands, [email protected] 2 ZonMw, The Netherlands Organisation for Health Research and Development, Den Haag, [email protected] In quantitative surveys on alcohol consumption, last week reports (LW) and typical week reports (TW) are very common question types. The main rule in comparison of questioning formats in this field of research is the ‘more is better’ principle: questions generating more reported alcohol 146 consumption are better than questions generating less reported alcohol consumption. This rule seems to oversimplify the matter of data quality, however. Therefore in our study we studied data quality in three ways: mean reported alcohol consumption at the aggregate level to measure ‘alcohol generating power’, face to face Three Step Test Interviews at the personal level to measure responding problems and item response at the aggregate level to measure motivating power of each question format. In this way we compared two formats on last week’s drinking (LW1 & LW2) and two formats on typical week’s drinking (TW1 & TW2). The results show that each method reveals specific relevant aspects of data quality, concluding that a) more is not always better, 2) better questions generate less answers. Elites at Home: A Multimethod Community Case Study of a Wealthy Chicago Suburb Al Hunter, Northwestern University, USA, [email protected] The research reports on a community case study of an elite, wealthy suburb of Chicago. The focus is on the day-to-day lives of its residents, their informal social networks and institutional participation in voluntary associations, local schools and churches, and local politics. The research explores attitudes and behaviors related to exclusion and inclusion with respect to the remainder of society. The research is based on a questionnaire survey of 10% of households in the community, in-depth interviews of residents, and participant observation of public events and festivals. The different methods address different beliefs, attitudes, and behaviors that distinguish public from private spheres of local community life. A Computer-Assisted Mixed Method and Mixed Model Analysis of Police DecisionMaking Jennifer L. Schulenberg, [email protected] Department of Sociology, University of Waterloo, Canada, In order to examine the organizational processes that occur in police decision-making, one needs to uncover what, when, how, why, and to what extent some decisions are made over others. Thus, a mixed methods approach was adopted in this research that not only uses quantitative and qualitative data but also analyzes the qualitative data through a grounded theory approach and statistically. The qualitative data are based on over 200 semi-structured interviews with more than 300 police officers in 95 police services across Canada – from all provinces and territories, all types of communities, and all types of police services. For purposes of quantitative analysis of the interviews, two SPSS databases were constructed: one with the interview as the level of analysis and one with the police agency as the level of analysis (through the development of aggregation rules when there was more than one respondent per police service). About 200 variables were coded in these databases from the interviews by means of content analysis in the same way that open-ended questions in a survey are coded for quantitative analysis. Statistical data on police decision-making were added to the police agency database. These were taken from the national Uniform Crime Reporting Survey which is maintained by Statistics Canada. Characteristics of the communities were taken from the Census and added to the police agency database. The quantitative analysis of the police agency database used the following techniques: univariate descriptives, bivariate analyses using cross-tabs, comparison of means, and one-way ANOVA, and multivariate analysis using multi-way ANOVA. The interviews were also qualitatively analyzed using a between-methods approach and multi-phase coding with the Computer-Assisted Qualitative Data Analysis (CAQDA) software called Qualrus. In this analysis, the interviews were coded qualitatively – segment by segment – in two ways: (1) using pre-established codes suggested theoretically and by the literature, and (2) coding within these 147 categories for salient themes and factors. The results of these analyses are presented concurrently as the quantitative and qualitative analyses clarify and illustrate each other. The resulting description and explanation of police decision-making are both positivist – showing the factors which are associated with outcomes – and constructivist – showing the ways in which police officers’ understanding of their occupational world conditions their behaviour. This paper will present one research question and one hypothesis in order to illustrate the use of both data types that are analyzed both quantitatively and qualitatively. The technical and epistemological issues that will be addressed are the appropriate use of statistical methods, and the meaning of codes, themes and variables when quantifying qualitative data. Keywords: police decision-making, mixed method, mixed model, CAQDA When Quantitative is Qualitative: An Integrated Analysis Rod Gutierrez1 and Pat Bazeley2 1 University of Sydney, Australia, [email protected] 2 Research Support P/L, Australia, [email protected] Making a qualitative interpretation from statistical analyses of data coded from free-flowing text is not a standard approach to research design, even within the domain of mixed methods research. The employment of such an intertwined analysis strategy has, however, revealed new insights into the causes and experience of occupational stress and psychological injury. Psychological assessments for 157 cases claiming work related mental disorders were examined, along with reports from the claimants, their superiors and an assessing medical practitioner, each version being 'scored' for presence or absence of a range of potentially critical factors. Level of agreement in the reports was assessed using non-parametric statistics, but of more significance in the context of integrated analysis, predictive tests (primarily logistic regressions) were conducted to identify factors most likely to be involved in the development of mental disorders and/or psychological injury. Results of these tests considered individually were inconclusive, but when they were considered together an explanatory pattern emerges to tell the story of an increasingly common but poorly understood workforce problem. The Combined Use of Qualitative and Quantitative Methods in Educational Research K. Niglas, Tallinn Pedagogical University, Estonia The ongoing debates about the relationship between qualitative and quantitative methods and the vigorous critique of the quality and utility of educational research led to a wish to contribute to the potential enhancement of research practice by systematically analysing several scorching questions of research methodology. The preliminary study, undertaken at Cambridge University during 1998-1999, focused on the paradigmatic confrontation of qualitative and quantitative approaches. The analysis of various texts on methodological issues and a small-scale empirical investigation suggested that the dichotomous nature of educational research, advocated by certain influential writers and many textbooks, is not characteristic of the actual research practice, nor proved the incompatibility thesis to be an ideal by which educational research could be enhanced. The results of a preliminary study, stating that quantitative and qualitative methodologies are not mutually exclusive and incompatible paradigms, and depending on the nature and the complexity of particular research problem, the design of a study can be either qualitative, quantitative or a combination of both, form the basic assumption for the main study. As the field of combined designs (also called as “mixed methods research”) is only in its early development, this study sets the aim to enhance and extend the existing systematic knowledge about the ways combined designs can be and are used in research practice, to explore possible justifications for a new kind of practice and to analyse the implications that might have in the context of educational research. 148 The main study integrates theoretical knowledge on and recent developments in the combined use of qualitative and quantitative approaches with the results of an original empirical study. The empirical part of the inquiry adopts a content analytical method with the aim to analyse the actual use of combined designs to filter out common ways quantitative and qualitative approaches are merged within different research projects. A meta-analysis was conducted, focused on methodological aspects of a purposive but relatively representative sample of articles, presenting the results of studies using various combined designs published in English-language academic educational research journals. A critical analysis of the existing knowledge and a synthesis of new ideas is characteristic to all stages of the study, but the last chapter integrates the emerged theoretical ideas with empirical results and relates the results of the present inquiry to the latest developments in the field. It also broaches the implications of the results for research practice and for the ways in which the methodology of educational research could be introduced to a new generation of educators and educational researchers, providing thereby an added value to the study undertaken. The main merits of the present investigation are the elucidation and clarification of complex methodological issues in the field of educational research, the regulation and harmonisation of related terminology, the systematisation of the classification of combined designs, providing a framework for the use of combined designs by emphasising the role of research purposes and aims on choosing a suitable design for a particular study and by eliciting the possible justifications and purposes for the combined use of qualitative and quantitative approaches, but also underlying the importance of coherence of the research design and indicating some potential inconsistencies of different combined designs. Keywords: educational research, research methodology, quantitative and qualitative methods, combined research designs, mixed methods research. SESSION: DISCOURSIVE ETHNOGRAPHY Organizor: Giampietro Gobo1 and Anne Ryen2 Contact: 1Dept. of Social and Political Studies, University of Milan, Via Conservatorio 7 - 20122 Milano, tel. ++39-02-503-18813 (office); fax: ++39-02-503-18840, [email protected] 2 Agder College, Norway Keywords: discourse, conversation, ethnography Abstract This session is devoted to those studies which combine ethnographic and discourse analytic methods. Against traditional approaches which promoted on the one hand purely discourse analytic methods, or on the other hand purely observational studies, this session aims to gather studies which investigate everyday life interactions, work practices or linguistic activities, focusing both the relationship between talk and its ethnographic context, and the relationship between observational data and participants’ accounts. These studies believe that ethnography without talk could be misleading, and talk without ethnography is blind. Will be particularly welcome studies based on Fieldwork observations with interviews with participants Fieldwork observations accompanied by transcripts of conversation or discourse The purpose of this session is not to show substantive results only but also the benefits of a such approach, methodological problems faced by these studies, different ways of managing and analysing this kind of data. 149 Using Conversation Analysis to Test Questionnaires R.J.W.M. Vis-Visschers, Statistics Netherlands, Division of Technology and Facilities, Department Methods and Informatics, P.O. Box 4481, 6401 CZ Heerlen, room R414, Tel +31( 0)45 570 6188, [email protected] At the Questionnaire Laboratory of Statistics Netherlands questionnaires are tested to asses the data quality and to investigate whether the questions work as they are intended. Bad questionnaires not only can cause invalid data, but can also cause an extra strain on the respondent and the interviewer. The goal of questionnaire testing is to achieve respondent and interviewer friendly questions and questions that result in valid data. Traditional methods to test questionnaires are among other things expert reviews, focus groups and cognitive interviews. One mutual drawback of these methods is that they intrude upon the natural course of the interview. In search for an unobtrusive test method, the usefulness of discourse analytic methods for testing questionnaires was investigated. In 2002 and 2003 interviews from two CATIquestionnaires were taped and the dialogues were transcribed and analysed. In this paper I will discuss the usefulness of discourse analytic methods in analysing telephone interviews and in identifying problems in questionnaires. This will be illustrated by means of transcribed dialogues. I will point out the problems we faced. Among other things: How to tape the interviews, and how to store them safely? Is there a way to make the method less labour intensive? How can the method be combined with other pre-testing methods? To conclude, I will present our solutions to the problems, and how we use discourse analytic methods today. Keywords: discourse analysis, pre-testing methods, questionnaire laboratory For “she who knows who she is…” Combining Ethnography and Conversation Analysis in the Study of Internet Newsgroups A. Vayreda1, F. Nuñez1, E. Ardevol1 and Charles Antaki2 1 AlmibarLab, Universitat Oberta de Catalunya, Spain, [email protected], [email protected], [email protected] 2 Loughborough University, UK, [email protected] Ethnographic methods have been proved suitable and successful for studying online social interaction and the emergence of virtual communities. Our proposal is to combine ethnographic approach and conversational analysis to study computer mediated communication as social practice. CA work has mostly been done on chat-rooms, because have been seen more like a conversation, but in this presentation we extend its application to the domain of newsgroups (forums). We focus on the users' deployment of two crucial features of sequential interaction: determination of next speaker, and accountability of speaker's design of their turn. We argue that, in spite of the differences between the asynchronous, written medium of newsgroups and the synchronous, oral medium of spoken exchanges, both of these features are made visibly present in users' postings. In examining a short thread of postings to a public-access site, we observe how the analyst can profit by recognising how users exploit both of these features. We show that the relatively 'free' determination of messagerecipient allows the user to launch generally-respondable turns which, in face-to-face (FTF) or telephone (T) conversation, would require response by a single, given individual. This affords a range of implications, and allows a range of social actions, impossible (or very difficult) in FTF or T conversations. The second observation we make is that the looser requirements of accountability permit "intimate" turn-designs which, in FTF and T conversation, would be non-normative; again, this allows a wide range of actions not possible or difficult in other media. Our case study is an intensive analysis of a short thread of postings, revealing the fine-grained detail of a user's accomplishment of social actions (including 'declaring love') by taking advantage of the communicative affordances of the 150 medium, and how, in subsequent messages, other users ratify the original message’s use of the medium. The ethnographic understanding of internet activity has, we claim, much to learn from CA's close attention to turn design and sequence, and can profit from its analysis of newsgroup 'turns' as vehicles for social action. Keywords: ethnography, conversational analysis, Internet newsgroups, computer mediated communication, qualitative methods Talk of Stakeholders, Talk of Players. Goffman’s Contributions to the Use of a RPG as a Social Investigation Tool William’s Daré, Cirad Réunion, 100 route de la Rivière des pluies, 97490 Sainte Clotilde, France (La Réunion), Tel: (33) (0)262 92 24 57, [email protected] Since J. Huizinga and R. Caillois, playing is supposed to be outside the reality. Associations of multiagent systems and role-playing games (RPG) are used in participatory processes as group decision support tools to favor exchanges of information between stakeholders. In these approaches RPG are supposed to impact social management of common resources. But how this could happen if gaming is not linking with reality ? A multi-agent system and a role-playing game were built up in a companion modelling approach to tackle the viability of Senegalese irrigated systems. They are used with villagers who farm in the same irrigated system. In this article, we show that reality and game are linked in the playing of role-players. Even if in a RPG, stakeholders are placed in a virtual world where roles and rules are characterized, they used their social relationship to evolve in the role playing game. Doing so, we show that the game can help the researcher to grasp social relationships between stakeholders. After introducing the context of the research, we focus on a crisis appeared in one game session. Talks of players interacting in the game are registered with a camcorder then translated into French. We use Goffman’s discourse analysis to describe the “front” of players developed to solve the water management issue, the interactive behaviors, the dynamics of social roles regarding to the evolution of the game situation. Then, we describe a decision process to solve a real crisis appeared in the management of the irrigated system. Thanks to collective interviews and history life studies, we analyze the stakes and the interests of the stakeholders involved in the conflict. In the third part, we put face to face information obtained in both context (reality and play). We show how the analysis of game sessions was helpful to reveal relationships between stakeholders. We conclude that the RPG can be considered as a dramaturgical scene. Ethnography interviews shed light on the interactions registered in the game. In this way, Goffman is helpful to analyze talks of players in the game and compare with talks in reality. So the game can be a new tool to investigate social reality. Keywords: Goffman, discourse analysis, Role-playing game, Senegal The Methodological Discource about the Complex Discourse of “The Child’s best Interest” Turid Midjo, Department of Social Work and Health Science, NTNU, NO-7491 Trondheim, Norway, [email protected] The paper explicates some challenges and problems that confront the researcher when analysing field observation, interview information and content analyse as a combined methodological strategy. More precisely my paper will discuss the following question: What kind of methodological discourses take place in the knowledge production process? The background for raising this question is my ongoing study where I’m using observation and interview as data gathering methods, as well as information from written documents/caseloads. 151 The aim of my study is to explore the communication process between the social worker, the parents and the child in order to understand how “participation practises” take place in the context of the Child Welfare Services in Norway. The “participation practises” are understood as ongoing discourses and one aim is to analyse how the discourses tend to keep, reconstruct and change the involved actors’ perception of what is to be understood as “The child’s best interest”. The study gives special attention to the relation between aspects of control, power and disciplination and aspects of empathy, dialog and reflexivity. To study the communication process I'm following "inquiry cases" in four municipal child welfare services – that means I do observation in all the settings where the social worker meets with the family members and/or other professionals. Through and after this process I interview the social worker, the parents and the child. I will follow between 10 -15 cases due on the information the cases give. So what methodological challenges do I confront in making one coherent research out of these three kinds of data? What does it mean to combine them? In the paper I will present my reflection about my own “way of doing” the analyse and the knowledge production. Keywords: child welfare, content analysis, ethnography, discourse analysis, interview. SESSION: THE ACHIEVEMENTS QUALITATIVE COMPUTING AND CHALLENGES OF Organizer: Lyn Richards, Contact: Research Services, QSR International, Australia, [email protected] Keywords: qualitative research, computing, software, mixed methods Abstract Fifteen years after the first international conference on qualitative computing (Surrey, 1989), there is little scholarly debate about qualitative software. This session is to host such debate – not to compare software products but to assess methodological change on the assumption that such change must be critiqued and directed. Papers are invited for a session to stimulate discussion of both the achievements and the challenges of software development for qualitative research. Topics include What has changed, and what has not changed, in qualitative methods? Where are the effects of software to be seen and how are they to be dealt with? How did early software affect methods? How to assess the adequacy of developments by which software has moved beyond these methods? What do we know of new qualitative techniques, encouraged by software? Developments and challenges of mixed qualitative-quantitative techniques, facilitated and encouraged by software development; Present and future issues of reliability and rigour of software-supported qualitative research; Adequacy of the variety and range of programs for the research fields requiring them. Qualitative Computing and Theory Building: A Methodological Debate César A. Cisneros Puebla, Autonomous Metropolitan University, Mexico, [email protected] The construction of theories by computer has been a long-standing objective of Artificial Intelligence research since the 1960's as well as an important task in qualitative research. Therefore, to understand the complex process of theory building is an essential topic to Qualitative Computing. Nowadays, a variety of qualitative data analysis programs are available, which are proposed as an alternative to 152 code-and-retrieve software and which also had been addressed as 'third generation' software for qualitative analysis in the middle of the last decade. But now we are facing new trends in the design of programs, new trends in using this qualitative software, also new challenges, tasks and goals are emerged. In this presentation the current debate about the theoretical links between qualitative computing and qualitative methods is analyzed focusing on the cognitive process involved in theory building. The new trends are discussed to highlight the role of conceptualization rather than description around the discussion on Grounded Theory Approach, and then the debate about differences and similarities between Grounded theory and Qualitative Data Analysis is the framework to the analysis. Keywords: qualitative computing, theory building, cognition, abduction, fuzziness Why Qualitative Computing is radically changing Qualitative Research Tom Richards, QSR International, [email protected] Qualitative computing software doesn’t merely support the qualitative data analysis (QDA) techniques carried out in earlier technologies – pen, paper, copier and even word processor. Nor does it simply speed up such techniques. It goes beyond them by providing tools for new qualitative data organization techniques, new analysis techniques, and new mixed-method techniques. New technologies have a habit of enabling people to do new things. Think of writing, think of motorized transport, or the piano, or telephones. It can take people some time to realize new powers the new technology makes possible – indeed often the inventors don’t foresee what wonderful things will be done with their inventions. Some qualitative researchers for example, coming from the earlier technologies, try to fit qualitative computing into their earlier ways of working – a sort of Procrustean Bed approach with all the agonies associated with it. Or they may simply bypass the newer and more powerful procedures that qualitative computing enables – which is like walking a bicycle rather than pedalling it. In this paper I will look at some of the most significant and interesting new ways of working. I will try to explain what the tools are really intended for, and by contrast some of their common and painful abuses. I will look at new types of data, at extensions of the very concept of data, and what can be done with them. I will discuss organization and management tools, and how they are the key to making fruitful analyses possible. I will look at various analytical processes, and what they can and cannot deliver (including delivering enough rope to hang yourself by). In all of this I will be looking at qualitative computing as the locus for the entire research project, not just the data analysis part. Forget CAQDAS – the game is not QDA by computer, but QR by computer. Keywords: qualitative research, qualitative analysis, software, computer Adopting and Adapting to Qualitative Software: Social And Methodological Considerations from the Periphery Diógenes Carvajal, Universidad de Los Andes, Bogotá, Colombia, [email protected], [email protected]. Qualitative analysis software is nowadays commonly used by most of qualitative researchers worldwide. These programmes have become very helpful in systematising and analysing many types of qualitative documents: text, image, sound, and video, making obsolete the antique clerical technique of cutting and pasting manually. Parallel to the growing interest of qualitative researchers in such software, every day there are more training workshops to cover the also growing training demand. This is a global process: researchers from all backgrounds participate on it. Therefore, the use of qualitative software has become a real necessity for us researchers from the periphery in order to relate to the methodological and technical mainstreams of the metropolis. But trying to be global is not as easy; one does not become global by copying the metropolis’ mainstreaming. And one field where 153 this is broadly seen is in the use of qualitative software. Does a researcher from the periphery recur to qualitative software to fulfil her necessities or just to be a la mode? In this sense, there are two aspects I want to mention; a pragmatic one, and a methodological one. The pragmatic one refers to the type of training in qualitative software that is being offered. The methodological aspect refers to the relation between training workshops and the current approach that novice qualitative software users have before facing qualitative data. Inexperienced researchers with vague or insufficient training in qualitative research are at risk of quantitatively approaching qualitative software. Thus, aspects like validity, reliability, rigorousness, and groundness are considered from a quantitative point of view, and are measured as if they were quantifiable. The qualitative analysis process is left aside. Considering the above, in this paper I will discuss three topics: First, the way we, researchers from the periphery, are compelled to adopt and adapt to technical demands from the metropolis, and its “social” consequences; second, how this relates to the growing supply of training workshops in qualitative software, and its methodological consequences, specially in relation to what I call the “quantitativation” of qualitative data. Finally, I will outline some key elements to lead future reflection on this topic, and will pose possible solutions. Keywords: Periphery, qualitative research, qualitative software, quantitative methodology, training, validity, reliability, rigorousness. A Context-Oriented Approach to Design of Computer-Assisted Qualitative Data Analysis Software Gennady Kanyguine, Sociological institute of Russian Academy of Sciences, [email protected] By what means CAQDAS can retrieve with help of the user more information from data as compared with the conventional coding and retrieval (C&R) operations? A basic tool of C&R is an imitation of the notion clarification in the shape of the relationship ‘one-tomany’. Ex. in any tree-like component of CAQDAS such clarification is represented as denotation for a given node (linked to the notion to be explicated) of its children (bound to the clarifying concepts). Actually, such CAQDAS clarification does not fit the practical act of explicating. On the practical issue it is always supposed that the relationship ‘one-to-many’ is accompanied by circumstances in evidence which appear as a component of the relationship’s meaning. Ex. SOCIAL STATUS is INCOME, EDUCATION, OCCUPATION, yet not as a matter of fact but under certain conditions that should be indicated at the moment of clarification. In this example such indication might be effectuated, for instance, with the notion of SOCIAL STUDY. Generally, it is proposed to imitate the explicating act in the shape of a special unit (context-fixed elucidation) that couples the relationship ‘one-to-many’ and an appropriate notion (context) suggested by researcher to denote the conditions the explicating act holds. How to make C&R operations as theory-building tools more consistent? A key point of applying CAQDAS as theory-building tools is secondary analysis of the codes resulted from the separation of informational streams (textual, visual, etc.) accomplished by means of C&R operations. In order to enrich inquiry’s techniques accessible to researcher in such an analysis, the software, in addition to the codes, gives him/her access to complementary descriptors, for example, NVivo’s attributes or MAXqda’s variables. Unfortunately, the codes and descriptors are of unmatched derivations. Literally, the heuristic problem of stream separation and formal operations on the codes obtained. To solve the problem of differing derivations in principle, a notion’s model (NM) is proposed. Unlike the code and complementary descriptor NM implements by itself all basic manipulations performed by researcher with notion in the process of theory-building - associating, definition in terms of data, searching, etc. Because MN grows up as a universal basic component of all computer-assisted operations of CAQDAS and the indication of the context by researcher is mandatory at every notion’s elucidation, the approach discussed extends considerably the set of relationships between the appearing computer analogues of the research notion as compared with the traditional tools of C&R. Keywords: problem-oriented qualitative data analysis software 154 SESSION: COMPUTER AIDED CONTENT ANALYSIS: RESEARCH, METHODS AND ITS FUTURE Organizor: Harald Klein Contact: Social Science Consulting, Brückengasse 12, 07407 Rudolstadt, Germany Department of Sociology, Friedrich-Schiller University, 07740 Jena, [email protected] Germany, Abstract The aim of this session is to present papers dealing with application fields of computer aided content analyis. Such papers should focus on the application of the method, the hypotheses, its operationalisation, and the results. Also papers concentrating on the methodogic and operationalisation problems are welcome, e.g. how to encounter problems of negation, ambiguity, or other linguistic problems. Otzher topics maybe software comparisons., new trends in content analysis, standardised category systems etc. The time for each paper will be ca. 15-20 minutes including a discussion. Part I Content Analysis of Short, Structured Texts: The Need for Multifaceted Strategies Harold W. Bashor, American University of Paris, France How can one analyze short, structured texts? Content analysis provides a systematic, replicable technique where validation normally takes the form of triangulation. Such mutual convergence lends credibility to the findings by incorporating multiple sources of data, methods, investigators, or theories. However, this paper examines the strengths and weaknesses of triangulation by comparing data collection methods for long, unstructured texts with short, structured texts. Considering the limitations of traditional triangulation, a multifaceted strategy has been tailored to meet the needs of specific, micro-linguistic analysis of a particular genre of text:international treaty preambles. In addition, a number of different perspectives to discern the following key elements and the dynamics of text have been examined: (1) purpose or overall agenda, (2) agent and his/her socialcultural standpoint, (3) the intended audience, (4) the socio-political context, and (5) most relevant to this paper, the language itself: syntax, semantics, style, rhetoric, and structure. The combination of these elements requires a multifaceted strategy that conventional triangulation often ignores in content analysis. This paper proposes a multi-layered approach with three different software programs. Although it has been argued in this paper that mutual validation of results is the ultimate goal, the capacity for methods to complement each other in the drive towards 'comprehensiveness' should not be assumed; nor should corroboration between methods be automatically considered unproblematic, and therefore precluding the need for further nvestigation. Extraction of Semantic Information: New Models and old Thesauri Jan Kleinnijenhuis, presumably with Wouter H. van Atteveldt and/or Ivar Vermeulen Vrije Universiteit Amsterdam, department of Communication Science, De Boelelaan 1081c, 1081 HV Amsterdam, tel. +31 20 444 6827 (secr.6854; fax 6820), [email protected]. The profusion of electronically available documents on the web has given rise to the quest for automatic representations of web content that enable comprehensive understandings of the content on the Web. The idea of a “semantic web” entails that documents may be marked up with metainformation in markup languages such as RDF in such a way that meaningful summaries, visualisations and classifications of large sets of texts are easily obtained (Daconta, Obrst, & Smith, 2003). Insights regarding the extraction of the meaning of large sets of texts may also be gained from 155 corpus linguistics (Manning & Schütze, 2002) or from the field of content analysis in the social sciences (Roberts, 1997) This paper presents a survey of the problems surrounding the automated extraction of information on relationships between objects such as organizations, persons and (economic or political) issues. Three approaches will be described and discussed in more detail: 1. the maximum entropy approach to identifify references to specific organizations, persons and issues (van Atteveldt, 2003). 2. the usage of electronically available thesauri (e.g. Roget’s thesaurus) to identify positive or negative relationships between these organizations, persons and issues .(Kleinnijenhuis & Vliegenthart, 2003) 3. the usage of dimensional reduction techniques such as ‘latent content analysis’ to obtain more robust meanings (Landauer & Dumais, 1997) These approaches will be applied to extract information on competitive relations between businesses and between political parties from the web and from electronically available newspaper archives (LexisNexis). References Daconta, M. C., Obrst, L. J., & Smith, K. T. (2003). The semantic web. New York: Wiley. Kleinnijenhuis, J., & Vliegenthart, R. (2003). Wetenschap in de pers:. Paper presented at the Workshop Science Alliance / Stichting Weten on Science Communication, Maastricht, 16 en 17 december 2003. Landauer, T. K., & Dumais, S. T. (1997). A solution to Plato's problem: The Latent Semantic Analysis theory of the acquisition, induction, and representation of knowledge. Psychological Review, 104, 211-240. Manning, C. D., & Schütze, H. (2002). Foundations of Statistical Natural Language Processing. Cambridge MA / London: MIT press. Roberts, C. W. (Ed.). (1997). Text analysis for the social sciences: methods for drawing statistical inferences from texts and transcripts. Mahwah, NJ: Erlbaum. van Atteveldt, W. H. (2003). Information Extraction using Sequencing with Large Label Sets. Unpublished Master Thesis, University of Edinburgh, Edinburgh. Why People Reject Super-fair Offers. Evaluating Video Experiments on Ultimatum Bargaining Run in PR China Heike Hennig-Schmidt1, Zhu-Yu Li2 and Chaoliang Yang3 1 Laboratory of Experimental Economics, University of Bonn, Germany, Tel.: +49 228 73 9195, Fax: +49 228 73 http://webserver.econ1.uni-bonn.de/ 2 Economic College, Sichuan University, 29, Wangjiang Road, +86 28 5417673, [email protected] 3 Laboratory of Experimental Economics, University of Bonn, Germany Adenauerallee 24-42, D-53113 Bonn, 9193, [email protected], Chengdu 610064, PR China; Tel/Fax: Adenauerallee 24-42, D-53113 Bonn, Subjects in Ultimatum Bargaining Experiments have been found not only to reject offers that are to their disadvantage. They also turn down offers that are to their benefit. It has been argued that rejecting super-fair offers indicate subjects’ misunderstand of the strategic features of the game. Other authors attribute the refusal of generous offers to the culture of status-seeking through gift-giving: Accepting gifts might imply strong obligations to reciprocate in the future which the receiver wants to avoid. To investigate subjects’ own reasoning, we video taped discussions of subject groups acting together as either a sender or a receiver in an Ultimatum Game experiment. We subsequently content analysed the discussions. Our study shows that the above-mentioned views on the driving forces for rejecting super-fair offers are far too narrow. Subjects’ motives range from other-regarding views of mutual fairness to the preference for certain numbers. The paper gives a detailed analysis of the arguments found and their impact on behavior. 156 Part II Measurement Instrument Choice for Tracking Email Communication Marguerite Kolar, Australia The purpose of this research is to measure the progress towards stated change objectives through weekly email communications from a single source. Measuring the progress of change provides researchers and practitioners alike, the opportunity to evaluate the extent to which messages are reflecting the progress of the change and whether the messages are reinforcing the organisation’s change objectives. This paper identifies potential methodologies and tools for the analysis of the email data set. Discussion on various message analysis techniques, the pros and cons of using computerised tools during qualitative analysis and identification of the most appropriate methodology and computerised tool for this research are covered. Criteria for determination of the appropriate methodology include ability to qualitatively analyse data to identify mutually exclusive themes, ability to compare themes against researcher determined concepts (e.g. transformation change characteristics), the ability to quantitatively analyse data to identify trends and availability of computerised tools associated with the methodology, as the data set is already on a computer. Content Analysis is determined to be the most appropriate methodology based on these criteria and the VBPro tool is identified as the most appropriate tool. Inductive versus Deductive Approach: Computer-assisted Content Analysis without Dictionary? Cornelia Zuell and Juliane Landmann, ZUMA, Mannheim, Germany Dictionary construction is a very time-consuming and tedious task. Furthermore, dictionary-free approaches promise to be able to analyse the texts, i.e. to find interrelationships among words and to identify the underlying concepts in a text. We will discuss such a dictionary-free approach and its advantages and disadvantages using texts from two different newspapers: the Frankfurter Allgemeine Zeitung and the New York Times. We analysed articles published at the beginning of the war in Iraq (March 18 to 20, 2003) and articles published at the end (April 7 to 9) in the two newspapers. Overview on Standardised Category Systems in Computer aided Content Analysis Harald Klein, Social Science Consulting, Brückengasse 12, 07407 Rudolstadt, Germany, Tel/ Fax: +49 3672 488494 One of the most time consuming parts in a content analysis ist he construction of a category system. Most studies have to develop their own category system, and often this will never be used again. This paper will show thaat standardised category system for general and speicifc research purposes are available. The category systems and their categories will be discussed, and one focus is which category systems can be used with which software. 157 Part III TCA: General Purpose Software for Semantic and Thematic Text Analysis Carl W. Roberts, 210D Snedecor Hall Tel: 515/294-8929; fax: 294-4040), 217A East Hall, Tel. 515/294-2114; fax: 294-2303), Iowa State University (Ames, IA 50011; USA) Text analysts generally use one of two strategies to develop a dictionary for classifying texts' words and phrases into "meaning categories" prior to statistical analysis. Either they use a dictionary developed by someone else, or they develop their own. TCA was developed as an aide to text analysts of the latter type, namely to those who use key-word-in-context routines to construct dictionaries with meaning categories that correspond more to the texts under analysis than to the analyst's preconceptions. The program enables users to generate data suitable not only for thematic text analysis (i.e., analyses of word/phrase counts within blocks of texts), but also for the type of semantic text analysis (i.e., analyses of words/phrases' relations within a semantic grammar) developed by the author. In addition, TCA has powerful tools for automatically (or manually) parsing texts into sentences, paragraphs, and larger text blocks, plus a capacity to input contextual data regarding these blocks. Routinizing Frame Analysis through the Use of CAQDAS Thomas König, Department of Social Sciences, Loughborough University, Loughborough, LE11 3TU, United Kingdom, Brockington Building U.01.03, Tel. +44 (0)1509 223730, fax +44 (0)1509 223944, [email protected] Even though frame analysis has become a popular analytical framework in media studies and social movement research, the methodological underpinnings of the empirical identification of frames lack systematization and have consequently remained underdeveloped. This paper consolidates recent advances in the empirical measurement of frames and explores, in how far computer-assisted qualitative data analysis software (CAQDAS) can extend on these methodologies. Because framing has become a fairly widely used but ill-defined concept, the paper will start with a brief delineation of framing theory as it is understood for present purposes. Next, current attempts to measure frames empirically in a systematic fashion will be discussed and a methodology, which synthesizes some of these approaches will be proposed. This methodology attempts in a first step to draw on existing knowledge on metanarratives to avoid a purely inductive identification of frames. Additionally, automatic word mapping tools such as Leximancer, Sphinx Survey Lexica are suggested as interpretative aids. In a second step, the analyst identifies a set of keywords, key phrases, and possibly audial or visual symbols that indicate frames in his data. These indicators are then used in a third step to semi-automatically identify frames in the remainder of the data. Keywords that might acquire different meanings in different contexts are inspected in their contexts by the analyst, who decide on their coding. This method avoids both the rigidities that come with fully automatic keyword clustering, which may lead to the inclusion non-interpretable keywords as well as the exclusion of so-called stop words such as prepositions and articles, which under certain circumstances might indeed be the strongest indicators for certain frames. At the same time it allows for a degree of routinization and systematization in frame analysis, whose quality has notoriously depended on the creativity of the framing researchers. Six CAQDAS – ATLAS.ti, Kwalitan, MAXqda, NVivo, Qualrus, and QDA Miner – are examined with respect to their usability in this type of framing research. 158 Qualitative Software 2004 and Beyond Raymond C. Maietta1 and Sharlene Hesse-Biber2 1 ResearchTalk Inc., 1650 Sycamore Ave, Suite 53, Bohemia, NY 11716, Tel: +1-631-218-8875, www.researchtalk.com 2 Department of Sociology, Boston College, Chestnut Hill, Massachusetts 02167, Tel: (617) 552-4139, Fax: (617) 552-4283, [email protected] In its current state, qualitative software accommodates analysis of data in text, graphic, audio and video format and supports a diverse range of data types and analytic approaches from quantitative verification to loose heuristic analysis. Dr. Maietta and Dr. Hesse-Biber examine important questions that arise from this point organized into key sections of their co-edited volume When Methods Meets Technology: Important Issues in Qualitative Software Evolution: Methodological Implications - How does accommodation to such a large, and even conflicting, set of analysis approaches affect implications and impressions of software use? What impact does qualitative software use have on the field of qualitative analysis? Pedagogical Approaches - Is there one 'right way' to teach qualitative software? When should qualitative software be introduced to scholars? Can the careful introduction of qualitative software integration techniques inspire improved qualitative scholarship? Applied User Experiences - How is qualitative software being used within different disciplines and industries (academia, government, not-for-profit, and corporate)? What lessons can be learned to ensure careful and creative application of these products? The session concludes with debate that focuses on just how much more innovation is needed for qualitative software to meet the needs for current and future directions in qualitative analysis. Part IV Comparative Computerized Content Analysis of Party Manifestos: Methodological Considerations and Empirical Results Paul Pennings, Dept. of Political Science, DBL 859, Vrije Universiteit Amsterdam, De Boelelaan 1081-c, 1081 HV Amsterdam, The Netherlands, tel. +31.20.4446852, fax. +31.20.4446820, [email protected], www.fsw.vu.nl/~pennings Since the 1970s different types of content analyses have been applied to party manifestos. These documents present the main party policy positions during elections and, although hardly read by voters, they are crucial for determining where parties stand vis-á-vis each other. The recent digitalization of these documents opens up new possibilities for content analysis. This paper explores the possibilities and limitations, discusses some problems and pitfalls, assesses the validity and reliability of the coding technique and presents some preliminary results. The approach which is used in the paper is a dictionary-based content analysis. The dictionary holds nearly 6000 key words and covers nearly 20 areas of policy-making. It is designed analyse digitalised party manifestos in 20 countries in the period 1960-present. It will be shown that it is especially useful to analyse the position of parties on main conflict dimensions (left versus right etc.), but that it can also be used to study one particular issue. 159 Concepts of Animal Kingdom in Brazilian Television Advertise: a Comparative Content Analysis Ana Laura Arruda1 and Claudia Rosa Acevedo2 1 Faculda de Orlândia, Rua Moreira de Oliveira, 132. Alto da Boa Vista, CEP 14025-300, Ribeirão Preto, SP, Brasil, Tel: 55 16 623-1647 or 55 16 3826-9888, [email protected] 2 Centro Universitário Nove de Julho – Uninove, Av. Francisco Matarazzo, 612. Água Branca, CEP 05001-100, São Paulo, SP, Brasil, Tel: 55 11 3665-9300 or 55 11 55.75.25.31, Fax: (55 11) 36659302, [email protected] The relative absence of content analysis studies applied to Brazilian advertise, allied to the increasing interest about animals, created by social changes and environment conscientiousness actions, beyond the dispersion of human – animal relationships scientific works justify this research. With the purpose to identify how Brazilian advertise represents the various concepts about animal kingdom that exist in literature of reference, a set of these concepts categories was worked out. Using both classic and interpretive content analysis methodologies, two groups of invited judges examined and classified the same set of ads selected from a collection of Brazilian laureate commercials that comprehend the period of 1950 to 2000, and the results are compared. Although this research presents limitations material order, its results confirm the formulated hypothesis and validate the conceptual framework proposed here. A larger knowledge of the content of Brazilian advertise with animals may be useful in the generation of new studies that require the conceptual framework developed in this work. Information about Brazilian cultural values of the animal kingdom may become practical contributions for the economy, facilitating the process to create new products and services oriented to animal consumption. Instrumentally Rational Approaches to Health Care Quality in the U.S. Health Care System: From Professional Education to Practice Guidelines Amit Nigam, Management & Organizations and Sociology, Northwestern University, Chicago, USA To assess how professions can promote the rationalization of social life, and how institutional change in professional fields can shape their role in promoting rationalization, I examine changing conceptions of and approaches to quality medical care among American physicians during a period of institutional change in the U.S. health care system. This study draws on content analysis of discussions of health care quality, drawing on abstracts focused on healthcare quality, as categorized by the National Library of Medicine’s MEDLINE database, in the Journal of the American Medical Association from 1983 until the present. At adopts a grounded theory approach, beginning with inductive analysis of the abstracts to develop a list of keywords and codes. It does an automatic coding of the abstracts based on 66 codes. It uses partitition cluster analysis, guided by qualitative analysis of potential groupings of the abstracts, to identify distinct approaches towards healthcare quality in the abstracts. It then draws on qualitative analysis of the abstracts, by cluster, to develop theory about changing approaches to healthcare quality over time. Based on this analysis, this study argues that physicians’ approaches towards quality medical care shifts from an educational professional approach focused on the education and licensing of physicians and on physicians’ knowledge and skills towards a guideline/standards-based approach that emphasizes adherence to established guidelines or standards for treating particular diseases or conditions. 160 SESSION: SURVEYS THE RESPONSE PROCESS IN ESTABLISHMENT Organizor: Tony Hak Contact: Faculteit Bedrijfskunde/ Rotterdam School of Management, Erasmus University, Rotterdam, The Netherlands, [email protected] Keywords: establishment surveys; response process; ethnography; debriefing Abstract The survey questionnaire is a document type for which a gamut of specialized test procedures have been developed, of which ‘cognitive interviewing’ and ‘behavior coding’ are the most widely used. These methods focus on the interaction between the document and the respondent, and on the cognitive tasks posed by the contents and the design of the questionnaire. Usability testing has been applied to design aspects of electronic questionnaires, separately from the non-electronic aspects for which the traditional test procedures are used. Establishment surveys differ from other surveys in one important respect: respondents are not reporting for themselves as people but for an organization, which means that the interaction between the questionnaire and the respondents is more complex. In many cases, the questionnaire will be read, discussed and manipulated by different persons, which implies that the questionnaire will ‘travel’ from desk to desk, or from PC to PC. This applies both to the traditional paper form of the establishment survey questionnaire as well as to the increasingly more common electronic form. This characteristic of the establishment survey makes it useful, but challenging, to develop methods for researching the response process in establishment surveys. Papers in this session might focus on the characteristics of the response process in establishment surveys or on methods (such as participant observation, debriefing and other kinds of interviews) that are developed to study this process. The Evaluation of Establishment Survey Questionnaires: Experiences from combining Quantitative and Qualitative Methods Deirdre Giesen, Statistics Netherlands, Department of Methods & Informatics, [email protected] At Statistics Netherlands one of the most valued methods for evaluating and improving establishment survey questionnaires is on site observation of the response process. A careful registration of how respondents understand and answer the questions yields rich information about measurement problems of the questionnaire and often provides ideas for possible solutions for those problems. These data however come at a high cost: both the data collection and data analysis are time consuming and require highly qualified staff. As we can only visit a small number of different respondents we have to be careful in the selection of the type of respondents we want to visit and the questions to focus on. Furhtermore, we would like to have some quantitative indicators that can be considered representative for the population that can help us “weigh” our findings from the qualitative fieldwork. Currently, we are working on the evaluation of the Structural Business Survey (SBS) questionnaire. The SBS questionnaire is sent out yearly to around 80,000 establishments, covering all size classes and almost all branches. The SBS questionnaire consists of a uniform core part and 179 unique parts, containing questions specific for type of industry and size class. Given the large heterogeneity of both the respondents and the questions to be evaluated, the concerns stated above about the restrictions of on-site observation as a method are extra significant. We have therefore decided to use both quantitative and qualitative methods as a preparation and a background for our qualitative fieldwork. These methods include: focus group interviews with among others field officers and editing staff unit non-response analyses 161 item non-response analyses data quality indicators spontaneous comments by respondents about the questionnaires All methods are used to assess for which questionnaires, questions and types of respondents problems occur. In this paper I compare and discuss the findings of these different methods. Keywords: establishment surveys, questionnaire evaluation, mixing quantitative & qualitative methods, response model. Redesigning Statistics Canada’s Annual Survey of Manufactures – Overview of Respondent Testing using Cognitive Interview Techniques Kevin Roberts, Annual Survey of Manufactures, Manufacturing, Construction and Energy Division Statistics Canada, 11-B6 Jean-Talon Building, (613) 951-6927, [email protected] The Annual Survey of Manufactures and Logging is conducted every year by Statistics Canada to obtain important information on the manufacturing sector of the Canadian economy. Manufacturing establishments in Canada are required to provide information on such aspects of their operations as sales, costs/expenses, salaries and wages and commodity detail. As part of an agency streamlining initiative the survey is being re-designed for reference year 2004. The main objectives are to implement measures leading to efficiencies in survey process and to reduce respondent burden. Consequently, the questionnaire has been re-designed to include a “user-friendly” set of financial variables and a new commodity classification structure that significantly reduces the number of categories collected. As part of this process, a large respondent test was conducted that involved using “one-on-one” cognitive testing techniques. Test participants selected were representatives from enterprises involved in various types of manufacturing. In-depth interviews were conducted at their place of business in several large cities across Canada. The objectives of the test were to obtain feedback from respondents on their overall impressions and reactions to the content of the new form. During the meeting, the respondent was asked to “think aloud” as he/she was completing the questionnaire. This technique allowed us to determine respondents’ understanding of specific concepts, questions and terminology. In addition, the ability of the respondent to answer the questions in the context of their own operations and accounting procedures needed to be measured. The paper reviews the results and experience gained using this qualitative testing method. Keywords: Respondent Testing, Cognitive Interviewing, Establishment Survey The Use Of Qualitative Interviewing To Determine The Acceptability And Take Up Of Web Data Collection Instruments for Mandatory Business Surveys Zoë Dowling, University of Surrey, Guilford, UK, [email protected] Many governments require establishments to provide details of their business activities. These details are often gathered via means of a survey by the country’s National Statistical Institution (NSI). Whilst responding to these surveys can be a mandatory requirement (for example, as in the United Kingdom), the NSI works within the parameters of minimising response burden whilst maintaining acceptable levels of data quality amongst other concerns. With ever-growing Internet connectivity within businesses the World Wide Web presents a new mode for data collection, one that appears to offer many advantages. For example in terms of response time and survey costs, amongst others. However, should NSIs offer the option to respond via the Web, the question remains whether businesses perceive this as an acceptable means to respond to a survey? In order to best address this question, series of qualitative, face-to-face interviews with business, which have experience of responding to mandatory surveys, was undertaken in the first half of 2004. More specifically, 30-40 semi-structured interviews were carried out within1 businesses to discuss 162 three main themes, centred around 1) their experience in responding to a survey, 2) their use and perception of the Internet, and 3) design and functionality issues within Web questionnaires. The interviews included the use of visual prompts – a paper version of a mandatory survey and three alternative Web versions of a section of the questionnaire. This paper will discuss the successes and failures of the methodological approach taken to address the research question. In particular, it will evaluate the use of the Web illustrations as visual prompts within the interviews. This paper arises from on-going PhD research at the University of Surrey. It funded by the UK Economic and Social Research Council with the UK Office for National Statistics as collaborating partners. The views expressed in this paper are entirely my own and do not necessarily represent those of my funders. Keywords: Establishment surveys, Web data collection, qualitative interviews, visual prompts 1 That is to say, the interviews will take places on the premises of the participating businesses The Contact and Response Process in Business Surveys: Lessons from a Multimode Survey of Employers in the UK Peter Lynn and Emanuela Sala, Institute for Social and Economic Research, University of Essex, UK, [email protected], [email protected] On surveys of businesses, the processes of making contact and obtaining co-operation are quite different from those on more frequently studied types of surveys, such as those of households or private individuals. We describe experiences and outcomes on a business survey and discuss lessons that can be learnt in order to maximise contact and co-operation rates on future surveys. The survey in question has some valuable characteristics, including a wealth of auxiliary information. It also presents an interesting perspective on issues of privacy and confidentiality. As part of a large-scale methodological study known as ISMIE (Jäckle et al, 2004), a survey was carried out in 2003 amongst a sample of employing organisations. The purpose was to obtain validating information regarding employees who had taken part in a survey interview, answering questions about occupation, industry, working hours, pay, pension and related issues. In the initial survey interview, consenting respondents provided contact details for their employer. The survey of employing organisations was initially carried out by post, including two reminder mailings. This achieved a 51% response rate. Employers who did not reply to the postal survey were subsequently approached by telephone. If possible, an attempt was made to collect the required information over the phone; alternatively, another copy of the questionnaire could be mailed or faxed if requested. This boosted the response rate to 72%. We analyse the extent to which each mode (post, telephone, fax) influenced response rate and nonresponse bias. We also analyse reasons for refusals, which were recorded for the telephone stage and have been coded where possible for the postal stage. We then focus on the telephone stage, for which detailed systematic information was collected about the process of attempting to make contact and gain response. This enables us to study the process and the relationship with the final outcome. Additionally, we draw upon information obtained in debriefing interviews with the telephone interviewers. We present evidence of the number and distribution of calls made, broken down into three phases: making contact with the organisation, making contact with the respondent, and completing the questionnaire. Several outcomes were possible at each call and we present indications of the prevalence and consequences of each. It was often necessary to get past “gatekeepers” such as receptionists or secretaries and we will describe some strategies that were adopted to achieve cooperation. With respect to completing the questionnaire, we investigate the extent to which different sections were completed by different people, the nature of the questions for which a different respondent was often needed, and we identify question types that employers were relatively more likely to be willing to complete by phone, rather than by mail or fax. A common request was for the employer to be sent (typically by fax) a copy of the signed consent form that had been provided to the 163 survey organisation by the employee. Overall, the response process during the telephone phase of the survey was typically long and complicated: we identify common sequences of contact types. Several practical conclusions are drawn regarding the design of surveys of this kind and the organisation of the data collection effort. In summary, we conclude that respectable response rates and response quality can be achieved, but that this requires considerable effort and a flexible approach to contact and response, allowing the preferences and requirements of the respondent (employing organisation) to dictate the methods used by the survey organisation. Keywords: Business surveys, co-operation, non-contact, non-response, refusal, survey mode Using Respondent Site Visits to Improve the Design of Establishment Surveys Stanley R. Freedman and Robert Rutchik, Energy Information Administration, USA The testing of survey questionnaires with potential respondents is a time honored method for improving the data collection instrument before the study is fielded. One of the more frequently used techniques falls into the category of “cognitive interviewing.” The response model it addresses is that the subject is reporting information about themselves. In establishment surveys the response model is different and more complex. The respondent is an intermediary between the information that a researcher wants and data that a business maintains. In large companies or with complex surveys, several people may provide input to the survey. Even if the questions are clear and well understood, a business may not keep exactly the data a researcher is looking for, or keep it in a format that makes it difficult for the company to respond. These additional elements in the establishment survey response model require modification to some of the traditional techniques. This paper discusses three methods used by EIA to address the issues raised in the establishment survey model. They are pre-survey design visits, cognitive interviews on draft survey instruments, and respondent debriefings. Findings on the effectiveness of each technique will be presented based on work done at the Energy Information Administration. Keywords: Establishment surveys, survey design; respondent site visits STREAM: QUALITATIVE METHODS (1) SESSION: THE ETHICS AND SOCIAL RELATIONS OF RESEARCH Organizers: Carole Truman1 and Anne Ryen2 Contact: 1Lancaster University, UK, [email protected] 2 Agder College, Norway Keywords: Research relationships; cross-cultural research; ethical issues; Power in research Abstract This session invites papers which adopt critical perspectives on ethical issues in social research. We invite papers which seek to widen the debates which relate to ethical considerations in research beyond that prescribed by processes of formal/standardised ethical review. We particularly encourage perspectives which draw from the feminist and/or post-colonial literature on research methodology which situates ethics within the context of the social relationships of the research process. We encourage theoretical contributions, as well as those which explore ethical issues within the context of doing social research. 164 The Best and Worst of Research Relations David Chavis, Association for [email protected] the Study and Development of Community, USA, This proposed paper will explore the best and worst practices in relationship building among researchers (including evaluators) and consumers, particularly in the community development area. We often focus on best practices as a way to learn, but worst practices need to be brought out. Research and evaluation uses information that is a resource that a local community has which is often used to benefit “outside” interests, often at great cost to local community. Low income and racial minority communities in modernized and developing countries are often the subjects of research and their self help efforts the fodder for evaluations by “objective” outsiders. In order to justify the outsiders’ collection and use of this resource, local institutions and individuals need to be seen as inadequate and incapable. In this common condition, the relationship is greatly imbalanced and assumptions of adequacy-inadequacy are paramount. The resource of information is used to benefit outside sources, since locals could not adequately use it. An emerging practice of collaboration (under many different names) reflects a perspective that relationships between scientist/evaluator are based on mutual benefit and competence. Information as a resource can be mutually beneficial to enhance local capacity and broader knowledge in a field. This paper will present examples, both good and bad, on how to build local capacity for change in low income and minority communities as well as how to inadvertently even a well intended researcher can create harmful relations. Examples will be drawn from the authors experience as a scientist/ evaluator and consultant often used to “repair” severely damaged relationships among research and low income and minority communities. Keywords: community development, research capacity, collaborative research On Vulnerabilities and Responsibilities in the Qualitative Interview Anne Halvorsen, Agder University College, Norway, [email protected] In defining the goal, setting the scene and sampling/identifying interviewees the researcher/interviewer constitutes the Self in the Self-Other relationship in the qualitative interview. Her responsibility is to make the best possible conditions for obtaining access to data, often expressed through the innermost feelings, thoughts and ideas of the Other – the interviewee. In doing so she is expected to downplay her own feelings, reactions and responses to whatever occurs during the interview, or to use them in as an instrument in ‘collecting’ the data (Fog 1994). In both cases she is expected to in control of the interaction between the two, she should not fall into the category of “the controlled”. The Self, who asks for revelations from the Other, and thus makes the Other vulnerable, is supposed to remain herself invulnerable (Fine et al 2000). This is in line with the general description of the relationship between researcher and respondent. The interviewee is usually depicted as the less powerful of the two, even if she is expected to contribute the lion’s share to the dialogue, both in quantity and substance. I think this largely holds even for perspectives which allow for the interviewer to take a more active part in the interview. The purpose of the interview is, regardless of naturalistic, constructivist or other epistemological approach to lay bare (reality through) ‘the interviewee’s point of view’, also when the role of the researcher as a coproducer of data is acknowledged. During a pilot study involving former child protection clients, I interviewed a 20 year old woman who made me question and reflect on this Self-Other relationship. Meeting her also challenged me on a more personal level; I had to admit and rethink my preconceptions of “the powerless client”, and acknowledge that I still had a long way to go as far as my skills as a qualitative interviewer is concerned. The reason for this has, I think, much to do with vulnerability – both that of the interviewee, and my own. In this paper I will discuss how my expectations to an interview with a “former child protection client” and “a vulnerable Other” proved wrong, and how I came to learn more about “the roles and rules of 165 the interview” (and about myself) in the way I tackled or “mis-tackled” what happened. This might add some nuances to the discussion about the relationship of the Self (“the research community”) and the Others (marginalised groups) in research (Fine et.al. 2000). References Fog, Jette (1994): Med samtalen som utgangspunkt. Det kvalitative forskningsinterview. København: Akademisk forlag. Fine, Michelle, Lois Weis, Susan Weseen and Loonmun Wong (2000): “For Whom? Qualitative Research, Representations and Social Responsibilities” I Denzin, N.K. and Y.S. Lincoln (2000): Handbook of Qualitative Research. 2nd ed. Sage. Informed Consent, Gatekeepers and Go-Betweens Sue Heath, Vikki Charles, Graham Crow and Rose Wiles, University of Southampton, UK, [email protected] Gaining informed consent from people involved in research is central to ethical research practice. Yet is it ever possible to achieve genuine informed consent within the research process? Certain research contexts may make this more difficult to achieve than others, for example, whilst the power relations inherent in some types of research may have a similar impact. This paper reports on an Economic and Social Research Council-funded project concerned to identify current practice with regard to the process of obtaining informed consent by researchers working in the areas of childhood, youth, ageing, palliative care, mental health and learning difficulties. These areas have been highlighted as areas where research participants are often constructed as potentially vulnerable within the research process. As a consequence, their involvement in research tends to be mediated by gatekeepers and gobetweens. This paper will focus on the impact of these concerns in research with children and young people. Drawing upon the views and experiences of researchers working in these areas, it will highlight the tensions experienced by many researchers between a personal commitment to an ethical framework which seeks to prioritise the agency of children and young people (based upon notions of ‘ethical symmetry’ or ‘process consent’, for example) and the requirements imposed upon them by a variety of gatekeepers and go-betweens, particularly within institutional settings such as schools, which may run counter to this commitment. It is argued that processes of formal/standardised ethical review are often inadequate to cope with the realities of negotiating informed consent within the field. Keywords: Informed consent, research ethics, children and young people. Wading through “The Jungle” with my Main Informant. A critical discussion of ethical practice in cross-cultural ethnography. Anne Ryen, Agder University College, Servicebox 422, 4604 Kr.sand S, Norway, [email protected] This paper deals with ethical practice when working with a main informant in cross-cultural ethnographic research. The author points to two delicate ethical issues. The first deals with the professional as opposed to the private in field relations. The second issue refers to the tricky outcomes of the referred professional code of ethical field practice. With reference to data the author calls for a more critical, less orthodox stance on ethics and ethnography. With reference to the first issue the author argues that a strong differentiation between the professional and the private may emerge as unethical, colonial, hierarchical and impossible as they get intertwined. Spending hours and days together in the field, the borderline between a research - and a social relationship becomes blurred paradoxically motivated by the Western standard interview prescriptions. As a consequence research relations may be more turbulent than described in most research reports. In this way ethnographic research also includes experiences deviating from expectations to the standard portrait of the idealized informant and researcher as the background to standardized ethical guidelines. Rather than a failure, this aspect of ethnographic field research is argued to inform our research with interesting contextual data. 166 In the second issue the author claims that the relational outcomes of so-called successful ethical ethnographic work following standard ethical codes may be more delicate than assumed. The professional may work to deceive rather than to protect the informant. Our quest for data makes the solution to this issue more delicate. The paper ends by relating the discussion of ethical complexity in research practice to two more profound aspects of ethics and research across cultures. One vital question is the social processes by which ethical issues come to be constituted as ethical issues. This includes the question if Western research ethics are universal and also applicable in Sub-Saharan research contexts or if Western research codes simply represent a prolongation of previous colonial relations? And eventually, do ethical codes reflect the dominance of a certain ontology and epistemology, and if so, what are the alternatives and the risks? The paper draws on data from a research project on Asian businesses in East-Africa. The author has worked with her main informant, an Asian businessman, for 2 ½ years, and her data consist of taped talk, participant observation, mobile calls, SMS and hand written letters. Keywords: research ethics, ethnography, colonial research, Western research ethics, ontology, epistemology Emotional Implications in Informant-Researcher Relations Trond Stalsberg Mydland, Agder [email protected] University College/ Agder Research Foundation, Qualitative approaches, such as participant observational fieldworks and interviews, are widespread used as methods in anthropological and sociological studies. Methodological theories and techniques, as well as formal ethical reviews, are important when researchers design and do suchlike approaches. But doing fieldwork and doing interviews always more or less involve implications that can be difficult to foresee in advance of face-to-face meeting involving informants and researchers. Several aspects evolve throughout these social meetings. Some of these are conscious, others unconscious, some are made conscious, others are not. Emotional implications can be such aspects, which can be difficult to predict prior to the actual meeting of informant and researcher. Likewise, how to establish meaning in emotional expressions challenges researchers’ professional and personal skills and experiences. Such challenges occur both in the definite meeting of informant and researcher, and afterwards. It can also be demanding for the researcher to get admission to informants’ experiences of emotions in the definite situation. To assure that informants are prevented from unpleasantness even more difficult. Meaning is negotiative, and emotional meaning can be challenging to determine. In this paper I will discuss approaches to two different empirical fields. First I will consider challenges in participant observation as method for exploring communication in social relations with informants who have a mental retardation and do not have a verbal language, or other fully developed sign language. Second I will examine challenges in qualitative interviews with persons who have sentences on rape and sexual assaults. In both these fields I have experienced that emotional implications can be challenging to handle, ethically, personally and professionally. Keywords: qualitative approaches, methodology, emotional implications, communication The Regulation of Research Ethics – Some Challenges of Social Diversity Carole Truman, Department of Applied Social Science, Lancaster University, Lancaster, UK, [email protected] One of the greatest challenges to social researchers is presented by the ethical regulation of social research. The role of research ethics committees has expanded across the UK and North America and the process of ethical review has become institutionalised by a variety of statutory and institutional 167 bodies. Ethics committees have gained a powerful role as gatekeepers within the research process. However, it is often the case that the re-constitution of ethical guidelines are underpinned by a range of measures which protect institutional interests, without necessarily providing an effective means to address the moral obligations and responsibilities of researchers in relation to the production of social research. This paper provides a discussion of research ethics from the standpoint of research participants who may come from diverse or marginalised social groups. This perspective provides a useful dimension to current debate. The paper discusses ethical issues which may emerge from the perspective of participants during the actual research as it is carried out. Individual and collective perspectives on ethics within social research highlights incongruities between the formal ethical regulation of research, and the experiences of research participants in relation to the research process. Keywords: Research Ethics; Participatory Research; Research Governance; Social Diversity; Marginalised Groups Ethics and Power in Community-Campus Partnerships for Research Susan Boser, Indiana University of Pennsylvania, USA, [email protected] The past twenty years has seen a strong emergence of social research approaches that seek to reduce the distance between the researcher and the researched. Grounded within the critical theory paradigm, these approaches bring particular attention to relations of power, often with an explicit agenda for altering power imbalance. Emerging in various forms such as action research, community-based research or participatory evaluation, such participatory approaches are marked by engaging local stakeholders directly into the research process itself. Typically these approaches include the researched in defining the questions, in data collection and analysis, and in interpreting and taking action based on the research findings. In removing the distanced objectivist stance, such community - researcher partnerships seek to share power in knowledge generation and, potentially, in decision-making based on that knowledge. Thus, this movement toward participatory research approaches brings new sets of social relations for research and, as such, presents a new set of ethical challenges. The current framework for understanding the ethical issues involved in research is predicated on post-positivist epistemological assumptions of a distanced objectivist research stance, and thus is ill-suited for examining the ethics of participatory research approaches. This paper shall address this gap, presenting a framework for considering the ethical questions involved in research partnerships between the community and university-based researchers. I will begin by offering an operational definition of ‘participatory research’, articulating the most common elements and describing some of the typical social relations that are often involved. Then, working from a critical theory perspective, I will consider some ways in which power imbalance might manifest within a participatory research project and outline potential ethical implications. Finally, I will suggest questions that participants in participatory research projects might use to explore such issues, highlighting some potential methodological implications. Examples from a four-year action research partnership between a Cornell University doctoral student and a community of local government officials, human service providers, and persons who used human services will illustrate key points. Keywords: research ethics; action research; participatory action research; participatory evaluation; social relations and research 168 SESSION: STUDYING RESEARCH SENSITIVE TOPICS IN QUALITATIVE Organizor: Hennie R. Boeije Contact: Department of Methodology and Statistics, Faculty of Social Sciences, Utrecht University, PO Box 80140, 3508 TC Utrecht, Netherlands, [email protected] Keywords: sensitive topics, qualitative research, intrusive threat, ethics Abstract Despite the sensitive nature of several social science topics such as, bereavement, illness and divorce, researchers hope and expect research participants to answer their questions and to elaborate in depth on them. Although most current academic works pay attention to ethical issues in general, it is unclear what researchers do with these guidelines and considerations and whether they are aware of the methodological issues concerned when they study sensitive topics. Lee (1993) described a sensitive topic as one that potentially involves a level of threat or risk to those studied and distinguished one type of research that poses an intrusive threat as it focuses on areas of personal experience that are emotionally charged. It is assumed that sensitivity potentially affects almost every stage of the research process from the formulation of a research problem, through the design and implementation of a study to the dissemination and application of the findings. In this session speakers are invited to think about and share their experiences with doing research into sensitive topics as they were demarcated above, especially with concern to the methodological and technical problems that were raised and how they dealt with them. Minerva’s Web: The Internet and Methodologies of Sensitive Subjects Indhu Rajagopal, Division of Social Science, York University, 4700 Keele St., Toronto, Canada, M3J 1P3, [email protected] Seeking new knowledge requires the use of all information and communication technologies both soft and hard. Despite the global expansion of the Internet, using the Web as a research tool faces many barriers and facilitators. Partly it is due to the newness of the medium, and partly due to the subject to be researched. On both accounts we have indeed made some progress by designing surveys, collecting narratives, exploring the Web, conducting focus groups using conferencing software, as well as selecting the subject matter uniquely suitable for Web research. In this paper we will explore whether there are subjects e.g. pornography, Internet piracy, etc., privileged to be researched through the Web, and whether there are kinds of information from the respondents – qualitative or quantitative - that are advantaged through the Web. Indeed, the limitations of the Web in scholarly research will be an integral part of this analysis. Keywords: Internet, research methodologies, sensitive subjects, technology, facilitator, research, knowledge Positioning Myself as a Researcher in a Politically Sensitive and Volatile Situation Chaya Possick, Dept. of Social Work, College of Judea and Samaria, Ariel, Israel, Ha'alon 37, Ginot Shomron, 44853, Israel, Telefax: 962-9-792-9885, [email protected] For the past six years I have conducted qualitative research projects revolving around the experiences of West Bank settlers. The political situation in Israel in general and, the West Bank in particular, have undergone rapid and dramatic changes during this period. These changes have included: PalestinianIsraeli negotiations about sovereignty, army redeployments, terrorism, retaliation, and the dividing fence. All of these developments are highly controversial both within Israeli society and in the 169 international community. They engender intense emotions and tendencies towards argumentation and defensiveness. As a qualitative researcher, my aim is to explore the inner thoughts and feelings of the settlers. In addition to the barriers to open and genuine expression on the part of the respondents regarding highly controversial and threatening political situations, there are numerous researcher factors that must be considered. Over-identification on the one hand, and the projection of outsiderness on the other, by the qualitative researcher, affect the research design, its execution, and the representation of the findings. This presentation will focus on the concepts of the middle voice, multiple positionings, and emotionalism, and their application in politically sensitive and volatile research situations. The application of these concepts will be traced throughout three phases of the research process—1) topic selection and formulation of research questions, 2) data collection and analysis, 3) writing up and publication of the research report. Keywords: qualitative research, sensitive topics, West Bank settlers Studying Decision-Making Issues at Home: The Nature of the Relation between Researcher and Participants María Elena Ramos Tovar, Universidad Autónoma de Nuevo León, Graduate School of Social Work, [email protected] or [email protected] Studying outcomes of decision-making issues at home has been traditionally analyzed using standardized questionnaires. However, such an approach is limited in analyzing the ways men and women persuade, negotiate, or confront in everyday circumstances. The dominant understanding of women’s power in Latin American is that women reign at home. Discursively, I accepted that answer but was unsure of up to what point it reflected reality. Several studies, conducted mainly in the U.S., concluded that women take responsibilities for children and household tasks and that men take care of “hard” chores, such as fixing things around the house, or helping their children with their homework. But does doing such things and deciding about such tasks give women the control and the moral authority at home? What do they think about doing and deciding such issues? At this point, it was clear to me that questionnaires would not be sufficient to answer those questions. Using a qualitative design allowed me to undertake an in-depth description of how power processes between men and women take place at home. A qualitative technique, such as the in-depth interview, allowed me to capture the different ways women and men negotiate their lives at home. Many times people are not conscious of their interactions and decision-making strategies. A qualitative design serves to describe complexities and processes that more quantitative techniques cannot capture. One difficult aspect of contacting people was explaining the objectives of the study. First of all, I had to make clear that I was not a psychologist who was looking to analyze any wrong doing in families’ interactions. To be under scrutiny would be threatening for respondents. But if this study was not to help solve their problems, what was it for? I did not say that the study was about power but about familial relationships. I assumed that the concept of power could be problematic since it implied relationships of authority and that people might feel it inappropriate to talk about authority within the family. Another difficulty was the role imposed by the participants. Some of them saw me as a familial counselor. Some thought of me as a mothering wife apprentice. Others as a confidant. Those roles as an expert and as apprentice were valuable ones. I learned from both of them, but definitely the one as a counselor was the most difficult one. Keywords: decision-making, contrasting qualitative-quantitative responses, interaction researcherparticipant 170 Some Functions of Probing Tactics in Interviews on Friendship Gerben Moerman, Department of Social Research Methodology, Vrije Universiteit Amsterdam, De Boelelaan 1081c, 1081 HV Amsterdam, The Netherlands, +31-20-4446776, [email protected] In qualitative interviews on sensitive topics, interviewers expect and hope that respondents elaborate in depth on their questions. Not many respondents however, will promptly elaborate on sensitive questions. So interviewers need probing while interviewing hesitant respondents. Unfortunately not much is known about the effects of probing tactics in these situations. Interviewing respondents on sensitive topics requires a certain amount of respondent’s trust in the interviewer. For example, if the respondent believes that the interviewer will misuse the information, the respondent will be reluctant to disclose sensitive information. Fontana and Frey take this requirement somewhat further by stating that “gaining trust is essential to an interviewer’s success” (Fontana & Frey, 1998: 59). Gaining trust is often viewed as part of a more general empathic attitude that the interviewer should demonstrate during the interview. A probing tactic in which the interviewer uses a repertoire of complying probes appears to support this empathic attitude. Consequently, other probing tactics with challenging or accommodating probes are seldom promoted in qualitative interviewing. Hence, there is still a lack of knowledge about the effects of different probing tactics in qualitative interviews. This paper focuses on the question whether respondents in interviews on sensitive topics tend to be more disclosing if a compliant probing tactic is used instead of another. In order to answer this question, I explore and compare how three probing tactics - compliant, accommodating and challenging - function in interviews on friendship. For this experiment nine interviewers were trained to conduct the interviews using solely one of the three probing tactics. In total 36 respondents were interviewed on the number of friends and the meaning of friendship. One surprising result among the results presented in this paper is that a compliant probing tactic does not necessarily leads to more trust then challenging probing tactic. Processes in Research into Sensitive Topics Pat Cox, Department of Social Work, Faculty of Health, University of Central Lancashire, Preston, PR1 2HE, United Kingdom, [email protected] Article 12 of the UN Convention on the Rights of the Child and Article 1.2 of the European Convention on the Exercise of Children’s Rights incorporate understandings that children and young people are subjects who can think, conceptualise, form views on past experiences and contribute to decision-making about their futures. However, the author’s experience is that such understandings are not consistently reflected by those who commission research involving children and young people, nor by all providers of health and social care services who may be approached to facilitate access to them. In this paper the author examines some of the methodological and ethical issues arising from endeavouring to undertake qualitative research with children, young people and adults. The research was undertaken in order to assess the need for a proposed community-based support service for children and young people who are survivors of sexual abuse, and for their safe carers. The subject of child sexual abuse fits Lee’s (1993) definition of a sensitive topic as potentially involving a level of threat to research subjects and addressing an area of experience that for survivors, safe carers, their supporters and for those who work with them is emotionally charged. The author describes and analyses the processes in the research project outlined above, focussing on how various people involved in the research, whether as research subjects, researcher, commissioners of research or gatekeepers, responded to the sensitive nature of its topic. She posits that both differing agendas (personal and professional) and levels of comfort with the topic influenced responses, and impacted on the outcome in ways which could not have been anticipated entirely, but which with hindsight are obvious. She suggests possible ways forward, both for researchers and for those who commission research. Keywords: sensitive topics; processes; methodological issue 171 SESSION: RECENT DEVELOPMENTS IN ETHNOMETHODOLOGICAL AND CONVERSATION ANALYTIC RESEARCH Organizor: Paul ten Have Contact: University of Amsterdam, [email protected] Keywords: ethnomethodology, conversation analysis, interaction studies, recordings-as-data, practical applications, workplace studies Abstract Since their emergence in the 1960s, ethnomethodology (EM) and conversation analysis (CA) have been relatively ‘stable’ in terms of their general methodological orientation. Their focus is on the detailed procedural study of the production of social order, often (and in the case of CA always) based on audio or video recordings of particular situated events. While these traditions have emerged within sociology, the approach and methodology are now used in a range of disciplines, including (interactional) linguistics, anthropology, communication studies & (discursive) psychology. Within the original overall framework, there have been, and still are, some interesting developments which deserve a focussed methodological discussion, including (but not limited to): the use of video- and computer-based technologies, combining recordings and ethnographic fieldwork, tentative quantifications of particular aspects. Of particular interest are issues concerning the practical ‘application’ of EM and CA studies in a number of fields such as: interaction of/with people with various kinds of communicative impairments secondary language teaching work organization in technologically complex settings human-computer interaction, computer-supported collaborative work & systems design. Part I Technomethodology Andy Crabtree, University of Nottingham, Nottingham, UK, [email protected] Harold Garfinkel has had occasion to note that “ethnomethodology is applied ethnomethodology” and that in its application it has ambitions to become a “hybrid discipline”, merging with different professions to address their occupational troubles. This paper discusses the emergence of one such hybrid discipline – technomethodology - in Information Technology (IT) research. Ethnomethodology has been involved in IT research for twenty years, where the approach initially helped the designers of IT systems address the troubles brought about by the mismatch of computer applications with the real world, real time circumstances of their use. In this respect, ethnomethodology’s original role in IT research was a critical one, drawing attention to the failings of psychological models on which design was predicated, and the need for design to be responsive to the social circumstances of work within which IT systems are embedded in their use. This early involvement resulted in the ‘turn to the social’ in IT research, where the challenge became one of understanding the collaborative character of work. This, in turn, raised the challenge for ethnomethodology to move from design critique to design practice. The move from design critique to design practice is a methodological move that requires ethnomethodology to become an active participant in the “invention of the future”. In the first instance, this resulted in the incorporation of “workplace studies” in traditional product development processes. This is a problematic relationship, however, as it casts ethnomethodology in a service provider role that inhibits the development of a foundational hybrid relationship. The situation is compounded by fact that the vast majority of ethnomethodological studies of work are not carried out 172 in product development settings but in universities and commercial laboratories, where emphasis is placed on innovation rather than on the construction of products. There is a need, then, to reconfigure the relationship to allow the emergence of a socio-technical methodology – technomethodology where technology becomes a vehicle for social research, the results of which in turn propel innovation and the invention of the future. Drawing on interdisciplinary work involved in the development of an internationally acclaimed “mixed reality game”, this paper articulates what technomethodology might be about, what it might look like, and what it might consist of concretely as a hybrid research methodology for IT research. Keywords: Ethnomethodology, IT Research, Hybrid Discipline, Technomethodology Exploring the Possibilities for the Use of Ethnographic, Contextual Information in Ethnomethodological/ Conversation Analytic Studies. Bregje C. de Kok, University of Edinburgh, Edinburgh, UK, [email protected] In this paper I will discuss whether, and in what way, information about the external context can be usefully drawn upon in studies based on the analytic traditions of conversation analysis and ethnomethodology. With ‘external context’ I mean the context external to the conversations on which the analysis is focused, like the wider cultural context, or biographical characteristics of research participants. I will address this question in two ways. First, I will discuss interview data in which Malawians talk about their own and others’ infertility, and explore how additional, ethnographic, contextual information may aid understanding of these data. The data analysis will be guided by an analytic interest in the interactional production and maintenance of ‘common sense’ and social constructions, rather than a ‘pure’ conversation analytic aim of the sequential organization of turns. Second, I will problematize arguments put forward by conversation analysts for discarding information obtained from outside the recorded conversations. Thereby I will discuss how these arguments point at differences and tensions between current conversation analytic work and its ethnomethodological foundations. Normally, conversation analytic studies refrain from using ethnographic data about the ‘wider’ context, because of adherence to the criterion of procedural consequentiality (Schegloff, 1992): ethnographic information is superfluous as the features of the external context which are relevant for the interaction will be ‘talked into being’ only ‘there and then’, in the local sequences of talk, and can thus be discerned in conversation partners’ orientations as displayed in the details of their conduct (Boden & Zimmerman, 1991, Ten Have , 1990). However, the vast majority of conversation analytic studies has used data gathered in the analyst’s own society or cultural setting. Conversation analysts tacitly assume and take for granted the analyst’s cultural competence to recognize the ‘actions’ which interaction partners bring off in their talk (Cicourel, 1992; Firth, 1996; Lynch, 1985) and the procedural consequentiality of contextual features. This raises the question of how talk-in-interaction can be analysed when conversation partners and analysts do not come from the same cultural and/or linguistic community, and whether at least in such a case drawing upon information about the external wider context is a useful, and perhaps necessary tool in the analysis. In addition, conversation analysis’ reliance on recordings seems to be based on its specific interest in sequential analysis of the turn taking system, and on an empiricist, realist epistemological perspective. However, this analytic interest, nor the realist epistemology are characteristic for ethnomethodology as founded by Garfinkel (1967). In conclusion, I will argue that discarding information about the external context does not necessarily follow from the ethnomethodological spirit, and it becomes potentially problematic especially when dealing with data obtained in a cultural setting different from one’s own. Keywords: conversation analysis, ethnomethodology, cross-cultural analysis 173 Shifting the Focus: The Impact of Audiovisual Technology on Empirical Research Antonia L. Krummheuer, University of Bielefeld, Germany, [email protected] In the last two decades, more and more ethnomethodological and conversation-analytical studies used video-based technology to collect their data (e.g. workplace studies). One advantage of audiovisual recordings is the possibility to analyse not only verbal but also non-verbal communication. Furthermore, the recordings can be replayed arbitrarily and a detailed analysis of the ongoing and often unnoticed processes of interaction is feasible (e.g. Heath, 1997). Until now, just a few studies discuss the limitations of the use of video-based technology, e.g. that efforts and costs are high (e.g. Jordan/Henderson,1995; Ruhleder/Jordan, 1997). The crucial point is, that the camera is not an “objective” instrument. The existence of the camera in the field, its position as well as its point of view on the situation (e.g. perspective) influences the field and shapes the material in special ways and therefore establishes its own structures of meaning. For example, the position of the camera prearranges the field into sections where relevant action should take place. Furthermore, not only the camera is influencing the field also the field has an impact on the video-recording - and the material as well. For example, the position of the camera is limited by conditions on side and reactions on the camera admit interpretation of the field. Therefore this presentation shifts the focus from video-recorded interaction to the video-recording itself. It will focus on the mutual influence of video recordings and the field and its impact on the collected material. This will be demonstrated with video recorded data of a situation in which a camera is usually absent. The data is derived from an audiovisual research of computer mediated communication between human actors and an interactive and personified virtual agent (software program) on a trade fair. To highlight special aspects, this data will be compared to television interviews. Keywords: video- and computer-based technologies; ethnomethodology; conversation analysis; human-computer interaction Order in Disorder: Rule-Governed Actions as Sociological Phenomena Alain Bovet1 and Philippe Sormani2 1 University of Fribourg, Switzerland, [email protected] 2 University of Manchester, UK, [email protected] Norms, values and rules are commonly considered in sociology as given and explanatory of social action. However, the plausible explanation of an action does not favour the detailed description of its course. In effect, the sociological emphasis on norms, values and rules has as an important consequence to focus upon the conditions of social action rather than its actual accomplishment. Therefore, the purpose of our paper is to examine two video recordings of how social actions that can be explained as rule-governed are actually accomplished: in one case, participants in a television debate struggle to take a turn at talk; in the second case, drivers at Mexico City crossroads struggle to manoeuvre their cars into the traffic. As problematic cases, these two instances will allow us to examine how participants produce, recognize and relate their actions to presumed common rules. Situations of apparent disorder provide a case in point to show how rules are used by participants as resources to coordinate their respective actions and re-establish a local order. Turn-taking seems to play a prominent part in that coordination and this re-establishment. Hence we will focus on how participants conjointly establish where and when turn transition may occur. Taking up Wittgenstein’s emphasis on the internal relation between rule and (rule-governed) action, and Garfinkel’s insistence on their reflexive constitution, the paper respecifies some of the conceptual and methodological problems involved in the analysis of rule-governed actions, including turn-taking, as ever investigable sociological phenomena. In particular, the notion of ‘transition relevance place’ will be investigated as the result of participants’ situated enquiries. As methodical achievements, these enquiries are amenable to ethnomethodological investigations. Keywords: Rule-governed action, traffic, interaction, turn-taking, ethnomethodology 174 Part II Tracking Psychotherapeutic Agendas with Conversation Analysis Charles Antaki1, Ivan van Leudar2 and Rebecca Barnes3 1 Loughborough University, UK, [email protected] 2 Manchester University, UK, [email protected] 3 Loughborough University, UK, [email protected] A recent development in Conversation Analysis is its application to the business of psychotherapeutic interactions. In this paper we show how CA can illuminate how therapist and client achieve certain therapeutic goals through the design of their talk, but we also address the extent to which CA can comment on the desirability or advisability of these goals (and the ways they are achieved). At some point these issues are properly the province of therapists. The tipping point, however, at which CA ceases to be useful and illuminating, is a matter of debate. We address the issue with an analysis of a central feature of therapeutic interaction, the therapists' offer of interpretations of the clients' reported experiences. We show how tensions can arise from the therapists' use of formulations, idiomatic language and yes-no question formats, and how clients can offer 'resistance' by dispreferred responses. Whether these discoveries reveal mis-steps in the therapy process, or whether they are subservient to more overarching therapeutic agendas not visible to the analyst, are matters we shall offer for debate. Keywords: (applied) conversation analysis; therapeutic interaction; therapists’ interpretations The Production of a live TV-interview through Mediated Interaction Mathias Broth, Stockholms universitet, Sweden, [email protected] This paper presents some results from an ongoing project on television production as an interactional process. The data used are video-recordings of screens and personnel in the control room and of the people in the studio during four live productions of the monthly French interview show called "Rideau Rouge". The interaction on which the paper focusses primarily is taking place between some key members of the production team: the director and the script posted before the numerous screens and the switch-board in the control room, and the operators moving around the participants to the studio interaction in the studio. Since the director and the script on one hand, and the operators on the other, work in different locations, communication between these parties has to take place using technological devices. Whereas the director and the script can talk to the operators through a microphone, the operators can only show the director and the script their respective images, not talk back to them. In addition, every time that the director puts a certain camera on-air, a tally light is activated both on top of it and in its viewfinder, which effectively communicates the on-air/off-air distinction to the camera's operator. How is it that members to the production team can understand correctly what their colleagues are doing through the technological configuration at hand, and collaboratively accomplish the wellordered product that the audience can view on their television sets? The answers to this broad question begin to show that the mediated and asymmetrical character of the interaction between the director and the operators is clearly consequential for the way they constitute recognizable action. Analysis starts off with the establishment of timing as a crucial resource for the production of intersubjectivity in this context. Sequencing and constitution of actions via the technological interface is identified as centrally a matter of producing actions in strict adjacency, lest they would not be recognized as belonging to a same sequence of actions at all. The paper then moves on to describe a basic action sequence in live television production, the "proposal - accept" sequence, and what constitutes its respective action types. This constantly repeated sequence generates the sequence of shots that the audience sees in the final programme. The paper's last analytic section concerns the projectability of the actions in the basic "proposal - accept" sequence. The possibility for members of the team to make successful hypotheses regarding the production of one of these action types by reference to other aspects of the interactional 175 context is demonstrated, but a case of conflicting projectabilities - generating an "error" in the production - is also analyzed. The results yielded by analysing recordings of naturally occurring interaction in this setting, and by focussing on how members to the team actually treat each other's actions in real time, underscores the relevance of CA methodology for studying the hitherto unexplored work of collaboratively producing live television. Keywords: conversation analysis; sequential analysis; television production; mediated communication, intersubjectivity; collaborative production Talk about Talk: The Methodological Status of Teacher Interviews on Classroom Interaction Tom Koole, Utrecht University, The Netherlands, [email protected] In this paper I will discuss one aspect of the methodology of an interdisciplinary project in the Netherlands that focuses on interaction in multilingual and multicultural classes. Since one of the project’s interests is in the relation between the teacher’s awareness of linguistic or ethnic differences, and his interactional behaviour in the class, we videotaped and transcribed maths lessons in first grades of secondary education (students aged 12-13) and in addition interviewed teachers and students. One of the methods used to analyse the interactions in the math lessons was Conversation Analysis (cf. Koole 2003). In this paper I want to focus on the relation between the interactions in the classroom and the teacher’s comments on these interactions in the interviews. Part of the interviews have a ‘stimulated recall’ character, which means that the teacher was shown parts of the lesson just recorded and invited to comment on them. Others parts deal with the more general views the teacher holds on education, multicultural classes, and specific students. I will present analyses both of fragments of classroom interaction, and of interviews in a search for teachers’ categorizations of students and their actions. I will show, in line with Sacks (1992), that categorizations are interactional activities and have a situated character. It is therefore methodically hazardous to treat the teacher’s categories used in the interview as identical with categories used in any specific interactional event in the classroom. On the other hand there is evidence from research of intercultural communication that categorizations (notably of co-participants) can have a less contingent, relatively stable status in interactional behaviour. In conclusion I will make the argument that the teacher’s interview statements show us the range of categories the teacher treats as relevant for categorizing students and their behaviour, but cannot be used as evidence of relevant teacher categorizations made in any specific event. What’s in a Name?: Revealing Social Identity Erica Lea Zimmerman, University of Hawai'i at Manoa, [email protected] Frequently, a practice of researchers of second language acquisition studies is to presuppose the identity of the participants as “language learners” in an interaction. These studies often assume before examining the data that the identities of “non-native speaker” or “second language (L2) learner” are the relevant identities for the interaction. Using a Conversation Analytic approach for examining identity in spoken discourse as proposed by Sacks (1992) and Antaki and Widdicombe (1998), this study will examine how identity is made relevant in naturally occurring speech. The methodological point of this study is that the analyst should abstain from presupposing the identities of the participants before examining the data. This means that the analyst must see how the participants themselves orient to and make relevant identity within the interaction. The data for this presentation is part of a larger study examining construction of identity by five multilingual speakers of Japanese, who are native speakers of Korean. Focusing on data from one of the participants from this larger study, this 176 presentation will examine a series of interactions by the same individual, focusing on repair that occurs once the participant states his name in telephone and face-to-face conversations. This repair occurred when the participant made phone calls or inquires at places such as the local police station and the post office. While the focus is on how identity is manifested in talk, the mechanism for alluding to the identity of the participant is the use of self-initiated, self-repair or otherinitiated self-repair (Schegloff, Jefferson, and Sacks, 1977). This study finds that the identities made relevant by the participants in the discourse are “Korean” and “otherness” meaning not Japanese. In order to provide a better understanding of issues of identity construction in talk in general, this study will help to open the door of international communication on a level that has not been previously examined. Keywords: Conversation analysis, Identity, Second language learning, SESSION: RESEARCH METHODOLOGY AND THE CHALLENGES OF SOCIAL DIVERSITY Organizer(s): Carole Truman1 and Donna Mertens2 Contact: 1Lancaster University, UK, [email protected] 2 Gallaudet University, USA Keywords: Transformative Research; Emancipatory Research; participatory research; Feminist Research; Marginalised Groups Abstract Contemporary society is characterised by social diversity. This presents particular methodological issues for social researchers who may find themselves working with diverse groups how may experience various forms of marginalisation. This session invites papers from practioners of social research who have addressed or confronted social diversity in their research practice. For example, how do methodological concerns relate to the way that research 'problems' are framed? How may research relate to those groups affected by research and meaningfully involve them in shaping the research process? How can researchers take steps to avoid designing a study that denies treatment to specific groups who then are "left behind". What methods have been used to collect accurate data which are culturally appropriate for diverse groups of people. The session organisers would be keen to receive contributions which have pursued the notion of mixed methods as a methodological territory to explore issues of doing research w/n culturally diverse groups. The Hermit’s Knowledge Isacco Turina, University of Padua, Italy, [email protected] In what follows we will try to develop a cognition-based reflexivity, hoping to reach the generalization and objectivity often denied to emotion-based reflexivity: we suggest that, dealing with its objects, reflexive sociology should move from the question “What do they do?” to “What do they know?”. Better, it should inquire the connection between them. These proposals have been applied in a field research on Italian Catholic hermits, people living a monastic life in solitude, out of traditional monastic communities. 35 in-depth interviews have been conducted with them. Reflexivity helped us to explain why these people living out of society, unanimously affirmed to be much closer to other human beings than one can be remaining in society. We think that, to understand such a paradox, an epistemological shift is needed, consisting in perceiving that hermits’ everyday life, which for a traditional descriptive sociology is the ends of knowledge, is to them on the contrary the 177 means of it. The sociologist, used to consider the act of learning as a work performed through methods and tools, should become aware that hermit knows with his whole life: that’s like saying that a blind person sees the world not just by his eyes, but by his whole body. In modern world, we can recognize two distinct sorts of knowledge: science, in which what people know and what they do have no links; and truth, where their knowledge obliges them to adhere to it by actions and behaviour. The former provides objective freedom but only a partial, relative and changing knowledge: for example, persons must behave and judge differently as they play different social roles. The latter imposes on its followers to give up freedom to organize their whole life around it, but gives them in exchange a holistic, stable and absolute understanding of world, history and life: militant members of religious and political groups are good examples. What we suggest is that science is not primarily a matter of technology, but a way of life that requires an a priori refusal of matching knowledge and action: so conceived of, even in modern democracies science could be much less spread than we usually think, challenged as it is by many kinds of ethical ideologies and religions requiring a commitment of belief and action. Methodologically, what we aim to show is the importance of paying attention to a reflexive dimension which we name cognitive distance, i.e. the difference between two possible meanings of the phrase “hermit’s knowledge”: the knowledge that the hermit has about the world, and on the other side, the knowledge that the social scientist has about the hermit. According to us, social sciences should not forget that: 1) Their objects are also, as the sociologist himself, knowing subjects. 2) Science is not the only way of acquiring knowledge about reality. Cognitive comparison between the scientist and his objects should prove relevant for that scientific understanding of them which is the very scope of research. Keywords: Hermit’s Knowledge, Reflexivity, Truth Science Philosophic Sagacity: Another Way of Doing Research Bagele Chilisa, Department of Educational Foundations, University of Botswana, P/Bag 0022, Gaborone, Botswana, [email protected] Countries that suffered from imperialism and colonisation critique research processes and procedures on the basis that they are informed and rely on literature and a world view that continues to perceive the coloniser as the knower and the colonised as ignorant. It is for this reason that some researchers believe the process of framing research questions, gathering data, analysis and interpretation may if not carefully examined exclude the colonised worldviews and ways of knowing and thus fail to communicate their experiences. The article explores processes through which African indigenous peoples can have greater autonomy in research and construct their own social realities in ways not so mediated by a Western knowledge systems that are framed solely within the boundaries of Western history, culture and world views. The article explores in particular philosophic sagacity or the use of sages in research. Philosophic sagacity is an analysis of accepted practices, believes and worldviews in African communities. Through Sagacity, the wisdom and beliefs of individuals who have not been schooled in the formal education system is exposed. A Sage is well versed in the wisdoms of his/her people and has a reputation for the knowledge; a thinker who is rationally critical and capable of conceiving excellent options and recommending ideas that offer alternatives to the commonly accepted opinions and practice. African communities with particular reference to Botswana have had a long history of diverse ways of processing and producing knowledge in centres such as the Kgotla (chief’s palace), shrines and religious centres. The production of knowledge was facilitated by indigenous researchers/intellectuals (sages) that included chiefs, poets, social critics, herbalists, visual artists, musicians, diviners and storytellers guided by the communities’ values and ways of perceiving reality. The article looks at the story of HIV/AIDS as revealed by a young artist through his art work. The article shows how the sage goes beyond explaining people’s behaviour to revealing and explaining to the people the socially organised powers in which his life and the life of the people in his community are embedded. The young sage creates the story of HIV/AIDS using materials and images that reflect realities and experiences as lived by his people. The different sites, and different 178 experiences portrayed through his art provide multiple perspectives and possibilities of diverse strategies towards addressing the Problem of HIV/AIDS in the country. Irrationality and Violence in the Postmodernity. Proposal of a Model of Analysis of the Social Reality under the Perspective of a Transdisciplinary Method Augusto Renato Pérez Mayo and Maricela Guzmán Cáceres, Universidad del Valle de México Campus Villahermosa, Mexico, [email protected], [email protected] The postmodernism has created a monstruos term of the reason, and later on it has been easy to criticize it, proposing other ways of thinking, another historical figures of intelligence. But it is not an aberration in proclaiming the preeminence of the reason. To our manner of understand the ideas of Popper when affirming extended that the election of the rationality is not based on its capacity to found a corroborated affluent knowledge, but in which it is the use of the intelligence that better can save to us of the horror. The irrationality leads after or before to violence, which prevails in the updated world like an answer to the crisis of the economic systems, moral and political everywhere in the world. Popper on the matter arguments “I believe that the critical form or the reasoning and, mainly, the belief in the authority of the objective truth are indispensable for a free society based on mutual respect. And this one is the reason for which it has such importance of not allowing on our thought the serious influence of intellectual misunderstandings like irrationalism, comprehensible results of the dogmatism and authoritarism.” It is not about to affirm the primate of the reason, but also to evaluate the most suitable model of intelligence for the aims of the human being. The critical reason is the only alternative to the violence that has been discovered so far. Every epoque has elected an intelligence model from which it considered its greatest creation. The modernity chose like ideal the reason and science, the posmodernity has welcomed an aesthetic paradigm. Now we know alredy the power and the weakness of the rationalism, and at the same time the force and the weakness of a weak thought. The question to answer is How to construct a new model? Is it time to do it? Is it possible to analyze the violence to the light of a transdisciplinary method and the lower the scientific categories of sciences more developed to apply them to the anaylisis of a social phenomenon of such relevance?. According to Luhmann, Maturana; Varela, Pask, Brown, Castell, Cornella, Vilar, Von Foerster, Gunther, Morin and others, our effort of abstraction seeks to unfold the new equations, new logics that would give to born a new model of interpretation of the reality of the being. We pretend to define a complex theoretical system that can give a treatment to the problems and emergent systems, understanding that the unit is possible to find it in the difference. Keywords: Emergent society, complex society, complexity, cognitive system, complex-cognitive system, trivial system, closed system, operationally closed system, selfpoietic system. Researching Civic Participation within a Diverse Context: Conceptual and Methodological Opportunities and Challenges Kien Shan Lee, Association for the Study and Development of Community, Gaithersburg, USA, [email protected] This paper is based on a study that examined the meaning and practice of civic participation among four immigrant communities in the Washington, DC metropolitan region: Chinese, Indian, Vietnamese, and Salvadoran. Topics explored included: definition of civic participation; individual and community capacities required for civic participation; and similarities and differences in social organization of each immigrant community and its effect on patterns of civic participation. The predominant approach was qualitative with some borrowed methods from quantitative research. The research questions were framed to deepen existing knowledge about civic participation in the United States. This knowledge base has been and continues to be informed primarily by the dominant 179 culture’s (i.e., White) values and norms regarding civic participation, resulting in incomplete portrayal of immigrants’ civic activities in the United States. The methodology was developed to include different types of leaders from the four immigrant communities using stakeholder interviews. This required the researcher to understand the concept of social organization in order to be able to ensure that a range of perspectives were captured. Issues of researcher and stakeholder credibility, which were unique given the ways in which credibility is defined in different cultures, had to be considered and addressed properly. The interview protocol had to be flexible enough to accommodate the diverse ways in which people from different cultures express themselves, while enabling the researcher to gather data in a uniform and rigorous manner. Stakeholder review of interview transcripts had to be built into the data verification process in order to ensure that their perspectives were transcribed accurately. Data coding had to be conducted by two persons to ensure reliability and precise interpretation of concepts and terms. The analysis included both an examination of content as well as quantification of responses in order to explore patterns and themes within and across communities. This study provided an opportunity to reflect on the special considerations required for conducting research in a diverse context, including the benefits, risks, and challenges. It also provided an opportunity to influence future research about civic participation that might be more appropriate in a diverse society. Keywords: stakeholder engagement, immigrant communities, qualitative research, civic participation Transformative Research and Dimensions of Diversity Donna M. Mertens, Gallaudet University, USA, [email protected] The history of social program research is firmly rooted in the theoretical and methodological traditions of post-positivism and social constructivism. Theoreticians and practitioners have come to a growing realization that the way in which research intersects with the social policy arena requires an expansion of thinking to accommodate the demands that result from activities associated with decision making about programs and policies (Greene, 2000). These accommodations address issues of moral behavior and ethics, social criticism, epistemology, and methodology. Researchers contribute to decisions about the need for and quality of educational and social programs and policies through a systematic investigation of such programs. They work in contested territory, laden with pluralistic values, and associated with real-life implications for resource allocation (Greene, 2000). Issues of power differences and vulnerability permeate the arena as well. These characteristics of the research landscape not only differentiate it from traditional social science research, but also logically connect it with the work of transformative scholars. Transformative scholars have as their central mission the accurate and credible representation of marginalized groups in and through the process of systematic inquiry towards the goal of bringing society to a point of greater equity and justice (Banks, 2000; Christians, 2000; Mertens, in press). Meaningful involvement of marginalized groups in research activities has been an on-going challenge for researchers (Chelimsky, 1998; Mertens, 1999; Weiss, 1998). Transformative/inclusive research emphasizes the deliberate inclusiveness of groups that have historically experienced oppression and discrimination on the basis of gender, culture, economic levels, ethnicities/races, and sexual orientation, and disabilities, and in a conscious effort to build a link between the results of the research and social action. A researcher's ability to be responsive to the challenges described above rests in part on their understanding of two complex concept: cultural competency and dimensions of diversity. The interpretation and operationalizing of these concepts is key to understanding the importance of and the mechanisms to address the deliberate inclusiveness of groups that have historically not been adequately, adequately, or appropriately involved in the research process. This issue is not only enduring, but it is a global issue. The purpose of this presentation is to bring an international perspective to improve understandings of these concepts for the theory and practice of research. Keywords: Transformative paradigm, cultural competency, disabilities, deafness, feminist 180 Culturally Engaged Research – A Specific Application of an Emerging Methodology Anthony J. Alberta1, Bonny Beach2 and John Whiteshirt2 1 Sonoran Research Group, Tucson, USA, [email protected] 2 American Indian Prevention Coalition, Phoenix, USA Members of ethnic, cultural and racial minorities are often reluctant to participate in research activities, especially those conducted by Anglo researchers or researchers employing experimental or quasi-experimental methods. Abuses of information gathered during research, interpretation of data without regard to how the cultural context affects the meaning of that data, disregard of the dignity of the people involved and the implementation of unethical research designs by researchers working in minority communities have contributed to a justified aversion to research activities. As the heirs to this scientific and ethical misconduct, current researchers from all backgrounds must work to not only gain the trust of members of minority communities, but to conduct the research activities in a manner that demonstrates respect for the people involved and their culture. During the past decade methodologies that involve participants, including participatory, collaborative and empowerment research call for including people in the design of research, the data collection process, and the interpretation of data collected. This level of inclusion transforms the research process, providing for real agency among “subject” groups and improving the accuracy and validity of findings by maintaining the cultural contexts in which they exist. Although these research methods significantly reduce the potential for abuse of research participants, they do not provide researchers, regardless of their own racial/ethnic/cultural backgrounds, with any guidance regarding how they might best interact with members of minority communities during the course of conducting research. In short, the successful application of these inclusive methods requires researchers to overcome barriers that include cultural conceptions of polite behaviors, the definition and locus of the phenomena under study, the definition of knowledge, and the role of communities in the research process. Culturally Engaged Research emerged as a distinct method during application of the Practical Skills Model for Multi-Cultural Engagement - a set of skills developed to assist behavioral health professionals to work in culturally diverse settings. Culturally Engaged Research combines the five skills from the Practical Skills Model and three distinct research practices into a unified approach to conducting research in a manner that emphasizes respect for the community participating in the research activities. This paper will discuss the application of the Culturally Engaged Research methodology in a project conducted in collaboration with the American Indian Prevention Coalition in Phoenix, Arizona, including both qualitative and quantitative findings. Keywords: Transformative Research, Emancipatory Research, participatory research, Feminist Research, Marginalised Groups SESSION: ACTION RESEARCH Organizer: Nancy Andes Contact: Department of Sociology, University of Alaska Anchorage 3211 Providence Drive, Anchorage, AK 99508, Tel: +1-907-786-4063, [email protected] Keywords: Action research; Participatory methodology; Collaborative research Abstract The last decade has seen an expansion of action research methodologies under the labels participatory action research, collaborative research, among others. Action research is fundamentally a democratic practice, acknowledging people’s capacity to create knowledge about and solutions to their own experiences. As a methodology, action research develops problem-solving, solution-oriented investigations and fosters improvement and self-determination. Knowledge generation is co-supported by members of higher education, organizations and agencies, governments, nonprofit organizations, and community residents who design and implement research. 181 The Catalyse Methodology and the Restoration of the Inquiry Results in the Action Research. Several Recent Experiments in the Rural Romanian Areas Mihai Pascaru, University 1 Decembrie 1918 Alba Iulia, Romania The Catalyse methodology is a creation of the MTI@SHS ....with the University in Besancon-FranceComte France. It develops methodological and instrumental solutions, which showing to advantage the principles of territorial intelligence defined by: the partnership of the operators, the global approach, the quality of and the access to information. It is a methodology for the action research. The essence of the Catalyse Methodology is represented by the restoration of results as method of shaping and putting into hierarchy the community problems in the local and regional areas. The sociological inquiries are just emerging stages for restoration, they enabling the constitution of an information set, which will raise and trigger off debates within the community with the end of identifying and putting into hierarchy the real problems of community and with the end of community development as well. In some recent experiments in the rural Romanian village (district Alba, Romania), the author of the present proposed paper used the restoration of the results of some sociological inquiries in two ways: the individual restoration (based on a semi-structured interview) and the collective restoration (by organizing restoration seminars). The paper intends to present some results obtained in the experiments in Romania, some limits discovered on this occasion and some deontological aspect noticed along the action research, by using the Catalyse Methodology, and restoring the results especially. Keywords: Catalyse Methodology, restoration of the results, community development Russian Drug-Addiction – NGO – Action-Research: Meeting on Crossroad Galina Saganenko, Sociological Institute of RAS, Russia, [email protected] Drug-addiction is a complicate and rather new social phenomenon in Russia, rapidly marshalling its destroying forces. Its parameters continuously have been changing, majority of notions about it are erroneous or intentionally misrepresented. Narcomania is what suits the Russian powers/authorities. The families with this problem are left without any support by the state. NGOs can give real help in crucial situation. As usual most of volunteers working in NGOs on a certain problem (as well on drug-addiction issues) are such people which have been touched by the problem, more or less dramatically. Activity in NGOs permits to a researcher to joint together some different statuses and to reach results of different types. I started to participate in activity of the Regional Charitable NGO “AZARIA” (“Mothers against drugs”, Saint-Petersburg) 6 years ago in order to get self-determination and to help in finding some solutions for the decision-makings for the NGO. The results of my participation in NGO activity are combined. What I have managed to achieve: (1) To carry out a system of investigations, which helped to find the potential in new kinds of social information and the manners of articulation of research and social goals and tasks in drug-addiction area. (2) To be in success in comprehension of social studies methodology, to articulate humanistic aspects of sociology, to substantiate new research types – research of reflexive and modules character. (3) To reveal significant facts in existence of the drug-addiction problem in Russian society on public and private levels. In particular, as for the macro-level there have been revealed and articulated the following: existence of inadequate Russian legislation related to drug issues, existence of 6-7 types of the modern Russian narco-business, absence of the state treatment infrastructure in modern Russia, absence of social work in this field, putting by the state the moral and financial burden only on family. (4) To get understanding in some principal components of drug-addiction field in Saint-Petersburg and Russia, in particular, to reveal uninterestedness of authorities in combating the use drugs: drug- 182 addiction gives them widening pseudo-activity for obtaining resources, rising one’s political and social capital, turning over money for their own profit. (5) To help the NGO to articulate its missions, purposes and tasks, to work out a conception of the NGO development, to offer some formats of efficient work and collecting relevant social information coming through the NGO for clarification of the social process and efficient help for suffering people. (6) To build up international projects among relatives’ NGOs in the Nordic and Baltic regions and initiate their uniting and networking. Keywords: drug-addiction in Russia, family and person in front of drug-addiction, parental NGOs’ potential, types of positive results against drug-addiction Research on Childhood and Children as Researchers: The Participatory Methodologies on Social Worlds of Children Natália Fernandes Soares, Manuel Jacinto Sarmento and Catarina Tomás, Instituto de Estudos da Criança da Universidade do Minho- Portugal, [email protected], [email protected], [email protected] In the second modernity the childhood participation is assumed as a fundamental principle to a large extent of the scientific speeches that are produced concerning childhood. The Sociology of Childhood, when considering the children as social actors and as citizens of rights, assumes the question of children’s participation as central in the definition of a social status of childhood and in the characterization of its scientific field. Consider children’s participation in research is a recurrent step in the construction of a discipline of social sciences that looks for listen children’s voice, that assumes children as full social actors, competent in the formularization of interpretations on its worlds of life and revealers of the social realities in which they are insert. The participatory methodologies with children attribute to the youngest a statute of knowledge citizens, and not of simple object, instituting collaborative forms of knowledge construction in social sciences, which are articulated with ways of pledged knowing production in the social transformation and the extension of social rights. The participatory research with children rises, however, special epistemological difficulties, related with the alterity of childhood and with the diversity of its existence conditions. The perception as the other of the adult, elapses from the recognition of childhood cultures as specific way, generationally constructed, of the world’s interpretation and representation. It is the translation work between the language of social sciences and the language of children (with its distinct cultural grammars) that the participatory methodologies are called to play, with the consequent refusal of the generational ethnocentrism and with the indispensable mobilization of a polyphonic speech, where the voice of the collaborative child-investigators keep up, side by side, with the interpretative work of the sociologists of childhood. The attention to childhood’s diversity, elapsing from social categories as gender, religion, ethnics, health and age group, imposes the refusal standardizing looks, defying the methodological imagination to an acceptance and respect for the differences and the diverse ways of its communication. This paper, steady in participatory research work, developed in the Sociology of Childhood, interrogates the knowledge’s construction of the social sciences with and on children and childhood, considering the alterity and diversity, and including an ethical debate on the knowledge creation about the social worlds of childhood. Keywords: children, rights, participation, participatory methodologies 183 SESSION: EVALUATING COGNITIVE QUESTIONNAIRES METHODS TO PRE-TEST Organizors: Ger Snijkers1 and Gordon Willis2 1 Statistics Netherlands, Division of Social and Spatial Statistics, and Utrecht University, Department of Methodology and Statistics Correspondence address: P.O. Box 4481, NL-6401 CZ Heerlen, The Netherlands, [email protected]; Tel: + 31 45 570 6995 ; Fax: +31 45 570 6272 2 Applied Research Program/NCI, 6130 Executive Blvd., MSC 7344, EPN 4005, Bethesda, MD 208927344, USA, Tel:+ 1 301 594 6652; Fax: +1 301 435 3710, [email protected] Keywords: Cognitive Laboratory Methods, Pre-testing Questionnaires Abstract In this session, methodological issues on cognitive pre-testing research will be discussed. Pre-testing of questionnaires in cognitive laboratories was initiated some 20 years ago. Today, a number of methods are available. However, a number of research questions related to the use of these methods remain unanswered. In this session, we would like to address two evaluation issues: 1) Improving questionnaires using cognitive pre-test methods. Questionnaires are pre-tested in order to improve them with regard to the Question-and-Answer process, i.e. getting questionnaires that are easily understood and better answerable. The ultimate goal is getting valid survey data. Here, the following research questions are put forward: Are pre-tested and (accordingly) adapted questionnaires really improved questionnaires? That is: Are those questionnaires actually more respondent (and interviewer) friendly? Do those questionnaires actually result in more valid survey data? 2) Standardisation of methods and design of cognitive laboratory research. Cognitive laboratory research is carried out on different ways. In order to make results comparable across labs, a discussion on standardisation of methods is worthwhile. This includes a discussion on designing pre-test research. The research questions that are put forward, here, are: How do you carry out cognitive laboratory research? Why? When pre-testing a questionnaire, what methods do you combine? Why? Researchers dealing with these issues are invited to present their research. Research papers dealing with cognitive research to pre-test questionnaires in general are also welcome. Examining Expert Reviews as a Pretest Method Terry DeMaio and Ashley Landreth, U.S. Census Bureau, [email protected] Expert reviews are frequently used as a method of evaluating pretest questionnaires. Either alone or in combination with other methods, people who have theoretical questionnaire knowledge or practical experience are asked to review draft questionnaires with an eye to identifying questionnaire problems. This can be done either by having individuals review the questionnaire alone or convening a group, also known as an “expert panel”. Presser and Blair (1994) included expert panels in their research on the effectiveness and reliability of different methods of pretesting questionnaires. But to our knowledge no work has been done to evaluate the consistency of the results produced by individual expert reviewers. In the interest of time and resources, is it plausible that in the early stages of pretesting most expert reviews are conducted by individual reviewers. 184 As part of an experiment on alternative cognitive interviewing methods, expert reviews were used to evaluate the completeness of the problems found by the three different teams in conducting their interviews. Three experts from three government agencies were recruited and given a set of instructions for reviewing the same questionnaire used in the cognitive interviews, a 48-question general population survey on recycling. Each expert was provided with paper forms and asked to enumerate problems question by question. They were also asked to identify the five worst questions within the questionnaire and the five worst (i.e., most major) problems with the questionnaire and the question numbers that reflect that problem. A coding appraisal scheme (used by Rothgeb, Willis, and Forsyth) was used to code the reported problems according to problem type. The paper describes the results of this exercise with expert reviewers. Keywords: expert reviews, pretesting evaluation Examining Think Aloud as a Pretest Method: The State of the Art of think aloud in Cognitive Psychology Tony Hak, Rotterdam School of Management, Erasmus University Rotterdam, the Netherlands, [email protected] In current pre-testing practice the term ‘cognitive interviewing’ refers to two main techniques, think aloud and probing. Although these two techniques are very different in terms of their aims and of their methodological status, they are normally not clearly distinguished in the practice of cognitive interviewing. It is, for instance, quite common to present different types of probes as instances of ‘retrospective think aloud’. There is also a certain resistance to practicing think aloud, either due to practical problems (“our respondents are not able to do this”) or motivated by a methodological criticism according to which think aloud, as an artificial device imposing a difficult task upon the respondent, is pictured as a technique which violates the principle of ecological validity. This paper describes how think aloud was developed originally and was used by (cognitive) psychologists as a technique for producing data about the process of thinking. Its aim was to make this process, that normally is hidden, observable by asking subjects to verbalize their thoughts concurrently, i.e. at the very moment they think them. Although it was debated from its very beginnings whether this can be done at all without changing the process of thinking itself, the think aloud technique is still widely used in cognitive psychology. This paper describes the state-of-the-art of think aloud in cognitive psychology. It discusses issues such as the current standards of application of this technique; the range of research questions it is used for; and its methodological status within cognitive psychology. Finally, the paper discusses whether and, if so, what we can learn from current practice in cognitive psychology for our pre-testing practices. Toward the Accumulation of Cognitive Interview Findings: Considerations in the Design of a Tested-Question Database Paul Beatty and Kristen Miller, National Center for Health Statistics, Centers for Disease Control and Prevention, USA, [email protected] While cognitive interviewing has made great contributions to the development and evaluation of survey questionnaires over the last twenty years, it has also been plagued by a few recurring problems. First, the method has evolved somewhat differently in different organizations, making it difficult to determine exactly what happened when someone claims to have done “cognitive interviewing.” It is therefore possible that researchers may sometimes use the term to describe evaluation projects that are actually quite different methodologically. Second, cognitive interviewing projects are generally geared toward the evaluation of particular questionnaires. While this is understandable, the result is often that findings do not accumulate over time, and thus every new cognitive interviewing project may “reinvent the wheel” by basically exploring the same issues repeatedly. Third, it is often difficult for 185 people not directly involved in a particular cognitive interviewing project to evaluate the quality of recommendations for questionnaire revisions. Such recommendations may be presented without clear documentation of the evidence that was used to make those recommendations; ultimately, this undermines the value of the work, again leading to unnecessary duplication of effort. The overarching problem is that cognitive interview findings do not accumulate, largely because documentation of efforts leaves much to be desired. In an effort to improve this situation, researchers at the National Center for Health Statistics have been working to develop a database of cognitive interviewing findings known as Q-BANK. This database catalogs survey questions that have been tested by various U.S. statistical agencies and links them to a variety of additional qualitative and quantitative data. For example, Q-BANK can be used to identify cognitive tested questions according to subject matter (e.g. asthma questions, cancer questions, demographics), question type (e.g. objective characteristics, behavioral reports, attitudes), response category type (e.g. yes/no, open-ended, quantity), and coded response error type (e.g. problems with terms, recall problems). The database also links questions to textual explanations of findings, documentation of methodology used, and other detailed qualitative information. The database effectively warehouses reports and allows for them to be searched systematically for a variety of purposes. Q-BANK has the potential to help cognitive interviewing findings accumulate toward larger-scale questionnaire design lessons, and to facilitate greater standardization of both cognitive interviewing and questionnaire design strategies. The purpose of this presentation is not to demonstrate the technical capabilities of Q-BANK as much as to discuss the methodological issues that it was designed to address, and to show how it might advance the art and science of cognitive interview evaluation of questionnaires. Keywords: cognitive interviewing, questionnaire evaluation and testing, Q-BANK, characteristics of survey questions, response errors Looking for Trouble and Finding It, Sometimes: Exploring Relationships between Presurvey and Post-survey Evaluation Data James L. Esposito, Bureau of Labor Statistics, USA, [email protected] Using a new Current Population Survey (CPS) supplement on cell phone use as the investigative context, this case study explores relationships between pre-survey evaluation data (drawn from cognitive interviews and a questionnaire appraisal coding system) and post-survey evaluation data (drawn from behavior coding and interviewer debriefings). Using qualitative data from cognitive interviews and the questionnaire appraisal system (Willis and Lessler, 1999), predictions were formulated as to where problems with the supplement questionnaire might occur during its administration in February 2004. Evidence of problems was based on behavior-coding data from 60 household interviews and on qualitative data from two focus groups conducted with CPS interviewers. Though subjective (i.e., no method of quantifying measurement error was available), the accuracy of predictions was assessed using post-survey evaluation data. A summary of predictive “hits” and “misses” is provided and discussed within the context of a larger questionnaire-design-and-evaluation framework (Esposito, 2002, 2003) that relates pre-survey and post-survey evaluation work. Keywords: Cognitive interviews, questionnaire appraisal system, behavior coding, interviewer debriefing, cell phone use supplement 186 SESSION: SECONDARY ANALYSIS OF QUALITATIVE DATA Organizor: Louise Corti Contact: UK Data Archive, Economic and Social Data Service, University of Essex, Colchester, UK, [email protected] Keywords: Secondary analysis; qualitative data; data archives Abstract Archived qualitative data are a rich and unique yet often unexploited source of research material that can be reanalysed, reworked, compared with contemporary data. This session aims to debate the methodological, ethical and theoretical considerations relating to the secondary analysis of qualitative data. In the first instance, papers will be considered that either present case studies based on re-using qualitative data, or draw out the strengths and weaknesses of particular secondary analytic approaches. Experiences with Secondary Use of Qualitative Data Material – a Feasibility Study with German Qualitative Social Researchers Diane Opitz1 and Reiner Mauer2 1 Graduate School of Social Sciences, Bremen, Germany 2 Central Archive for Empirical Social Research, Cologne, Germany The growing importance of qualitative methods since the 1970s in Germany, the progress of computer-supported data collection and preparation as well as the development of data analysis software, they offer - in combination with meanwhile extensively produced data material - a chance to test whether and how secondary analyses of digitalized qualitative data can contribute to reduce the costs arising from collecting, storing and transcribing new interview data. Due to time and/or financial restrictions, research projects are often not able to make use of the whole variety of possible qualitative analysis methods applicable to the data. Moreover, they usually cannot exhaust the complete analysis potential of their own data. On the background of these considerations the German Research Foundation (DFG) financed a cooperation project of GSSS and ZA to explore the possibilities of an archive for qualitative data. In this context a survey of researchers was undertaken to find out whether there is an interest for the secondary use of archived data and if researchers are ready to give their own data away for research and academic teaching. Furthermore, the extent of the concrete interest in secondary use of available verbal data material should be found out, including sources of data being used and the purposes of secondary use. The aim was to learn about the advantages and disadvantages from a researcher’s point of view of carrying out secondary- and reanalysis. The contribution wants to present first results of the feasibility study for a German archive for qualitative data. Are German researchers interested in secondary use of qualitative data, and which criteria should the data meet? We will try to estimate the chances for building up a professional nationwide archive for qualitative data in Germany. Secondary Analysis of Panel Data: Establishing and Using a Text Databank Irena Medjedović and Andreas Witzel, Bremen, Germany Despite the posibilities of using a secondary analysis in social science there are furthermore reservations and ideas of problems about this method. In our opinion the lack of methodological experiences dominate scepticism. In our contribution we want to show with an example of biographical interview-data of a panel-study, how to work with secondary analysis using the progression of data administration programs, which functions of text analysis systems. 187 We want to demonstrate coding- and retrieval-procedures in order to overcome the special demands of panel-data. Taking into account of the discussion of using a secondary analysis which concentrates on data criteria for a good documentation, we want to show, that in different evaluation phases and depending on type and complexity of the objectives of analysis special memos are necessary, which also could be used for secondary analysis if prepared for a data documentation. The demand for secondary data are mostly restricted to original data. But on the basis of our experiences even inductive analysis is possible when falling back on category schemes of the original study. If such schemes have the function of a heuristic way as “container” or “oversized card-index box” the analysing of text data using categories must not contradict to open-ended coding in the process of creating in-vivo-codes. The Anonymization Process for Secondary Use of Qualitative Data D. Thomson, L. Bzdel, K. Golden-Biddle, T. Reay and C. Estabrooks, University of Alberta, Canada Secondary use of qualitative data is only possible if that data has been “anonymized” – that is, if information which could identify research participants has been removed. This anonymizing process is required by most if not all Research Ethics Boards; it allows researchers to honour commitments of confidentiality originally made to participants, while permitting the use of this data to develop further knowledge. Anonymizing is a part of qualitative work that does not receive much attention, yet close analysis shows that the process is full of methodological, ethical and theoretical tensions. Qualitative research focuses on how people live and act in very particular, situated contexts. Removing identifying information also, inevitably, removes contextual information that has potential value to the researcher. Hence, the central questions for a study of anonymization are: How can researchers best anonymize qualitative data for secondary use? What important contextual information is lost in the anonymization process? How do we minimize these losses, preserving the situated and contextual nature of the information, while honouring the confidentiality and privacy of research participants? How do we assess quality of research when only anonymized stories are presented? We propose to present a case study of working with anonymized data on the research project, Knowledge Utilization and Policy Implementation (KUPI), a five-year program funded by the Canadian Institutes for Health Research. This project involves the secondary use of qualitative data sets from three separate research projects across Canada. Each project employed different methods of qualitative data collection that were originally anonymized following different protocols. Furthermore, the researchers on the KUPI project have differing levels of familiarity with the original, nonanonymized data. Based on this case study, we will consider the strengths and weaknesses of the anonymization process and provide useful recommendations that address some of the central questions of anonymization. Revisiting ‘Classic’ Qualitative Sociology Mike Savage, University of Manchester, UK This paper explores methodological issues regarding the revisiting of ‘classic’ qualitative studies. Classic studies pose particular issues for secondary analysis over and above those which specifically relate to the secondary analysis of qualitative data itself. By virtue of being ‘classic’, the findings and arguments of such studies define a subsequent ‘canon’ of theoretical and methodological scholarship, and hence shape the thinking of subsequent researchers conducting secondary analysis. Secondary reanalysis therefore should be not only of the archived data itself, but of the published work itself, but this raises a host of complex methodological and ethical issues 188 Using my own reanalysis of Elizabeth Bott’s ‘Family and Social Network’ archive, and John Goldthorpe and David Lockwood’s ‘Affluent Worker collection’, I examine possible analytical strategies for re-analysis, including ‘debunking’, the alternative of ‘sacralisation’, and ways in which original data can be read ‘against the grain’. Strategies in Teaching Secondary Analysis of Qualitative Data Louise Corti and Libby Bishop Unlike for quantitative approaches, the published literature on secondary analysis of qualitative data is sparse. Among qualitative researchers there is neither a pervasive culture that exists to encourage the secondary analysis of data nor is there an extensive body of published sources that can help guide or instruct a researcher wishing to understand better the benefits and limitations of re-analysis. So where does the student or novice researcher turn for guidance and training? In this paper we set out ways in which EDSD Qualidata has been trying to facilitate both usage of archived qualitative data and methodological debate among the wider academic communities. We will present an overview of the current published literature and existing training provision for secondary analysis of qualitative data and describe the various approaches to support and training taken by our service. Finally, we will cover the preparation of potentially useful kinds of training materials (teaching datasets, user guides, commentary and exercises) that we believe will help support teachers and learners. Reanalyzing Qualitative Interviews from Different Angles: The Merits of Sharing Qualitative Data Harry van den Berg, Department of Social Research Methodology, Vrije Universiteit, Amsterdam, The Netherlands, [email protected] In contrast to survey interviews, qualitative interviews are seldom reanalyzed. Besides obvious reasons such as ownership – and especially the culture of individualistic ownership – that impede reusing data, there are also important methodological arguments against secondary analysis. The main methodological argument against secondary analysis concerns the supposed risk of decontextualization. In this paper, I will argue in favor of sharing qualitative data on behalf of secondary analysis. The argument is partly based on – and much inspired by - the results of a collaborative project of thirteen researchers who were invited to analyze the same set of interview data from their own theoretical/methodological viewpoint (Van den Berg, Wetherell and Houtkoop-Steenstra, 2004, Analyzing Race Talk; Multidisciplinary Approaches to the Interview. Cambridge: Cambridge University Press). These results show the possible merits of sharing qualitative data such as 1) testing the validity of the original analysis, 2) enriching, refining and complementing the original analysis, 3) reusing interviews on behalf of different research goals and 4) comparing different theoretical approaches used to analyze the same data. In order to evaluate those possible merits more carefully, it is necessary to scrutinize the argument concerning the risk of decontextualization. Different theoretical and methodological positions concerning the contextualisation of interview discourse are discussed. The main conclusion is that the kind and measure of contextualization of interview data needed on behalf of discourse analysis is mainly dependent on the research goal. As a consequence two opposing methodological positions are criticized: 1) On the one hand the methodological position that promotes the inclusion of the ever-widening societal and historical context as a necessary precondition of the analysis of interview discourse. This position runs the risk of speculative social theorizing as a framework for interpreting interview discourse. 189 2) On the other hand the methodological position according to which the analysis of interview discourse should be restricted to the activities of interviewer and interviewee. It may be doubted that the neglect of every social context outside interview talk – as advocated by some strands within Conversation Analysis – is fruitful or even possible. This neglect runs the risk of abstract empiricism. Keywords: Qualitative interview, discourse analysis, secondary analysis SESSION: RETHINKING PROFESSIONAL ETHICS: THE HIDDEN RESEARCHER AND THE UNINFORMED INFORMANT. Organizors: Liora Gvion1 and Diana Luzzatto2 Contact: 1The Kibbutzim College of Education, Tel Aviv, Israel 2 The Academic College of Tel Aviv Yaffo, Israel Keywords: public location, informants, research field, ethic dilemmas, professional ethics. Abstract Traditionally, professional ethics has required the researcher to notify the community of interest of his intentions to conduct research. Unobtrusive methods, although efficient, have been regarded as illegitimate as they have violated the informants' rights of privacy. However, unconditioned transparence poses certain limitations especially when conducting short-term observations in public locations in which encounters tend to take place occasionally and on an on-and-off basis. Moreover, taking place in a public sphere, interactions among participants tend to be general on one hand, yet reveal upon interesting life aspects of private lives generally sealed off to the researcher in fixed, close boundary bind communities. This session brings forward major dilemmas, often shouldered aside in qualitative research. First, is it unethical to choose public locations, such as restaurants, schoolyards, or beauty parlors as research fields without acquiring permission to do so? Second, who is entitled to grant permission as such? Third, to what extent can a research integrate data observed out of the field into his current research? Fourth, what is a public domain and is it a potential legitimate research arena as it is open to all? Fifth, is there an unwritten contract between researchers and potential uninformed informants? 'Doing Ethics’ In The Field: The Art And Politics Of Covert Research David Calvey, Manchester Metropolitan University, UK This paper discusses the ethical dimensions of fieldwork with reference to a six-month covert participant observational study of door supervisors, or more traditionally ‘bouncers’, in the leisure and entertainment sector of Manchester, which has been part of an urban mythology as regards gangsterism. The context of this study is the increasing public discourse on the professionalization and criminalization of this group in the wider context of the governance of the night-time economy. More generally, the examination of the work culture of bouncers is one, which offers some insights into a sociology of cultural practice and social organization. Central to the accomplishment of an ongoing door order is the interactional management of respect on the door. That is door work as involving a range of organizational issues as removed from romanticized and mythologized treatments. Despite the increasing work done on club cultures (Thornton, 1995, Redhead et al, 1997, Malbon, 1999), new masculinities and crime (Winlow, 1999) and violence and governance in the night-time economy associated with bouncers (Hobbs et al, 2003) the dedicated and faithful treatment of the topic has been marginalized, subsumed or obviated. In short, the phenomenon has been lost under exotica. 190 More specifically, key vignettes are taken from the ethnomethodologically inspired ethnography of the work culture of bouncers to articulate some critical reflections on self and other in the field via the concepts of nomadic ethnography and post-fieldwork biography. The idea of methodological appropriateness is used to articulate the covert stance taken with the study. In particular, and arising from my dualistic lived field experience of ‘going bouncer’, the paper wishes to raise for scrutiny the relationship between reflexivity and ethics in covert fieldwork. Despite the increasing recognition and discussion about risk and danger in field research (Lee, 1993, 1995; Treweek and Linkogle, 2000) the activity of managing situated field ethics remains glossed over. The standard professional discourse on ethics is abstracted from the actual doing, which is a situated, contingent, collaborative and temporal matter. This study forms part of a wider engagement with reflexive ethnography, research ethics, cultural studies, organisation studies and ethnomethodology. Methodological Issues Experienced in a Comparative Study of Children’s Free Time Activities Sidsel Hadler-Olsen1 and Doug Springate2 1 Bergen University College, Bergen, Norway 2 The University of Greenwich, London, England This paper discusses methodological considerations in relation to a 3-year comparative study focussing on children’s out of school activities. The study concerns selected groups of English and Norwegian 10 year old children in various week periods between 1996 and 1999. The main aim was to explore what these children did, where they did these activities and with whom. The problem was how to get meaningful data. The informants were young and were being asked to record their activities and behaviour. The children were not passive in the research process but active in their interpretation of what was being asked and in the selection of what they wanted to share. In this paper we discuss the particular issues of dealing with children as informants in studies of this kind. The researchers must be both innovative and respectful in dealing with young children. In the first year of the study the main research methods were a combination of diary keeping by the children and selected interviews. These methods were improved from year to year and were supplemented by group discussions, mapping, questionnaires and television diaries. Throughout the study we became more aware of the issue of how researchers can meaningfully access children’s lives outside their school time. How did the children perceive the tasks we gave them? Are there ethical issues implied in such tasks when the tasks are given within s school context? Do the children perceive the tasks as part of their school work? Are there communication issues between children and adult researchers? These issues are compounded by the comparative nature of the research as concepts such as “play” and “childhood” have different meanings in the two countries. Keywords: Children as subjects, ethics and legalities; Children as respondents; Communication issues; Cultural differences; Joint research; Qualitative versus quantitative research traditions Making Private Issues Public: Ethnic Dilemmas Versus Moral Obligations Liora Gvion1 and Diana Luzzatto2 1 The Kibbutzim College of Education, Tel-Aviv Israel 2 The Academic College of Tel-Aviv Yaffo, Israel This paper discusses means to construct and create an informed consent in cases in which information is revealed in the public sphere yet not made consciously available to an ethnographer. It is our claim that in cases as such, informants’ privacy is maintained by bridging between the private sphere, where personal and family matters revolving around emotions such as pain, happiness, and so forth, are 191 managed, and professional duty to uncover meaningful data, available in public settings. Such a state of affairs raises a moral and ethical dilemma. One the one hand, private information revealed through dialogues among individuals talking in public settings is meant, according to normative perceptions, to remain private. On the other hand, the assumption above rests on what one may call imaginary social boundaries not protected by any formal code. The purpose of this study is threefold. First, we reveal the possibility of obtaining spontaneous information, produced in an un-structured event and hard to obtain in interviews or explicit participant observations. Second, we claim legitimacy to make use of information as such as research ethics provide enough means to conceal private identities and protect the informants. Third, this study points out the richness embedded in the ethnographer’s social circles and personal network and their potential contribution to research. The case study analyzes the negotiation process held by parents of children with minor communication disorders with the local placement-committee in the Tel-Aviv area in order to gain control over their children’s placement in the school system. Data were obtained in a sport club that provided physical education classes for children with special needs. While the children were taking their classes the parents were discussing their experiences as parents of children with special needs. One of the researchers was part of the setting and therefore often involved in the conversation. She was sharing information as well as listening to other parents’ experience. Once realizing the amount of information as well as the social issues involved, she decided to turn her personal experience to a research field. In order to overcome ethical problems, parents were informed about her decision, their consent was obtained and another researcher conducted all in-depth interviews. SESSION: FEMINISM, METHODOLOGY AND METHODS Organizor: Sharlene Hesse-Biber Contact: Department of Sociology, Boston College, 140 Commonwealth Avenue, Chesnut Hill, MA. 02467, USA, Tel: 617-5524139 Fax: 617-55204283, [email protected] Keywords: feminism, methods, methodology, feminist methods, epistemology Abstract This session will consider cutting edge papers which deal with the intersection between feminism methodology and methods. We will consider both theoretical and empirical papers on any aspects of this topic. We are interested in some of the following issues: What is the role of power and authority in the research process? Are there feminist ways of knowing? Is there a feminist method of knowledge building? How to understand the role of the researcher and researched within the interview process. Issues of difference in Feminist practice: How do race, class, gender and sexuality together form standpoint ? How to account for and work with difference during the lived experience of research ? What are the ethical questions and challenges researchers face as they incorporate difference into their projects? How to critically deconstruct the interplay of race, class, gender and sexuality during the practice of research. What are some critical issues in the analysis/interpretation, data management, and writing up of feminist scholarship? Part I Critical Rationalism as a Tool for Feminist Comparative Historical Sociology Judith Buber Agassi, Tel- Aviv University, Israel Critical rationalism is the assumption that we can approach the truth about social reality. As in the natural sciences, this process of inquiry begins with problems and proceeds with tentative solutions to be critically examined. Criticism of extant theories leads to the search for alternatives to them. 192 I begin with the observation that in Nazi concentration camps characteristics of larger organizations of prisoners were similar for males and females, but that small groups – so-called camp-families – were common among female Jewish prisoners and rare among males. This seems to agree with the popular theory of the importance of gender-specific psychological factors. Yet any detailed study shows that this characteristic is the outcome of Jewish education that demands much more of girls than of boys that they act protectively towards the weaker, young and old in the family. The second popular extant theory – of Friedrich Engels – that the inferior social status of women is due to the division of society into classes, so that the socialist revolution will wipe it out by the abolition of both private ownership of the means of production and the raising of children in the nuclear family. As in Kibbutz society property was communal and children were raised in groups outside the home of their parents, the expectation was that women would enjoy equal status there. This did not happen. My explanation for this disparity is that the division of labor in the Kibbutz was sextyped, assigning to women overwhelmingly service, non-progressive and localized work duties. The Clear Feminist Rejection of an Obscure Positivism Maria Ignez Silveira Paulilo, Universidad Federal de Santa Catarina (UFSC), Brazil The purpose of this essay is to discuss feminist research methods which give special emphasis to the rejection of positivism. This rejection is based on a stereotyped and simplified vision which considers significantly different thinkers, such as the sociologists Durkheim and Weber, as equally “positivistic”. This leads to a strong opposition between quantitative and qualitative research methods. Researchers eventually follow, in practice, principles that theoretically would be denied. The criticism of established science winds up approaching one of the most classic formulations of positive empiricism which is that “well collected data should speak for itself” and that is enough to “given voice” to the oppressed women. The use of interview as a consciousness-raising process needs to consider the possibility of a lack of correspondence between aspirations and concrete situations, i.e., the non-existence of corresponding channels available for action. Sometimes the interview-based research as a form of making women aware of their rights is more an authoritarian intrusion and imposition in the life of people than benefits. Third world women want to be regarded not as victims but as people with their own culture and values. Keywords: Positivism; Feminism; Qualitative Analysis; Interview-based Research Entering Sites; Entering Ethnography Anastasia Norton, Sociology Department, Brandeis University, MA, USA Nancy Scheper-Hughes says that the ‘work’ of ethnography entails, at its very core, “the working out of an ethical orientation to the other-than-oneself.” These strategies of successful ethnographic “arrival” in research sites have real theoretical and epistemological implications. The manner in which researchers enter, live, depart and return and the ways that we understand those experiences define the practice of ethnography. And yet, successful arrival includes more than the physical movement from point ‘A’ to point ‘B’. In this paper I explore the processes involved with arrival in a field-site and the implications for theory construction. I have developed a typology of arrival that explores both the multiple realms involved in travel as well as the complex levels located within each realm. I use my own field experiences in a Maroon village in Suriname to explore the spatial, temporal and emotive travels that I embarked upon in order to arrive in my field-site. This typology of arrival, coupled with an investigation of feminist methodology and theory can lead to the development of more critical, selfreflexive and contextually mediated ethnography. Keywords: Ethnography, Feminist Methods, Suriname Maroons 193 Power Tools: A Feminist Approach to Large Sample Quantitative Data Analysis Allison J. Tracy, Wellesley College, Wellesley, MA, USA Overgeneralization has long been of central concern for feminist researchers. On one hand, a feminist critique of traditional approaches to data analysis has been that reliance on aggregate statistics, such as means and covariance structures, captures only the most typical experiences in the sample while “disappearing” experiences less typical. Many feminist researchers have dealt with this concern by carefully attending to each voice in their sample, assuming heterogeneity of experience but necessarily limiting their sample to a feasible size. A common critique of these small scale studies is that the results, again, may not generalize well to a larger group. While some feminist researchers downplay the role of generalizability as a primary goal of their research, the influence of study results on broad public policy frequently hinges on a demonstration of the applicability of the findings to a much larger subpopulation of society. Indeed, evoking large scale shifts in social policy is a primary focus of feminist activism. The apparent tension between quantitative and qualitative data analysis paradigms has generated a great deal of debate and has resulted in the proposal of a variety of mixed methods approaches. This paper will offer and demonstrate a newly available mixed method approach that integrates the assumption of population heterogeneity with the power of large sample statistical modeling, yielding results that simultaneously honor both typical and atypical experiences. Recent innovations in quantitative statistical software have made possible the analysis of a large range of latent variable mixture models. These models allow for unobserved heterogeneity in the means and covariance structure. Simply put, researchers may now empirically identify qualitatively diverse subpopulations in their sample data representing sets of different experiences, without anticipating these differences prior to analysis. Further, model-based residuals can be used to identify and study in greater depth those individuals whose experiences are not adequately captured by the mixture model. To illustrate this new approach, an empirical example will show how gender, race/ethnicity, region, social class, and other social location indicators predict various classes of sexual risk in young adulthood and how physical activity in adolescence influences the degree of sexual risk within each distinct class. Followup analysis of the model-based residuals will show how the use of extreme and typical case sampling can be used to explore mechanisms that underlie these processes, generating directions for the development of culturally appropriate interventions and public policies. Keywords: Methodology, Feminism, Mixture Modeling, Mixed Methods Part II Gender Imago: Searching for New Feminist Methodologies Niza Yanay and Nitza Berkovitch, Ben Gurion University, Israel Invoking Butler’s notion of gender performativity, Kristeva’s concept of foreignness and Laplanche’s reconceptualization of otherness, we examine the power of fantasy to change the women and men that we always already are. Using ‘writing- in-response’ we discuss the meaning of gender performance in relation to our theoretical commitments. The paper is structured around three different forms of dialogue: (1) Two lectures that we presented, each one contesting accepted ideas of gender, self, and society; (2) Seven e-mail correspondences that develop the ideas presented in the lectures and that dramatize the transition from speaking to -writing-in-response-; and (3) A discussion, developed both together and separately, that raises the possibility of exploring a new language of gendered subjectivity. The paper challenges the concept of “direct experience”, the separation between psychology and sociology, and destabilizes the space between gender fantasy and performance. 194 The Problem of Methods on Feminist Epistemologies: Discussing their Ideological and Political Background as Opposed to Solutions of Concrete Scientific Research Subjects Maricela Guzmán Cáceres and Augusto Renato Pérez Mayo, Universidad del Valle de México campus Villahermosa The term of feminist epistemologies is commonly utilised in relation to a heterogeneous number of studies embracing a wide diversity of standpoints, both concerning epistemology and feminism. All of them disagree with that traditional epistemology, whose weakness lies in the fact that it pretends to construct a general theory that ignores the specific social contexts. From this perspective, the individual is regarded as an abstraction expressing universal and untainted qualities in terms of reasoning and feeling (do you mean perceiving? If so you can put perceiving). From the feminist view, it is argued that that one to acquire knowledge is a particular and historic individual, whose body, interests, emotions and consciousness are made up by a concrete historic context and are relevant for epistemology- This relevance lies in the fact that knowledge is always located (probably contextualised is more convenient) (Haraway, 1991). This means that (knowledge?) is shaped by the individual and his/her particular situation. (spatial, temporal, historic, cultural and social) and that, the margins of justification are always contextual (this ideas was not quite clear in Spanish) From this location, derives the connection between knowledge and power. The political commitment with social change is one of the main constituent aspects of feminist epistemologies and also one of the main features, which differentiate them from other theories of knowledge. This paper discusses the eminently ideological and politic character of the proposals established by the so called feminist epistemologies. It stresses their weakness in relation to the methods that they use to operationalise their epistemological paradigms. Following Bunge (1980:21-22) I argue that these methods do not meet the five requirements, which according to this author, every epistemology ought to have: a) if it concerns science in the proper sense; b) if discuss philosophical problems which are present through scientific research; c) if propose clear solutions for such problems; if it is able to distinguish between authentic science and pseudo-science; d) if it has strength to criticise programs and suggest new results. Keywords: Feminist Epistemologies, Methods, Methodology, Feminist Methods. The Feminist Practice of Oral History and the Merging of Personal and Social Problems Patricia Leavy, Stonehill College, Easton MA, USA This paper explores the feminist practice of oral history as a method of linking individual narratives to macro phenomena. In particular, this paper examines how through a thematic data analysis process, feminist researchers can use oral history transcripts as a way of individualizing social problems by allowing thematic categories to emerge out of the narrative and showing how while the individual’s particular narrative is unique, the themes can tell us about the experiences of many. The paper discusses how the researcher analyzed an original oral history transcript about one woman’s struggle with an eating disorder as a method of individualizing a large-scale social problem (that primarily impacts females). Womanhood at the Margins: Notes of Researching and Resolving the Stigmatization of “Undesirable” Women” Yu-Wen Fan, New School University, NY, USA The unions of Mainland Chinese veterans and their wives have been widely stigmatized in Taiwan. Exiled from the Mainland, low-ranking Mainland veterans were forced into the bottom of the economic strata. They married Taiwanese women usually characterized as “undesirable.” Lacking indepth observations and feminist perspectives, previous researches have reiterated the stigmatization 195 without questioning or exploring it. Via case studies, this paper presents the images of two disabled women married to Mainland veterans: how they and their husbands viewed the disabilities and to what extent the disabilities and resultant images have affected their lives. In case one, the husband constantly ridiculed the wife’s being unable to speak clearly. This may serve as a way for the husband not to “downgrade” himself to the category of his wife in the presence of me, a female researcher. In case two, a wife with a crippled leg attributed her misfortune to her disability. She could hardly get away from it in spite of her being articulate and having a unique view of life. The other aim of this paper is to reflect whether in-depth observation and meaningful interaction between the researched and the researcher yield a potential discourse of de-stigmatization. STREAM: QUALITATIVE METHODS (2) SESSION: THEORETICAL APPROACHES AND METHODOLOGICAL STRATEGIES TO PRODUCE AND ANALYZE QUALITATIVE EMPIRICAL DATA Organizors: Ruth Sautu1, Silvana K. Figueroa2 and Jochen Dreher2 1 University of Buenos Aires, Argentina, [email protected] 2 University of Konstanz, Germany Keywords: qualitative methodology, hermeneutics, reflexivity, collective identity, actor’s perspective, moral order, collective protests. Abstract Based on empirical research that privilege the social agent´s perspective in the reflexive interpretation of social reality, the objective of the session is to discuss: The influence of theoretical and methodological approaches in the formulation of research objectives. The problems related to “second order interpretation”: data validation. The paper presenters must take these two issues into account to facilitate the interchange of ideas during the session. Papers will be sent in advance to all participants to promote discussion. The papers to be presented cover a variety of recent trends in qualitative research. Authors, substantive areas and methodological approaches are the following: The emergence of new identities in Argentine Armed Forces: a life course research. Alejandra Navarro, University of Buenos Aires. People’s representations on corrupt practices and the construction of a new moral order: in-depth interviews and answers to open-ended questions in a middle-class survey. Ruth Sautu, Paula Boniolo & Ignacia Perugorría, University of Buenos Aires. People’s perceptions on the construction of collective goals: interviews in street meetings of collective protest. Luciano Brom, Pablo Dalle & Rodolfo Elbert, University of Buenos Aires. Changes in the identity of high school students: an etnographic study. Analía Meo, University of Warwick. Following the traces of symbols: an hermeneutic an hermeneutic analysis of paradoxical constructions within Argentine collective symbolism. Jochen Dreher & Silvana Figueroa, University of Konstanz. Images of the Argentine ruling class: focus groups using mass media photographs as stimulus. Valeria Maidana, Gabriela Medín & María Alejandra Otamendi, University of Buenos Aires. 196 Part I: Theoretical and Methodological Issues Transcendental Realism in a Particular Methodological Secuency Beatriz Irene Wehle, University of Quilmes, Argentina, [email protected] This paper aims to describe and analyze the consecutive steps and strategies we followed in our reserarch about the perspective of the juridical agents in the Tribunals of Justice. The subject of our research were the changes in labour oganization and the competencies of the workers in the Courts of Justice in Buenos Aires, Argentina. During our research we made use of different theoretical perspectives in order to make a qualitative analysis of empirical data. We applied qualitative methods which were discussed and examined through putting them into practice. Then, we chose those which were more effective for attaining our objectives. With regard the chosen theoretical frame, from the starting point of inductive inferences, we developed a theoretical perspective close to that of the ethnographical methodology. But this perspective is close too to the action-research: this is so because this research team was asked by an union organization in order to study some current conflicts in the workplaces and work conditions of juridical agents. In this first stage, the prevailing theoretical perspective was related to the participant observation, connected with the interests and concerns of union activists and focused on the descriptive data of labour in a holystic analysis. This first stage was followed by other theoretical perspective, more strictly focused on the critical ethnography, which allowed us to emphasize, through sistematical analysis, hidden structures which were present in these workplaces. From there, we started to work in a stage of no standard instrumentation in order to find underlying organizational models. In the third stage of our research and from the perspective of the reflexive phenomenology, we identified the labour situations in the Tribunals of Justice and analysed the more significant data which were obtained through deep interviews to key informers. Basing on the phenomenological and social interaction perspectives, we categorized the agent’s accounts and explored the possible relationships between the subject’s representations and labour processes as for the agent’s perceptions about their jobs and other more global judgements about service and procedures in justice departments. From this more ethnographic perspective, our work was based on the social antropologists. In short, from the agent’s perspective, in order to come out the particular points of view and reconstruct the processes of colective representations in the enviroment of labour organization in justice administration, through inductive methods and the pragmatism of “trascendental realism” we worked out explaining structures of the present configurations. Keywords: Actor’s perspective, labour processes reform, collective representations Post-communist Systemic Transitions, National Identity and Mass Media: Theoretical and Methodological Aspects of Analysis Izabella Zandberg, Westat, USA, [email protected] This presentation outlines theoretical and methodological aspects of analysis of the relationship between systemic change, national identity and mass media. The author suggests that revolutions, including those of 1989, need to be analyzed as cultural processes and that systemic transitions are, in part, attempts at cultural engineering. Construction of collective identities is an important part of those efforts since collective identities have the potential of becoming national identity, capable of exercising political sovereignty and playing a defining role in the developments of national and international communities. Their analysis needs to be undertaken in reference to a particular historical, social, and economic context in which these collective identities are being spun. Mass media are a forum in which this construction takes place and therefore their contents provide rich research material for analysis of meaning-making processes accompanying structural change. 197 Based on a proposition by John J. Pauly (1991) that literature review in qualitative research maps the discourse of the field by simply “identifying an ongoing conversation that the researcher now proposes to join” (p.9), this study reviews related research conducted in various geographic and historical locations, and applies the proposed methodology to the particular case of post-Communist Poland. Drawing on the insights of social constructionist theory, this methodology involves discourse analysis of selected media texts that were produced within different ideological frames (liberal and antiliberal), and pertained to the construction or assertion of collective identity, contributed to the construction of systems of knowledge, and expressed beliefs about reality. A thematic analysis of media contents is proposed, involving a three-stage process: 1) identification of patterns such as “conversation topics, vocabulary, recurring activities, feelings, folk sayings” that are related to the analytic focus on collective identity and reality construction (Taylor & Bogdan, 1989, p. 131); 2) combination of patterns into themes by “bringing together components or fragments of ideas or experiences, which often are meaningless when viewed alone” (Leininger, 1985, p. 60); 3) identification of themes through seeking “ideas or components that would fit together in a meaningful way when linked together” (Ibid.). The proposed methodology is applied to the analysis of the contents of a radical ultra-right Catholic medium – Radio Maryja (Radio Mary) and findings are briefly presented. Possible methods of data validation are discussed. Keywords: collective identity, national identity, post-communist transitions, mass media, discourse analysis “Representations” of Death: Theoretical and Methodological Approaches Christine Matter, University of Konstanz, Germany, [email protected] The experience and the perception of temporal boundaries of individual and collective life has become an important topic of social scientific research in recent years. Death and dying which have been neglected as research issues within sociology for a long time are individual and social events which correspond to a unique boundary ‚experience‘. What is particular about this boundary ‚experience‘ is that although death is certain for everyone it cannot be experienced directly: death is a nonexperiential „empty telos“. This poses specific challenges to the production of qualitative empirical data if one wants to examine conceptions of death, dying, and an eventual ‚after-life‘. The fundamental problem that the phenomenon of death cannot be related to genuine real life experience has to be taken into account in methodological and theoretical approaches. The relationship between the interviewer and the interviewee is shaped by this specific fact: the interviewee is not in the position to talk about an experience he or she has actually gone through and, therefore, is no „expert“ concerning the issue. In the case of death people have to talk about a „limit“ to their existence that has no „other side“. Nevertheless, death is increasingly becoming a topic of discussion in today‘s society: in terms of one‘s own end of life as well as the experienced end of others. The proposed paper addresses methodological and theoretical problems which arise in this context. How is the interview and the interpretation of the material affected by the fact that the subject has no more experience on the topic of his death than the interviewer? The paper is based on empirical data from an ongoing qualitative research project at the University of Konstanz which examines perceptions and „representations“ of death by comparing two different generations of East- and West-Germans: those born around 1930 and those born around 1970. We apply in this analysis Karl Mannheim‘s theoretical concept about the problem of generation and the methodological approach of Grounded Theory. The paper explores the methodological and theoretical possibilities of the social construction of temporal boundaries. It examines the symbolic constructs people apply to the problem of the end of the individual in the context of the two aforementioned groups in German society. Keywords: generation, death, symbol, time, Grounded Theory, hermeneutics 198 Tracing Symbols. The Hermeneutic Analysis of Paradoxical Constructions within Argentine Collective Symbolism Silvana K. Figueroa and Jochen Dreher, University of Konstanz, Germany, [email protected], [email protected] The symbolic work by means of which we extend a network of meaning over our visible and invisible constructions of the world allows us to live within paradoxes. These illustrate one fundamental form of dealing with contradictoriness and oppositionality; they simultaneously represent contradiction and the process of its harmonization. Symbols consolidate and secure these constructions and enable us to combine various seemingly irreconcilable meanings, emotions, values, and tendencies into a unified image. They contract the contradictory into a unit, the not simultaneous into the simultaneous, the congruent into one form. Symbols, especially those we consider collective symbols, fulfill this function of unification. They are products and instruments of human effort, which in turn modifies and maintains the conditions under which we live together in a group, community or society. Collective symbols constitute the feeling of community just as they help ensure the community’s (collective) consciousness and continuity. On the basis of the interpretations of two case studies from our research project “Constructions of Identity in Pluralistic Societies. Processes of Constitution of the Other and the Self in Argentina” we will demonstrate the methodological process which is used to analyse Argentine collective symbols. We consider tango and gaucho to be phenomena responsible for the constitution of communities and collectivities within this society. Many of these collective symbols, that are unifying contradictory elements of specific social worlds, disclose the means not only to analyse the specific “problems” present within these social worlds, such as social cohesion attained on the basis of the construction of identities, the creation of a mutual horizon of meaning or the typification of forms of action. Furthermore, they point us towards concrete “solutions” to specific “problems” within the Argentine society, that can be resolved through collective symbolism. Keywords: collective symbolism, identity construction, hermeneutic analysis, Argentina, tango, gaucho People’s Representations on Corrupt Practices and the Construction of a New Moral Order: In-Depth Interviews to Middle-Class Residents of Buenos Aires Ruth Sautu, Paula Boniolo and Ignacia Perugorría, University of Buenos Aires, Argentina, [email protected], [email protected], [email protected] The objective of this paper is to evaluate the validity of a methodological strategy based on in-depth interviews to study people’s representations of corruption and their beliefs about its effects on the working of a democratic social order. The paper draws on a social representations theoretical approach applied to analyze common people’s own words when describing corrupt mechanisms, the conditions that favour its emergence, and its effects for the democratic social order. In-depth interviews were conducted among a purposeful sample of professionals and business men and women who had information about mechanisms through which corruption operates in both the public sector and private activities. The testimonies content consists in the description of cases of their own experience and in their own interpretations and value judgments of the causes and consequences of corruption. The analytical approach consisted of three interrelated stages: first, the description of the corrupt situations, actors and mechanisms based on the thematic analysis of interviews textual transcriptions; second, the analysis of the categorization system employed by people and their attribution of either social and/or individual causal explanations; and finally, the meaning of corruption in everyday life and the feelings of resignation, acceptance or refusal implicated in the interpretation of corrupt practices. While the first stage leads to a reconstruction of the factual mechanisms of corruption, the 199 second and third aim at achieving an understanding of its significance for the construction of a “desired moral order”. Keywords: social representations, corruption, middle class, qualitative analysis Social Mimesis: A Study about Cultural Performances Vincenzo Romania, University of Padua, Italy, [email protected] Most recent developments in sociology of cultural processes frequently focus on the dimensions of practice and performance as pragmatic expressions of attitudes and behaviour, always associated to a cultural role, or a cultural meaning. This interesting direction of research , both in sociology and in cultural anthropology, has mainly born in to the ambit of ethnographic studies. My paper focuses on social mimesis, a strategic performance which immigrants often play to avoid the stereotypes and the stigma related to their identity, in the local public space. The strategy is played by processes of passing, of occultation, of amputation of individual identity. Following this point of view, I set an empirical research based on the ethnographic paradigm, and conducted in Italy 100 qualitative interviews to a snow-ball sample of Albanian immigrants. The results of my research have important consequence for a dynamic and anti-systemic theory of culture and for a model of social integration which renounces to consider immigrants as ‘simply’ organized in communities, otherwise analysing the individual strategies of insertion in the host society. Keywords: social mimesis, performance, integrations, theory of culture Part II: Methods & Procedural Techniques Impact of Interviewer Attitudes and Socio-Economic Status in an Inter-Ethnic Survey Peter Pillók and Dávid Simon, Eötvös Lóránd University of Sciences, Budapest, Hungary, [email protected], [email protected] Our goal in the recent paper is to investigate the presence and pattern of impact of the interviewers’ attitude towards ethnic minorities in an inter-ethnic comparative study. In our case study we analyse a research conducted in 1999-2000 in Hungary. The purpose of the study was to investigate the Hungarian Roma population and compare it with the majority especially in terms of living conditons. The findings of the research were published by János Ladányi - Iván Szelény and György Csepeli - Dávid Simon, but in these papers the impact of interviewers’ attitudes and socio-economic status was not given detailed investigation. In our paper we try to find biases in the answers of the interviewees caused by the interviewer. Also, because of the design of the research the interviewers had to answer some questions about the interviewees - among others about the ethnicity of the interviewee. For our investigation we used the data of the original research and a survey that was done among the interviewers. The survey consisted of usual socio-economical questions and standard questions concerning attitudes towards the Gypsies, and the items of the F-scale. We also used the data of a national representative sample about the questions of the latter. We investigate the distinction between the interviewer and the representative sample in order to test the hypothesis that the interviewer can represent the majority. We also try to correct the possible biases caused by these distinctions and give more exact estimations from the answers of the interviewers about interviewees from the viewpoint of the majority. In the second part of our paper we describe the possible effects on the answers of the interviewees taking into consideration the similarities between the interviewers and interviewees caused by the sampling design (e.g. location). We also compare these effects on the minority subsample and majority subsample. For these investigations we use multilevel analysis. As a final result, we give an evaluation and suggest some impovement for further similar multicultural surveys. 200 Video-based Analysis of Social Situations: Comparing the Analytic Range of Ethnomethodology and Genre Analysis Dirk vom Lehn1 and Bernt Schnettler2 1 King’s College, London, England, [email protected] 2 TU Berlin, Germany, [email protected] There is a growing interest in the social and cognitive sciences to use video-recordings as principal data to explore human action and interaction. Apart from an increasing body of experimental research a variety of methods and techniques are employed to investigate naturalistic data. This paper focuses on two approaches that have recently begun to use video-recordings for the analysis of ‘naturally’ occurring social situations, ethnomethodology and genre analysis. Studies drawing on ethnomethodology and conversation analysis have developed an analytic and methodological framework to reveal the social and sequential organisation of participant’s action and interaction. They concentrate on the sequential character of participants' conduct and the ways in which they accomplish their action and activity in, and through, interaction with others. The aim of these ethnomethodological studies of interaction is to uncover the “ethnomethods”, that is, the practices and reasoning that participants use to accomplish their actions and activities. Genre analysis was developed mainly to study the structures of oral communication. It combines basic ideas of the social construction of reality with instruments for the detailed investigation of spoken language from linguistics and pragmatics. Genres are socially sanctioned practices and, as they serve as instruments for the communicative construction for the social world, their reconstruction is central for determining the central contents of a particular culture. The aim of genre analysis is to identify typical patterns of speaking practice and investigate their systematic relation to aspects of the social structure and the cultural dimension of a certain society. The paper will use a video-recorded fragment of a social situation to compare and contrast the range of the analytic and methodological frameworks provided by ethnomethodology and genre analysis. It will use the fragment to formulate research objectives based on different theoretical and methodological concepts of the two approaches and analyse the fragment to reveal the different kinds of observations and findings that the two approaches generate. In conclusion, the paper will begin to explore how ethnomethodology and genre analysis may enrich each other’s analysis of naturally occurring social situations. Keywords: video-analysis, ethnomethodology, communicative genres Giving Visibility to Hidden Rural Localities: Contributions of an Ethnographic Research Fernando Ilídio Ferreira, Universidade do Minho, Portugal., [email protected] A qualitative and ethnographic research developed in the ambit of a PhD at the University of Minho, Portugal, is presented in this paper. It was developed in a rural municipality and the part which is presented corresponds to a study of a social solidarity institution called OUSAM. Founded in 1985, as a sequence of a multidisciplinary project developed by the local health centre, OUSAM has been carrying out activities involving children, families and communities in isolated parishes by a system of bussing whereby children are picked up from their homes and driven to five small pre-school centres and later returned home. From its origin, the orientation for public space of the “community” has been the principal characteristic of the work of this institution, but the members of the direction and the professionals consider that since the 90s the pre-school teachers’ orientation for the restricted space of “the classroom” is increasing. The ethnographic research confirms that phenomenon but also reveals that factors still exist which make OUSAM very peculiar. In fact, one of its most distinctive characteristics is the work that it has developed in the interior of the small and dispersed communities. The bussing corresponds to the first moment of the education process of children and the possibility of daily contact between pre-school teachers and families and other members of the communities. Besides that, 201 the fact that pre-school teachers circulate daily throughout the isolated parishes, has given them a better knowledge of the social and family environment of the children of these areas. Given the fact that the movement of the mini-buses, which transport children, pre-school teachers, auxiliary staff and other social workers, are almost unique, these interior localities have become less “invisible”. In fact, the professionals carry out the role of mediators with other local entities, such as social security, local authorities, a health centre and a local project whose objective is to fight against poverty and social exclusion, in order to contribute to the resolution of identified problems. Therefore, what was at first considered an obstacle and which is still today seen by some people as a precarious solution, reveals potentials which over-pass the domain of pre-school education. It is a broader process of social and community work of mediation which is essential in rural areas confronted with problems of de-population and isolation. Keywords: qualitative and ethnographic methodology, community and mediation work, actor’s perspective, rural areas The Emergence of New Identities in Argentine Armed Forces: A Life Course Research. Alejandra Navarro, University of Buenos Aires, Argentina, [email protected] The objective of this paper is to discuss the theoretical and methodological decisions undertaken in the study “Pensamiento y Accionar Militar. Los Carapintadas en la Argentina: 1987-1990" ["Thought and Military Drive: the Carapintadas in Argentina: 1987-1990"] emphasising the advantage of a qualitative approach instead of a quantitative one. This study reconstructs four military uprisings that took place in Argentina between 1987 and 1990, known as "levantamientos Carapintadas” [Carapintadas' uprisings]. The paper’s objective is to reconstruct those events from the Carapintadas’ point of view (lower and middle ranking officials), highlighting their evaluations and emotions, as well as their ideological speech. The frame of this study is people's narrative which is permeated by their attitudes, values and beliefs. In this sense it contains not only a description of events but a selection and evaluation of their lived reality. There are events in people's life that leave traces and scars which permanently stay in their memory. Sometimes they could become turning points through which these people reinterpret their past and future experiences. Memory is an individual property but it is strongly influenced by socio cultural contexts. Hence, activities like memory and forgetfulness must be considered basically as social processes. Sometimes it is possible to find not only dissent between what different people remember about the same event but also faulty but valid reconstruction of a past episodes. What is important is to know the actors’ perspective in the recalling of events that took place in a certain way at a certain time. The studies which dive into peoples' narrative memories and emotions are usually shaped by qualitative approaches which recover their life course. The paper I am presenting to this meeting analyses each of the theoretical and methodological decisions that were implemented in our study, emphasising the validity of the information. Data come from in depth interviews with the participants in the four military uprisings and from documents written by them. Keywords: identity, actor’s perspective, military institutions/identity, life-course narratives, Argentina People’s Perceptions on the Construction of Collective Goals: Interviews in Street Meetings of Collective Protest Luciano Brom, Pablo Dalle and Rodolfo Elbert, University of Buenos Aires, Argentina, [email protected], [email protected], [email protected] The present paper has two objectives, one of substantive order and other of methodological nature. In first place, the article delves into the opinions and beliefs of participants of street protest demonstrations regarding the impact of corruption on Argentine democracy and economic development. In second place, it discusses methodological issues related to the validity and reliability 202 of both the construction of the instrument, the fieldwork experience and the data analysis. Data comes from 196 survey interviews with open-ended questions conducted among participants of demonstrations and popular assemblies that took place in the Metropolitan Area of Buenos Aires during April/May of 2002. Taking in account that the survey was implemented during moving scenarios such as that of demonstrations, the paper draws on both the cognitive and emotional elements present in the respondents’ discourses. This is coherent with Raymond Williams’ definition of ideology as the characteristic world-view or general perspective of a class or other social groups including formal and conscious beliefs but also less conscious, less formulated attitudes, habits and feelings, or even unconscious assumptions, bearings and commitments shared by the individuals of these groups. This paper also discusses the methodological issues related to the validity and reliability problems implied in the different stages of the study, from the construction of the selected instrument to the data analysis. The paper describes the limitations and benefits of implementing a questionnaire with openended questions among demonstrators and the particularities of a fieldwork experience in a noisy and crowded environment. Finally, it addresses the benefits of implementing a strategy of data analysis which combines both a qualitative and quantitative approach. In first place, the thematic analysis of the data allowed us to analyze the respondents´ opinions and emotions. The emergent categories led us to construct a typology that resumes the respondent’s ideological orientations towards the actual political and economic system. This typology was a key tool in further statistical analysis. Keywords: democracy, ideology, qualitative/quantitative analysis Images of the Argentine Ruling Class: Focus Groups using Mass Media Photographs as Stimulus Valeria Maidana, Gabriela Medin and María Alejandra Otamendi, University of Buenos Aires, Argentina, [email protected], [email protected], [email protected] The objective of this paper is to discuss methodological aspects of the study “Images of the Argentine ruling class: focus groups using mass media photographs as stimulus”. The study compares middle class young people’s opinions about social and political ruling class leaders in two different historical moments: June 2001 and June 2003. While June 2001 may be considered representative of the argentine political crisis, June 2003 represents a relative stability after the crisis. People develop their opinions, perceptions and attitudes in interaction with significant others in conversation encounters or in processes of self-reflection. Therefore, as a focus group consists of group’s open discussions we used this method to produce data about participant’s social representations, knowledge, beliefs and perceptions about social and political leaders. We utilized photographs of leaders, which appeared in mass media magazines, as a stimulus for discussion. From the analysis of the group’s discussions we constructed the following categories: appearance and physical characteristics; moral characteristics; and attributed attitudes and behaviors. Comparing data collected in two different historical moments, we found that the leaders in the photographs were judged consistently in a similar way. There are certain elements that remain stable: the categories used for describing the actor’s social membership and morality appear as the core of the social representation. From a methodological point of view, it is important to point out that focus groups have certain disadvantages such as introducing a bias when one of the participants speak too much or has a dominant attitude that could make other members to feel uncomfortable or not have an active participation. In this presentation we are going to emphasize the moderator importance to stimulate participation among everybody, the convenience of a careful selection of the participants and the influence of photographs as a stimulus for discussion. Keywords: qualitative methodology, focus group, reflexivity collective identity, social and political leaders, actor’s perspective. 203 Evaluation of the Internet Use: A Call for Mixing Method Approach Boris Kragelj, University of Ljubljana, Slovenia, [email protected] In the paper we try to present a special approach for the evaluation of the use of the Internet that is based on mixing qualitative and quantitative methods. The approach originates from the evaluation research tradition, applies classical evaluation criteria from the public policy analysis and systematically combines various traditional methods of social science with new methods of Internet research. Using that kind of methodological approach for the evaluation of Internet use results in more relevant, complete and broader assessment of Internet use by certain organization as it takes into account its specific communication goals and considers broader social context of their realization. This new approach is universal, so that the Internet use of different organizations could be assessed in uniform - standardized way, and at the same time particular, so that specific goals or different communication strategies of Internet use by each of the particular organization could be considered as the leading criteria in each case of the evaluation. “Evaluation study” by definition is systematic use of procedures, methodologies, and approaches of social science research for assessing ideas, plans, projects, results and effects of public or political programs considering broad social and political context and including all involved parties (Rossi and Freeman, 1993). We can understand evaluation study as methodology for reviewing activity (program monitoring) or assessment of success of some kind of program (program evaluation). The main purpose of evaluation study is not only to assess how effective are specific programs in achieving its specific goals, but also to find some better solutions of reaching it. Originating from the theoretical approach of evaluation study, we are trying to transfer its basic ideas for the purpose of evaluation the use of the Internet, where the program under evaluation are goals or communication strategy of certain organization in using the Internet. We argue that this kind of evaluation study is possible only with systematic design of mixing qualitative (public policy analysis) and quantitative (social science and internet research) methods and in this way suggest the following three-step triangulation model: Identification of organization’s goals or communication strategy in using the Internet Selection of relevant theoretical (public policy) evaluation criteria that captures the effectiveness of organization’s planned Internet use Complete empirical analysis of organization’s website (measurement of its performance in all aspects related to its goals and strategy defined in step 1: e.g., accessibility, usability…) corresponding to the selected evaluation criteria in step 2 Comparison of goals (from step 1) with results of analysis (step 3) according to selected evaluation criteria (step 3) We demonstrate this methodological approach on a case study of Internet use by the Government of the Republic of Slovenia. Keywords: mixing methods, evaluation study, public policy analysis, Internet research methods SESSION: INTERROGATING CONVENTIONALITY IN QUALITATIVE RESEARCH. BRICOLAGE AND INGENUITY IN INTERDISCIPLINARY FIELDWORK Organizor: Marsha Henry Contact: Centre for Health and Social Care, School for Policy Studies, University of Bristol, 8 Priory Road, Bristol BS8 1TZ, United Kingdom, Tel: +44 (0)117 954 6755, Fax: +44(0)117 954 6756, [email protected] Keywords: feminist methodology, authenticity, time, insider/outsider, reflexivity, accountability Abstract The panel will consist of 5 papers on various qualitative methodological issues raised by doing research from an interdisciplinary perspective and/or using multiple methods. The overall aim of this 204 panel is to explore how qualitative researchers cross conventional methodological and disciplinary boundaries in their research and how they conduct fieldwork in ways that go against the methodological 'grain'. Furthermore, the papers will explore how researchers are able to creatively respond to dilemmas arising in the field by adapting old or inventing new methodological techniques. Tom Slater's paper challenges the use of tape recorders for in-depth interviewing. Paul Higate's paper argues that gender and military experience play equally important roles in the experiences of military sociologists. Suruchi Thapar-Bjorkert's paper examines the difficulties raised by using historical and sociological methods across the colonial and postcolonial period. Finally, Marsha Henry's paper explores problematic issues of representation and authority raised for diasporic researchers. The Tape Recorder as a Major Player in the In-Depth Interview Tom Slater, Centre for Urban Studies, School for Policy Studies, University of Bristol, UK This paper offers an account of the problems encountered when using a tape recorder to capture the words of people interviewed for a research project on gentrification in inner-city Toronto, and how the problems were resolved. Numerous textbooks, fieldwork guides, and evidence-based practitioners in several sub-disciplines of social science treat the tape recorder as a sacrosanct accessory - without it, interview quotations are assumed to be invalid and inauthentic, and the researcher somehow casual and irresponsible. This paper argues that jettisoning the intrusive whirr of tape spool can have a very positive impact on qualitative work depending on the research context, and the researcher's experience as an interviewer. It is pointed out that, first, an engaging interview can only take place when both interviewee and interviewer are at ease, for they influence each other, back and forth, and second, treating an interview as an informal conversation, to be documented immediately afterwards and later checked with the respondent, can be just as effective and informative as capturing everything that is said on tape. While I do not advocate this as a general, normative strategy, I point out that qualms over authenticity of accounts and 'reliable data' are secondary to the needs of an inexperienced researcher trying to come to terms with the tremendous challenges posed by qualitative investigations in tense inner-city locations. If the Shoe Fits: Authority, Authenticity and Agency in Feminist Diasporic Research Marsha Henry, Centre for Health and Social Care, School for Policy Studies, University of Bristol, UK Many feminists have been critical of the inequalities that arise between feminist researchers and their participants, especially so in the case of research on third world women, and proposed some ways to minimise exploitation such as using participatory methods. However, one area that has been relatively neglected is how feminist researchers are actually trained and prepared for fieldwork. As an IndoCanadian, I was trained for my doctoral fieldwork in India by white, middle-class, English academics. My supervisors did not assume that I was, in Narayan's terms, an 'Authentic Insider', merely because I was part of the South Asian Diaspora. Instead, they trained and prepared me based on their own methodological 'traditions' and histories. What was problematic, I realised after completing my fieldwork, was that I had fashioned some aspects of my researcher identity based on my supervisors as model researchers. This was particularly complicating in relation to issues of authority, authenticity and agency. In this paper, I argue that the centrality of the white, western, middle-class academic as a dominant model of feminist researcher poses considerable difficulties for researchers who are racially and ethnically marginalised within their 'home' communities. The 'authority' that I was expected to diminish as a western researcher, seemed surprisingly absent in many of my research encounters. My 'authenticity' as a westerner or non-westerner was challenged both at 'home' and in the 'field' making hyrbidity and hyphenated identity a compelling option. My abilities to exercise agency in the research process were complicated by my attempts to fill the shoes of a feminist researcher that did not reflect my different social and institutional positioning. I argue that diasporic researchers might be better prepared for research by acknowledging the ways in which they are academically positioned both at the centre and at the margins, at 'home' and in the 'field'. 205 Colonial Contexts and Post-Colonial Understandings: Researching on Women in the Indian Nationalist Movement Suruchi Thapar-Bjorkert, Sociology, University of Bristol, UK The project of 'nation building' during the anti-colonial struggle in India has primarily been associated with the public sphere, which is also conceptualised as the sphere of all 'political' activities. I was interested in the 'domestic' narratives of 'ordinary' middle class women (like my mother and grandmother) who could not negotiate public roles for themselves as easily as elite women. This methodological journey would enable me to deconstruct unexamined assumptions about women and their domestic activities in the Indian nationalist movement. My own emphasis on personal narratives would raise both methodological and theoretical dilemmas. In particular, I came to question western feminist theoretical models and their relevance for understanding non-western 'colonial' experiences. Women's autobiographical narratives were to expose the interconnections of the public and private: their active negotiations within the nationalist movement, the processes through which public activities regulated private spaces and finally the politicisation of the home/family through diffusion of the public/private boundaries. I was able to demonstrate that the domestic was an equally important site of nationalist activities as the public sphere and saw two transformations - the domestication of the public sphere and the politicisation of the domestic sphere. Insider/Outsider Debates: Reflexivity in Military Research Paul Higate, Centre for Health and Social Care, School for Policy Studies, University of Bristol, UK Research paradigms within the sub-discipline of the sociology of the miliary continue to be influenced by positivist approaches. These derive from organisational and psychological studies and tend to be framed by more traditional 'objective' perspectives. Rarely is the role of the researcher and his or her impact on the ensuing research enterprise invoked; subjectivities in these contexts remain unexplored. In this paper, using the concept of reflexivity, we consider the ways in which the biographies of the research team shaped engagement with ex-military personnel and their partners. Here, the 'insider' - a former member of the Royal Air Force, and the 'outsider' - a life-long civilian with earlier affiliations with the Campaign for Nucelar Disarmament, discuss how their biographies influence not only thinking about their role in the research project, but also how it shaped experience of interview amongst service leavers and their families. Through the lens of reflexivity, we show how our approach to a particular research project, together with the findings we produce, are creations brokered intersubjectively at the interface of participant and researcher biographies. SESSION: MISCELLANEOUS Organizor: Henk Kleijer Contact: SISWO, Plantage Muidergracht 4, 1018 TV, Amsterdam, The Netherlands, [email protected] Funding of Social Research in Pakistan S. Hamidullah, Government Postgraduate College, Mardan, NWFP, Pakistan, Tel.(Residential) 92 923 612913, Fax.(Residential) 92 923 611098, [email protected] In Pakistan all the students of various universities are supposed to conduct an empirical research on the methodology they are taught during their graduate and post-graduate course and submit the report to their faculty of concerned department for the partial fulfillment of their degrees. The methodology and quality of research in no way lags behind the course taught at western universities. However the students use their own methodology when they conduct their researches because of the facts that no financial help or sponsorship is extended for such activities. Consequently when the 206 students get their degrees in social research, they divorce their skill and do not touch such “rubbish” activities. Similarly the teachers of social sciences have no interest for the research because there is no funding facilities or the incentives / encouragement, therefore they are confined to teaching only and do not innovate their skill of research. Ultimately the research is confined to personal and private activities with out any appreciation. The present study discloses how the students collect the primary and or secondary data available in the various corners of the universe. Do they go from door to door, from individual to individual and interview / probe them all to over come biased data? If the students cannot afford the field survey through interview schedules or questionnaires, do they do the TGD (Targeted Group Discussion) methods for their researches? Or they use self-created methods for the compilation of their dissertation acceptable for the post-graduate degrees. The present study enlightens how the research activities are carried out and presented in various conferences of national and international forums by the researcher / scholars who are addicted to research writing and given no incentives in the globe. The report also depicts their constrains and expectations. Children as Respondents: Developing Questionnaires for Children Edith de Leeuw, Natacha Borgers and Astrid Smits, Department of Methodology and Statistics, Utrecht University, The Netherlands, [email protected] Children are no longer neglected as respondents in surveys. They participate more and more in surveys. However, systematic methodological knowledge on survey techniques and questionnaire development for children is scarce, and researchers have to rely on ad-hoc knowledge from diverse fields as child psychiatry and educational testing or on methodological knowledge on how to survey adults. The purpose of this paper is (1) to integrate the current theoretical and empirical knowledge regarding questionnaire research with children as respondents, and (2) to present instruments and strategies for the development of questionnaires for children. Keywords: data quality, cognitive development, social development, youth questionnaires Feminist Method and Methodology and the Struggle to define Social Inquiry and Epistemology. The Logic behind the Measure. Ronald Marshall, The University of the West Indies, St. Augustine Campus, Trinidad en Tobago, [email protected] Critical thinking according to Harding (1987) must be no less than what is traditionally defined as relevant evidence. This paper critically reviews the question of method, methodology and epistemology in feminist research and attempts to demonstrate that feminist research is at crosspurposes. The question of qualitative research, “bounded generalizability” and feminist methodology represent attempts by feminists of various suasion to come to grips with the question of logic and rationale, even as they challenge traditional researchers on biases in their research. In so doing, they unwittingly create systems that are not logical and force people to reconceptualize the meaning behind research. In comparing the two methodological systems, I come away with the impression that social science research and sociology are at the crossroads. These impending changes have implications for the emergence of a new method, methodology and epistemology and the ability of sociology in particular to be able to continue to undertake research in order to uncover various complex phenomena. 207 Disambiguating Culture: The Cultural Element of Survey Response in Perspective Eleanor Gerber, US Census Bureau, Statistical Research Division, USA, [email protected] Interest in examining surveys in a cross-cultural perspective has risen as surveys have been fielded in complex, multi-national and multi-ethnic contexts. Survey response and respondent behavior in general is understood to vary by cultural, ethnic and linguistic factors. The aim of this paper is to examine some key concepts by which these behavioral differences may be understood. An understanding of these factors is useful both in predicting where differences may occur and in guiding future research. The data on which this paper are based are drawn from qualitative work over more than a decade surrounding the US decennial census. For this discussion, culture is regarded as a set of mental constructs; definitions, assumptions, values and beliefs by which people construct and interpret the world. It is the central insight of anthropology that these beliefs differ by culture. However, this single conceptual insight does not do justice to the complexity in respondent behavior to surveys nor to the range of differences to which survey designers should attend. These factors include: 1. Conceptual differences: An example is the race categories in the US census, which are problematic for persons from Latin America, who often do not find acceptable categories for themselves. 2. Social adaptation: Some relevant differences are not codified in belief, but represent common structural adaptations to external circumstances. Thus, immigrant households from many different systems of belief adapt to high rents by sharing space, regardless of varied ideals of household structure. In addition, wide similarities in attitude towards government data collection are shared by diverse groups with similar histories of conflict with the Federal Government. These social adaptations can create wide similarities in behavior among very diverse groups. 3. Assimilation: Despite very different beliefs, it often does not take immigrants long to learn what they consider to be the expected response. Thus, although “Hispanic” is an unfamiliar concept, many immigrants adopt it as a self-identification in the US context. 4. Different understandings of respondent behavior. The survey itself is a cultural representation. Respondents have a set of cultural expectations about what surveys are and how to participate, which vary by previous experience. These factors affect survey design and procedures, and should be the subject of further research. Keywords: Cross-cultural methodology, multi-cultural survey design, nonresponse and survey errors SESSION: MIXING QUALITATIVE AND QUANTITATIVE METHODS: EMPIRICAL DESIGNS AND THEIR IMPLICATIONS Organizor: Udo Kelle Contact: Vechta, Germany, [email protected] Abstract Despite ongoing „paradigm wars“ there is a growing tendency to combine qualitative and quantitative methods of social Research to „Mixed Methods“ Designs. At present „Mixed Methods“ represents a rapidly developing field of social science methodology. However, the discussion is still characterized by a lack of common ground. A variety of proposals have been made for a taxonomy of Mixed Methods Designs, but there is yet a lack of agreement regarding basic concepts and definitions. One reason for this is that the relation between empirical designs and theoretical concepts has been rarely discussed so far, although many authors in the Mixed Methods debate stress the preeminence of the research question over abstract methodological and epistemological considerations. In this session we will discuss the relation between sorts of mixed methods designs on the one hand and types of research questions and theoretical rationales of research projects on the other hand. 208 Researchers who want to develop and discuss ideas about the relation between designs, research questions and theory esp. on the basis of the results and experiences of empirical projects are invited to submit proposals for presentations. Part I: Mixed Methods Designs and their Relation to Theoretical and Methodological Debates Qualitative and Quantitative Methods between Paradigm and Pragmatics: Mixing, Integration or Triangulation? Uwe Flick, Alice Salomon University of Applied Sciences, Berlin, Germany, [email protected] The relation of qualitative and quantitative research has been discussed from different angles, ranging from the so-called paradigm wars to consequently pragmatic combination and use of both. Mixed Methodologies and integration of qualitative and quantitative research are two key topics and two slogans that are very attractive both in the methodological discussion and in research practice. Both of these attempts to end the qualitative-quantitative debate still have to face several key questions: How far do they operate on theoretical and reflexive grounds? Do these grounds allow to frame the methodological relation of qualitative and quantitative research as well as the methodological training for both types of methods and to lead to a theoretically and empirically sound social research? In this paper, the concept of triangulation shall be taken up (again) to set out a framework for a reflexive and pragmatic use of qualitative and quantitative research. In spelling out the concept of triangulation a bit more than the way it is used in both debates may contribute to develop adequate research designs that comprise qualitative and quantitative research. A main point in this conceptualisation will be how far both forms of research are taken serious in their epistemological, methodological and theoretical backgrounds and research principles and are not seen in a premature dominance of one form over the other. Keywords: Triangulation; Qualitative and Quantitative Methods; Mixed Methods; Research Design Mixed Methods Sampling Considerations and Designs in Social Science Research Anthony J. Onwuegbuzie1 and Kathleen M. T. Collins2 1 Department of Educational Measurement and Research, College of Education, University of South Florida, 4202 East Fowler Avenue, EDU 162, Tampa, FL, 33620-7750, USA, [email protected], Tel: 813-974-1163 (Office), Tel: 813-632-9226 (Home), Fax: 813-6329227 2 College of Education and Health Professions, Dept. of Curriculum and Instruction, 201 Graduate, Education Building, University of Arkansas, Fayetteville, AR 72701, USA, [email protected]; [email protected], Tel: 479 575-4209 (Office), Tel: 479-527-6955 (Home), Fax: 479 575-6676 Because every study, whether quantitative or qualitative, is time-and context-bound, sampling decisions are made, whether consciously or subconsciously, by the researcher. These sampling decisions are made at the data collection (e.g., choice of sample, units, instrument), data analysis (e.g., choice and amount of data types of data units [e.g., numbers, words, observations], and data interpretation (e.g., choice and amount of data units to interpret) stages of the research process. Central to all sampling designs in both quantitative and qualitative research are issues of sampling cases and data units because some type of generalization nearly always is made. In particular, either numbers/words/observations from the sample are generalized to the population. In other cases, the sample of words or observations from the participant is generalized to the participant’s population of words or observations in order to capture the voice. In the context of mixed methods research, 209 sampling considerations are even more complex because sampling decisions must be made for both the quantitative and qualitative components of the study. Therefore, this paper provides a framework for developing sampling designs in mixed methods research. First, we discuss the role of sampling in both quantitative and qualitative research. Second, we discuss sampling tenets common to both quantitative and qualitative studies, including the following: (a) sampling decisions should stem from the research goal (e.g., predict, understand complex phenomena), research objective (e.g., exploratory vs. confirmatory), research purpose, and research question; (b) the sample of cases and/or units must be adequate such that the underlying characteristic, experience, and/or process is captured; (c) the sample of cases and/or units must be compatible with the research design; (d) inferences should stem directly from the sample of units that is extracted; and (e) the sampling strategy should be viable, efficient, practical, and ethical. Third, we provide a typology of sampling designs in mixed methods research. Here, parallel, concurrent, and sequential mixed methods sampling designs are distinguished. Also, we outline 24 sampling schemes. Fourth, sample size issues pertaining to cases/units in the quantitative and qualitative components are discussed. Finally, sampling design considerations are discussed in terms of the triple crises in mixed methods (i.e., research of representation, legitimation, and praxis.) We contend that carefully reflecting about sampling design represents an appropriate way of achieving verhesten in social science research. Keywords: Sampling, Sample Size, Mixed Methods Why it is Necessary to Combine Qualitative and Quantitative Data and Methods to Examine the Thesis of Individualization? Jens Zinn, School of Social Policy, Sociology and Social Research, Cornwallis Building, University of Kent at Canterbury, Canterbury, Kent CT2 7NF, UK, Tel: +44 (0)1227/82 4165, [email protected] One of the main concepts of the theory of reflexive modernization is the individualization theorem. The releasing of individuals from traditional bonds, the pluralization of options for action, the burden of being forced to make choices and not simply having the choice of doing so are at the centre of the hypothesis of a modernity undergoing modernization, as Ulrich Beck puts it. The proof of increasing individualization is however controversial. Previous attempts of its empirical verification or rejection may have lead to the impression that the operationalizations lacking depth were deliberate attempts to prompt confirmation or rejection of the thesis of individualization. If this thesis is however not to remain a blurred catchword but rather to be used as a theoretical proposition which should serve the purpose of scientifically verifiable theory, it must be formulated more precisely and operationalized. This paper aims to make a contribution in this effort. My argument presupposes that neither the observation of simple pluralization nor alone the change in the semantics of social or subjective individual description suffice to confirm or reject the thesis of individualization. Much rather the practical separation in research between the action outcome (e.g. social structure analysis) and meanings of action (e.g. biography research) in the discourse on individualization must be overcome if based on scientific analysis a society-wide process of individualization is to be said to have taken place. Individualization can only be understood and tested in the sense of a ”new mode of societalization, … a kind of ‘metamorphosis’ or ‘categorial shift’ in the relation between the individual and society” (Beck 1992, 127). This can be made possible, I argue, by using strategies of constructing types provided that not only identities, lifeplans or patterns of action but rather biographical actions influenced by contextual experiences from a longitudinal perspective are typified. Keywords: individualization, mixed methods, biography research, typology, longitudinal Research, triangulation Reference Beck, Ulrich (1992). Risk Society. Towards a New Modernity. London: Sage Publications. (German: 1986: Risikogesellschaft. Auf dem Weg in eine andere Moderne. Frankfurt a.M.: Suhrkamp). 210 Stratification and Social Mechanisms: An Empirical Point of View Flavio Ceravolo and Laura Accornero, Dipartimento di Ricerca Sociale (Department of Social Research) Università del Piemonte Orientale, Via Cavour, 84, 15100 Alessandria, Italy, Tel: +39 0131 283 727, Fax: +39 0131 283 704, [email protected], [email protected]> The paper focuses on the relation between the perspective of social mechanisms on one hand, and the controversy between qualitative and quantitative social research. In this regard we present an empirical study about the social reproduction mechanisms in an Italian local context (Alessandria) performed interviewing a sample of people homogeneous by high educative qualification (degree). This study presents a mix of qualitative and quantitative methods: a survey of 600 people together with some unstructured interviews. In this paper we will describe the principal results of this study and how the approach of social mechanisms can contributes to the “war between paradigms”. Using this empirical strategy we have pointed out four main reproduction patterns between generations: a “strong occupational” reproduction from the high position of the father’s job to the son’s collocation on the labour market; a “group reproduction” pattern from family position from daughter, an acquisition pattern based on the reach a high educative qualification starting from a working class family, and finally a double exclusion mechanism that can be ascribed the results of a cumulation of inequalities axis (gender, class…etc). The conjunction of qualitative an quantitative approach give us the possibility to identified the patterns described and evaluate their incidence in the local society. On the other hand we must point out that empirical studies have usually been focused on the construction of models that describe the relation between factors only at a macro-sociological level, thus building very complex – but also over-general – causal models. A valid alternative is to identify empirical patterns based on micro regularities defined by theoretical assumptions, which characterize homogeneous populations. In particular, this approach takes into account either structural conditions (macro) or individual choices and actions (micro) [Hedström e Swedberg, 1996, 1998]. Our aim is to map a plurality of different social mechanisms (or micro-causal patterns) operating in different social circles or spheres. Explanation in terms of mechanisms, however, allows us to develop the embedded dimension of individual choice and action. Social actors, in fact, make their choices on the basis of constraints, opportunities, beliefs and desires (interests) that are developed not only by the primary socialization process, but also by the social circles in which they are involved. On an empirical ground, we have to consider the possibility to specify assumptions and hypotheses about the different mechanisms operating in homogeneous social circles. This methodological concept was proposed by Lazarsfeld, who referred to “specification”. Coherency Analysis for an Updating of Sociological Thought Mascia Ferri, Department of Sociology and Communication, University of Rome "La Sapienza", Italy, [email protected] The study of the authors who have made the history of Sociology, has created definitions and interpretations that are often in contrast with one another. Therefore, the issue is one of understanding the reasons for this incongruence in thought - for this we are indebted to Sociology of Knowledge and to activate a historical reflection to recompose the pieces of the History of Sociological Thought. To this end, we needed a methodology to act as a bridge between historical reflection and the Sociology of Knowledge. Coherency Analysis has as its purpose identifying interpretive instruments available to the author under scrutiny, and to verify if these same instruments were used by the author in the research. A researcher must create an interpretive scheme/model organized in such a manner as to allow a comparison of the author's instruments, the instruments of the social-historical era where the author lived and the researcher's instruments. A study of coherency, should reveal the incongruence between the author's model and the researcher's, the instruments available and those actually used, the ideological positions and the viable positions. 211 No concept can be taken in isolation from the hypothesis, that it is stereotype of femininity, as is shown by Rousseau's work, be severed from the concept of social control, to illustrate an ideological position that is more conservative than progressive. The analysis of the dimensions of a concept reflects a logical, semantic and pragmatic process, for this reason its indexes cannot be created by a simple mathematical calculation: a researcher's reflective intervention is necessary. The researcher must distinguish a relevant conceptual combination from an arithmetically based one. However, the researcher's work introduces partiality by creating the data and in directing the observation. Therefore, out of necessity the researcher must works in a state of surrender to where, as indicated by Kurt H. Wolff, parenthesis are placed around the notions received with respect to a particular social reality that one is trying to understand, without resistance created by pre-constituted interpretive models: one's own point of view, prejudices, and stereotypes must be suspended. A study through this method does not mean renouncing one's critical role as a sociologist, but rather better defines one's ability to present a proposition over one's interpretive abilities. Coherency Analysis could lend consistency to previous theories by updating rather than revising them. Keywords: Coherency Analysis, stereotype, interpretive instruments Part II: Methodological Problems and Theoretically Informed Solutions in Mixed Methods Research Danger: Data Overload! Mixed Methods in a Cross-National Study of Care Home Staff Ingrid Eyers, University of Surrey, England, [email protected] This paper presents the experience of a mixed methods approach utilising structured interviews, selfcompletion questionnaires and semi-structured interviews using vignettes and a lifecourse grid to research care home staff in England and Germany. The research aimed to establish how government policies impacted on the acquisition of care skills and the daily provision of care. The theoretical framework related to emotional labour and institutional time in the provision of care and how this impacted on the quality of life experienced by older people living in English and German care homes. Data was collected in four English and four German care homes. The structured interviews with eight care home managers established a profile of the older people living in the care home and the staff employed to provide their care. Self-completion questionnaires were distributed to all care staff in the sample homes to establish a profile of how the respondents gained the skills to work in a care home. This was followed by interviews conducted with qualified staff and care assistants in each of the sample homes. The resulting 50 interviews focused on establishing a picture of the daily routine in the care home involved the use of care related vignettes in order to compare the provision of care by qualified staff and care assistants in England and Germany. The interviews concluded with the collection of lifecourse data in grid form. There was an acute danger of data overload in this blend of research methods which resulted in a wealth of inter-linked empirical data sets. However, the mixed methods provided a balanced, coherent insight into the impact of government policies on the care of older people. Furthermore the findings have theoretical implications on emotional labour and the sociology of time. The German Research Foundation Project “Mobilzeit” as Example for the Integration of Quantitative and Qualitative Methods Andrea Tina Booh and Siegrid Wieczorek, Justus-Liebig-University Gießen, Germany, [email protected], [email protected] Our study is placed in the field of labour research especially on the topic of working-times and worklife balance. Its design contains “between methods triangulation”. An important aim of our study is the integration of quantitative and qualitative methods. So we focus on the term of “between methods triangulation” as a central concept for method integration. 212 The design of our study includes a survey of the working population in Germany. Participants of our quantitative study are 1677 people in paid work and people who are not employed at the moment but with the intention to find paid work in future. The sample includes 55,9 % (937) women and 44,1 % (740) men at the age of sixteen to sixty. Data were collected from April to May 2003 in standardized telephone interviews. Our design implies a connected sample from the representative survey. It consists out of 40 persons who were asked in face-to-face interviews about topics of the standardized questionnaire as well as other topics like work-life balance. We chose this procedure to combine qualitative and quantitative data of the individual cases and to analyze them simultaneously. The presentation includes the type of multi-method design and the combined results of the representative sample together with the qualitative interviews of the connected sample. Keywords: Between methods triangulation, quantitative methods, qualitative methods, mixed methods, working-time, work-life balance. Do Media Discourses Influence Opinions? Connecting Discourse Analysis of Newspaper Data with Time Series Analysis of Survey Data Nina Baur1 and Christian Lahusen2 1 Methoden der empirischen Sozialforschung, Otto-Friedrich-Universität Bamberg, Feldkirchenstr. 21, 96045 Bamberg, Germany, [email protected] 2 Lehrstuhl für Soziologie II, Otto-Friedrich-Universität Bamberg, Lichtenhaidenstr. 11, 96052 Bamberg, Germany, [email protected] Using a current research project, we want to discuss problems and solutions of linking qualitative and quantitative longitudinal data. The project aims at linking debates on long-term media effects and the acceptance of welfare states. The main question is: Does acceptance of welfare states change over time as a result of debates on social security in the mass media? In order to be able to operationalize this question, we focus on acceptance of state programmes for preventing and reducing unemployment and combating poverty in Germany. We used a three-stage-process in order to answer our question: First, we analyzed newspaper articles using a combination of qualitative content analysis and discourse analysis. Second, we traced both individual and collective change in attitudes towards the welfare states by using different waves of the Socio-Economic Panel (SOEP) and of ALLBUS. Third, we used the “Media-Analysen” (data collected by ag.ma giving information on readers of German media) in order to construct reader profiles of the newspapers we analysed. Both qualitative and quantitative research are valuable and that – depending on the research question – researchers very often have to rely in a combination of both strategies, because no other data are available or because of the nature of the research question. Still, at least in our case, the gap between qualitative and quantitative data could not be closed without using theoretical hypotheses. In this sense, there are a good number of practical reasons for using Mixed Methods. However, our research grounds on the assumption that Mixed Methods is also indispensable in order to overcome the gap between interpretative and statistical analysis. In our case, we argue that the acceptance of the welfare state within the population can certainly be traced back by opinion polls. However if we want to understand the possible effects of public debates on individual attitudes, we have to overcome the limitations of standardized data and reach out to interpretative data. Undoubtedly, this broader approach increases the practical and methodological challenges of the research process. Our study indicates, however, that Mixed Methods open new insights into a broader empirical understanding. Keywords: Qualitative Research, Quantitative Research, Mixed Methods, Discourse Analysis, Survey Data, Time Series, Longitudinal Research, Validity 213 The Conceptual Net of Internal Consistency and the Es of Praxiology in Human Inquiry Arne Collen, Saybrook Graduate School and Research Center, 450 Pacific, San Francisco, CA 94133 USA, [email protected] At the heart of every research study of human beings is a set of core concepts that comprise a network of constructs, from which arises a meta construct termed internal consistency. This first conceptual system is complemented by a second comprised of another set of core concepts, known as the Es of praxiology. This second conceptual system imbues the first at each phase of human inquiry. Some relationships between the two systems are discussed to delineate the manner in which the second system serves to enable the first system to become manifest as research process. Keywords: Constructing research methodology, research process, validity. SESSION: ANALYZING INTERVIEWS (PART 1): QUALITATIVE ANALYSIS Part I: Systematic Qualitative Analysis of Interview Data Organizors: Manfred Max Bergman1 and Véronique Mottier2 Contact:1SIDOS Switzerland and Cambridge University, [email protected] 2 Université de Lausanne, Switzerland, [email protected] Keywords: Interviews, discourse analysis, narrative analysis. Abstract Interviews are one of the main research methods to assess social context and subjectivity. A substantial increase in the general popularity of qualitative interviewing as a cost-effective research tool on the one hand, and its proliferation in evaluation research and policy analysis on the other, has raised questions about the relevance and quality of interview-based research. Most quality concerns focus on data collection (e.g. question design, intersubjectivity, sampling, rapport), while neglecting the analytical process. This session will explore what happens to interview data after it has been collected and, wherever relevant, transcribed. Based on case studies and theory, session presenters will discuss what methods they use and which steps they take to arrive at substantive conclusions. Strategies for Analysing In-depth Interviews of People with Disabilities in Northers Aegean, Greece Dimitris Papageorgiou1, Theodoros Iosifides2, M. Sidiropoulou3 1 Performance Theory and Performance Practices in Everyday Life, Department of Cultural Technology and Communication, University of the Aegean, Sapfous and Arionos, 81100, Mitilini, Lesvos, Greece. Tel. 0030-22510 36617, [email protected] 2 Social Research Methods, Department of Geography, University of the Aegean, University Hill, 81100, Mitilini, Lesvos, Greece. Tel. 0030-22510 36405, [email protected] 3 Department of Physical Education and Sports Science, University of Thessaloniki, Laboratory of Developmental Pediatrics and Special Education, DPESS, University of Thessaloniki, Greece, Tel. 0030-2310 992215, 0030-210 9345825 The basic aim of this paper is to present certain methodological and analysis strategies of a series of in-depth qualitative interviews of people with disabilities in two islands of northern Aegean area (Lesvos and Hios), Greece. This qualitative research was part of a broader European research program (EQUAL) concerning various dimensions of social and economic life of people with disabilities. Through forty-four (44) in – depth interviews with people with disabilities a series of issues were 214 researched and more specifically the meaning of being a person with a disability, social networks and socialization, family and social life, economic and labor market participation and social integration problems and prospects. This paper concerns mainly the basic analysis strategies employed after the empirical material was collected and took a textual form. After a brief account of the qualitative research process (formation of research thematic areas, construction of the interview guide, training of the interviewees, selection of participants and sampling, problems of entering the field and establishing rapport, field relations and actual interviewing), special emphasis is given to the following analysis strategies: - Gradual construction and reconstruction of the coding framework of analysis and the transition from the interview guide to the coding scheme - Links between different codes and sub codes and the avoidance of decontextualization - The use, advantages and disadvantages of qualitative analysis software - The emergence of core categories of analysis and the role of the grounded theory approach - From categorization to theorizing and the relations between empirical data and theoretical – analytical notions - Problems of validity and reliability and the question of generalization of findings Finally the paper concludes by summarizing the major methodological and analysis challenges with an emphasis to the overall picture of the research experience. Keywords: qualitative data analysis, grounded theory, in-depth interviews, people with disabilities, coding, categorization theorizing The Interpretation of Qualitative Data from a Critical Theoretical Framework Claire Wagner1 and David JF Maree2 1 Department of Psychology, University of Pretoria, Lynnwood road, Brooklyn, 0181, RSA, [email protected] 2 Department of Psychology, University of Pretoria, Lynnwood road, Brooklyn, 0181, RSA, [email protected] This paper describes how data, generated from semi-structured interviews, was analysed. A brief background is provided of the aim of the research and as well as a description of how critical theory informed the methodological approach in the study. As there is no unified critical theory and much debate about appropriate methodologies within this theoretical framework abound in the literature, the researcher describes the path that was chosen to do a critical interpretation of the data. Furthermore, the paper informs the reader about how the text was prepared for analysis and the rationale behind this preparation. Elements of Carspecken’s (1996) approach on how to analyse data from a critical hermeneutic epistemology were adapted and applied to the interview material. A reconstructive analysis was performed on the data where codes were assigned to establish patterns and generate themes. This in turn enabled the researcher to choose suitable parts of the data for meaning reconstruction (fleshing out and explicitly stating what is said by respondents). Nine categories referred to as ‘beliefs’ were generated. The researcher then applied a hermeneutic approach to the categories exploring the ideas of virtual intersubjectivity, meaning-making through familiarity with the culture of the actors, reflecting on and identifying the researcher’s norms, the normative circle and the personal characteristics of research participants versus typical cultural behaviour. The paper concludes with some reflections on the methodological contributions and limitations of the study. Keywords: critical social theory, qualitative data analysis, critical hermeneutics, methodology 215 Contradictories in the Interview: Public and Personal Olga B. Savinskaya, The Institute of Sociology, the Russian Academy of Sciences, [email protected] In my presentation I would like to debate about the subsidiarity narrative and discursive analysis. Narrative analysis allows reconstructing the consequences in life stories; determine the cost-effective links, heroes, allies and enemies, un/favourable conditions and so on. But sometimes, the analyst is faced the situation, when fair frank interview gives, in the result, a lot of subjective contradictions. The attitudes to a topic of the interviewee are looking like discrete, and even contradictory. To join the history in whole could be used discursive analysis. Analyzing discursive contexts of each utterance pointed out, that history was told by several identities of an interviewee. And the questions are how the interviewee presents different identities in a talk, why she/he refers to each of them, how identities involve each other? On a personal level, these questions are more important for the analysis of interviews with persons that have marginal status. On the societal level, they become more acute in times of transitions then identities themselves are infixed and floating. I would like to demonstrate my reasoning on the one interview with a young woman working in a private firm on telecommunication service. The interview lasted about an hour. The subject of the talk was the work, working place, but concerns all aspects involved the work: home, family, child, combination of professional and parenting responsibilities, leisure, and so on. Despite consequently narrated work experience, the interviewee gave contradictive description of her attitudes to the work and the role of work for her life. Several discourses that were produced by different identities of the interviewee were marked out during discursive analysis. I conditionally put them in order from public to personal: some utterance expressed the common stereotypes, some of them reflect common opinion in the professional sphere, and some of them were the specific culture on the workplace. And the last part of expressions, that don’t match any named above, was marked me as unique character of the interviewee. “Public” discourses seemed strong: they had more reasonable argumentation, “private” were more emotionally, and they came up through the powered commonsense discourse. In that way, the discursive analysis shows how current discourse struggle in one story involves the narrative structure, and how present reconstruct past. Keywords: discursive analysis, origins of contradictive attitudes The Interview as a Text: Analyzing In-depth Interviews in Social Science with the Conceptual Framework of Linguistics Veronika Sieglin, Facultad de Trabajo Social, Universidad Autónoma de Nuevo León, Monterrey, Nuevo León, Mexico, [email protected], [email protected] The methodological reflections of this paper are part of a research project which intends to recreate the effects of governmental training programmes in the self image and self-worth of Mexican rural midwives. As much of this imaginary does not operate on the level of the conscious – not only what is said but how it is said is significant – I use linguistic analyses (e.g. rhetorics, semantics and narrative theory) in combination with social theory in order to make visible unconscious self-conceptions. This does not mean looking for meaning hidden behind the text but within the text itself. This methodological procedure expands the materials subject of analyses from the manifest to the latent contents of the texts. The paper - based on 42 in-depth-interviews – presents the steps taken in order to prepare the materials for analyses and, on the next level, to interpretation. The paper presents two ways of linguistic analyses: one which takes into account segments of different interviews in order to explore a determinate topic; the other focused on one and the same discourse fragment as source for case study. Keywords: self-concept, unconscious, linguistics, discourse analyses 216 Analysis of Interviews with Russian Entrepreneurs using Critical Discourse Analysis. Tatyana Kosyaeva, Department of Social Problems, IEIE, Russian Academy of Sciences, Novosibirsk, Russia and Economic Institute, Czech Academy of Sciences, Prague, Czech Republic, [email protected], [email protected] The contemporary field of discourse analysis is very diverse. I am going to discuss why and how one sub-field, Critical Discourse Analysis (CDA), is especially useful in making substantive conclusions from the raw interviews-in-depth with Russian entrepreneurs. It is especially useful for the following reasons. First, CDA is considered both theory and method by leading practioneers (van Dijk , Fairclough, Wodak). Second, CDA stresses the view of discourse as a moment of social practice. "The motivation for focusing on social practices is that it allows one to combine the perspective of structure and the perspective of action." Fairclough (2001, p.121). Third, CDA contains important tools of inquiry which mediate between the "micro" level of life stories and the "macro" level of social institutions. As an example, consider Fairclough's 3-dimensional model, Fairclough (1992, p.73). Theory and method of Critical Discourse Analysis contains tools that allow interpreting the Russian entrepreneurs' life stories, namely connections between biographies and change in system of the social institutions in Russia in transition period. In CDA change of language is always a change of social practices, and ".word is not used in terms of a definition, but rather against a set of social and cultural assumptions that constitute a cultural model," Gee (1999). Within this approach, emergence of new words in life stories indicates appearance of new social practices and of the corresponding cultural models. In-depth interviews with entrepreneurs have been conducted by the author in several regions of Russia since 1991. Usage of the chronological strategy - repeated interviewing of the same respondents-entrepreneurs revealed not only the dynamics of their identities, but also the changes in discourse about entrepreneurship. As a result of analysis of Russian entrepreneurs' "life stories", several stages of the process of transformation of system of social institutions were identified. Keywords: Critical Discourse Analysis, interview-in-depth, Russian entrepreneurs, life stories Analysis of Semi-Structured, In-depth Interview Data Kathleen W. Piercy, Department of Family, Consumer, and Human Development, Utah State University, Logan, Utah, USA In a semi-structured interview, the researcher asks a series of open-ended questions, with accompanying queries that probe for more detailed and contextual data. Respondents’ answers provide rich, in-depth information that helps us to understand the unique as well as shared circumstances in which they live, and meanings attributed to their experiences. When several research questions need answering, the semi-structured, long interview is an ideal way to collect data. Such an approach produces a considerable amount of textual data for analysis. As a social scientist, the analytic approach that I use is the 5-step method of McCracken (1988) for long interviews. The first step consists of reading and reviewing each interview transcript twice; the first time, for understanding of its content; the second time, for identification of useful comments noted as observations. In the second stage of analysis, observations are developed into preliminary descriptive and interpretive categories based on evidence presented in the transcripts, one’s literature review, and the theory or conceptual framework used to guide the research. The third stage consists of thorough examination of these preliminary codes in order to identify connections and develop pattern codes. The fourth stage of analysis involves a determination of basic themes by examining clusters of comments made by respondents and memos made by the researchers. A search for inter-theme consistency and contradiction is conducted (McCracken, 1988). The final stage examines themes from all interviews within each case, as well as across all such groupings, to delineate predominant themes contained in the data. These predominant themes then serve as answers to the research questions, and form the basis for writing up the data. Other aids to data analysis are described in this paper, including analytic memos, contact summary forms (Miles & Huberman, 1994), the researcher’s use of self through a reflexive journal, and 217 computer software (QSR NUD*IST). Computer software is an invaluable tool for management and analysis of large amounts of qualitative data, and can be used in all stages of analysis. Reports generated by this software are used to improve coding accuracy, discover new patterns and themes, and as a tool for discussing evolving coding schemes and interpretations of the data. Several examples of these analytic strategies and the ways in which they were used in the author’s research projects are offered throughout the paper. Keywords: semi-structured interviews, stages of data analysis, memos, computer software aids, use of self, reflexivity References McCracken, G. (1988). The long interview. Newbury Park, CA: Sage Publication. Miles, M.B., & Huberman, A.M. (1994). Qualitative data analysis. Thousand Oaks, CA: Sage Publications. Part II: Sorting Qualitative Data Organizors: Manfred Max Bergman and Anthony P.M. Coxon Contact:SIDOS Switzerland, and Cambridge University, United Kingdom, [email protected] Keywords: Qualitative data analysis, sorting, categorization Abstract What happens between data collection and the presentation of the research results? In so-called quantitative studies, analysis is focussed on the statistical manipulation of variables. What, precisely, happens in qualitative studies? All qualitative research involves the examination and sorting of bits of information. One typical recommendation proposes to “familiarize” oneself with the raw material (i.e. re-read research notes, transcripts, and other texts repeatedly) until “patterns emerge.” In this session, we will try to disentangle what this may mean. More specifically, we will study how researchers working with qualitative methods arrive at their conclusions, whether through formal or systematic explorations, or through less formal, intuitive processes. From Interview to Results: The Processes of Analysis Barbara B. Kawulich, State University of West Georgia,USA, [email protected] Many qualitative studies conducted today fail to fully explain the procedures used to code and analyze qualitative data, particularly interview data. This paper shares a variety of data analysis techniques, including coding initial interviews, organizing categories in various ways to make sense of interview data, and making interpretations of the data for reporting purposes. Using suggested approaches by LeCompte and Schensul (1999), Strauss and Corbin (1990), Wolcott (1994), and Miles and Huberman (1994), among others, this paper provides numerous strategies used to analyze and interpret qualitative data. Also included are several techniques for teaching these strategies in graduate qualitative research courses. Keywords: qualitative data analysis, coding, interpretation, teaching qualitative research Intersubjectivity and Beyond: The Use of Flexible Content Analysis in Sorting Qualitative Data Margrit Schreier, International University Bremen, School of Humanities and Social Sciences, P.O. Box 750 561, 28725 Bremen, Germany, Tel.: +49-(0)421-2003406, Fax: +49-(0)421-2003303, [email protected] This paper focuses on 'flexible content analysis' as a method for analysing qualitative data (using data from a reception study as an example). Flexible content analysis is conceptualized as lying mid-way between various types of coding as used in qualitative data analysis and quantitative content analysis. 218 In a first part of the paper, the method and the steps it involves will be introduced. Essentially, the method consists of determining the meaning of verbal material by assigning the textual units to categories of meaning that have been assembled into a coding schedule. Construction of the coding schedule will often begin with a deductive outlining of potentially relevant categories, each category constituting a potential 'pattern' based on presumed similarity between texts. In order to assure an optimal 'fit' between categories and the material to be analysed, the categories will then have to be supplemented inductively by examining part of the material. The final coding schedule will comprise precise descriptions of all categories, specifying the conditions under which a textual unit is to be assigned to any one category, including examples, and, if relevant, decision rules. On the basis of this coding schedule, a first trial coding is carried out by at least two coders, and inter-coder agreement is calculated. Usually some revision of the coding schedule will occur, and then the actual coding of the main body of the material takes place. In the second part of the paper, the method will be compared to quantitative content analysis on the one hand and to coding on the other hand. It constitutes a type of content analysis by sharing in that method's double intersubjectivity: First, patterns are established by looking for intersubjective content across different texts, i.e. for conceptual similarities. Second, the categorisation of the texts or textual units is carried out intersubjectively by at least two coders. Flexible content analysis is, on the other hand, like coding and different from quantitative content analysis in allowing for the inductive ad hoc specification and thus 'discovery' of patterns and by extending beyond the surface characteristics of texts. Flexible content analysis, however, reaches its limits where idiosyncratic meanings found in a single text only are concerned. In order to capture these, different patterns have to be established, based not on conceptual similarity, but on locations along theoretically relevant dimensions. Examples of how this is done in coding will be provided. Keywords: qualitative data analysis; content analysis; coding; categorization Towards Rigorous Practice in Qualitative Research Jacques de Wet and Zimitri Erasmus, Department of Sociology, University of Cape Town, South Africa, [email protected] Qualitative research, in particular data analysis, is too often seen as ad hoc, intuitive, unsystematic and thus without academic rigour. We challenge this view. The central purpose of this paper is to illustrate that qualitative data analysis can be systematic, procedural and rigorous. In this paper we provide an overview of analytical procedures we followed during a study about students' experiences and perceptions of 'race' and racism at a Medical School in South Africa. We outline our implementation of these procedures and reflect on their value in optimising our research outcome. We track steps in our analysis by working backwards from one cluster of key findings in the study concerned in order to demonstrate the ways in which we came to these particular findings. Where appropriate, we note the ways in which Computer Assisted Qualitative Data Analysis Software (CAQDAS), specifically the qualitative software package QSR Nvivo, contributed to systematic and rigorous practice at critical points in the analysis. When we outline and reflect on our analysis, we draw on Miles and Huberman's (1994) and Wengraf's (2001) approaches to qualitative data analysis and on Morse et al.'s (2002) understanding of rigour. We conclude that our use of QSR Nvivo facilitated systematic organisation and procedural analysis of our data. QSR Nvivo provides a system of electronic tools for organising, retrieving and verifying data thus enabling one to work with data more efficiently. It does not do the analysis, nor does it think for one. Well-organised data enables researchers to implement procedures more effectively, which in turn contributes to rigorous analysis. Finally, we conclude that while it may be true that some of the accusations that qualitative research is sloppy and unscientific are unjustified, our experience is that qualitative researchers do not instill confidence in their research by continuing to produce research reports where their methods of analysis are not well formulated. The challenge is to identify methodological and analytical benchmarks for 219 qualitative research. This paper demonstrates our latest attempt at “working at sensible canons for qualitative data analysis” (Miles and Huberman, 1994:2). Keywords: qualitative data analysis, rigour, analytical procedures, CAQDAS References Miles, M. B. & Huberman, A. M (1994) Qualitative Data Analysis: An Expanded Sourcebook. Thousand Oaks, CA: Sage. Morse, J.M., Barrett, M., Olson, K., & Spiers, J. (2002, Spring) Verification Strategies for Establishing Reliability and Validity in Qualitative Research. International Journal of Qualitative Methods 1 (2), pp.1-19. Wengraf, T. (2001) Qualitative Research Interviewing. London: Sage Analysing Relationships in Qualitative Data in the Context of Applied Policy Research Josie Dixon, Qualitative Research Unit, National Centre for Social Research (NatCen), United Kingdom, [email protected] In this paper, I begin by exploring the needs and understandings that policy-makers have of qualitative analyses. I go on to consider the implications of these needs and understandings for the conduct of qualitative analysis in applied policy research. In particular, I focus on methods for identifying analytical relationships within qualitative data, the evidential status of these identified relationships and the role of this evidence in informing policy. Policy research is different to academic or ‘pure’ research. The main purpose of applied policy research is not to develop and test academic theories (although there may be theoretical gains from the conduct of applied research). Rather, the commissioners of applied policy research are primarily seeking to better understand and address specific social problems and to do this within relatively limited time scales. Consequently, policy-makers need qualitative research to: deliver quick and timely findings focus on policy-relevant research questions provide findings that are clearly communicated and unambiguous in their relevance and meaning provide findings that are persuasive provide findings that are operationally focused be transparent about how conclusions have been reached and be clear about the status of the evidence engage with what is already understood be demonstrably unbiased, accountable and quality assured I argue that these needs have implications for the analytical approach taken in applied policy research. Specifically, I distinguish between the identification of relationships within qualitative data (within a broadly causal model) and rich, contextual description and I explore the role of each in informing policy decisions. I go on to consider assumptions about causality and the role and treatment of analytical variables in qualitative analyses. In doing so I highlight the tension between policy-led and inductive approaches to qualitative analysis and I propose ways of accommodating this tension with a view to producing rigorous research findings that are useful for policy-makers. I also consider the challenges involved in identifying, and communicating, the strength and status of the various types of evidence generated by qualitative analysis. I propose some practical and defensible approaches to addressing these issues with the aim of reducing unnecessary ambiguity for policy-makers who are using the research to inform policy decisions. Throughout I make reference to Framework, a tool for conducting qualitative analysis used by the Qualitative Research Unit at NatCen, to illustrate some of the issues and points raised within this paper. As well as drawing upon relevant literature, I draw upon my own extensive experience of conducting qualitative research for policy purposes and upon work carried out within the Qualitative Research Unit at NatCen. This includes a recent study, of which one aspect included discussions with policy makers and research managers in Government about quality criteria in qualitative research. Keywords: Qualitative data analysis, applied policy research, evidence-based policy making, causality in qualitative research, generalisability in qualitative research, communicating research findings, Framework 220 STREAM: SURVEY RESEARCH SESSION: COGNITIVE AND MOTIVATIONAL DETERMINANTS FOR RESPONSE EFFECTS IN SURVEY RESEARCH Organizer: Volker Stocké Contact: University of Mannheim, D-68131 Mannheim, Germany, Tel. +49(0)621-181-3432, Fax. +49(0)621-181-3451, [email protected] Keywords: Aquiescence, Cognition, Motivation, Question Order Effects, Response Effects, Response Order Effects, Social Desirability Bias, Theory of Response Behavior, Wording-Effects Abstract Research about the determinants of a variety of response effects in survey research has made considerable progress in recent years. Several cognitive and motivational factors have been found to influence the probability of wording-effects, whether question and response order affects the response behavior or under which conditions answers are biased by social desirability or aquiescence. Among these factors are for instance the cognitive availability of the requested information, subjects’ accuracy motivation, their conversational norms or need for social approval. Despite this progress in research about response effects, the underlying cognitive processes and motivational foundations of several effects are not yet completely understood. It is therefore more research is needed in this respect. Furthermore, at present theoretical explanations and empirical studies often concentrate either on cognitive or motivational determinants for response effects. However, the interplay of both factors can be regarded as an interesting and fruitful question. For instance: How does respondents with high or low accuracy motivation cope with an insufficient cognitive accessibility of information or time pressure? Which information do subjects with a strong approval motive use in order to determine the most desirable response option? Another, but related topic is the still relatively weak theoretical foundation of research about response effects. It would be a great progress to integrate the already available knowledge into a general and complete theory of response behavior in surveys. This session invites contributions relating to all of the topics addressed above, or to any other topic, which improves the explanation and prediction of response effects in survey research. National Identity, Nationalism, or Patriotism: Do we Really Measure what we Assume? Horst-Alfred Heinrich1 and Karsten Stephan2 1 University of Stuttgart, Germany, [email protected] 2 University of Siegen, Germany, [email protected] The usually employed models reflecting national identity, nationalism, or patriotism are tainted with some inconsistencies when empirically tested. Theoretically it is assumed that nationalism and patriotism should form two factors independent of each other. The former should measure respondents’ relationship toward the nation whereas the latter should refer to universalistic values like human rights. Nevertheless, empirical studies report always a positive correlation between the two factors. Therefore, the question is raised whether we really measure what we claim. One problem is that the items of both concepts refer to the same reference object: the nation. Therefore, we think that the standard instruments do not measure patriotism in a way as it is defined. Instead, it seems to us that this concept corresponds to a soft version of nationalism. As a consequence, we wanted to know in how far respondents are able to differ - between the reference objects of nation and universalistic values - and between evaluation of and emotion toward the reference object. In our study we wish to answer these questions. First, we changed the question wording of the instruments usually used. We eliminated double stimuli referring to both nation as well as values. Then, applying several split versions of the same questionnaire to student groups we tried to find out 221 whether variations with regard to question wording enables respondents to differentiate clearer between their attitudes toward the nation and those toward democratic values. Working with this design we can show that patriotism as it is usually defined does only measure a slightly different facet of nationalism. Finally, our work leads to theoretical conclusions concerning the link between universalistic values and the nation. Keywords: nationalism; patriotism; question wording A Cognitive Model for Survey Response Effects Ákos Münnic1, Jaap Murre2 and Willem Saris3 1 University of Debrecen, Hungary, [email protected] 2 University of Amsterdam, The Netherlands, [email protected] 3 University of Amsterdam, The Netherlands, [email protected] A lot has been written about survey response effects. Schuman and Presser (1981) gave a nice overview of different effects. More recently Tourangeau et al (2000) formulated a model to explain these effects. Their model was based on the idea of Zaller (1991) that people create their responses at the spot and use as information the most salient information. However this model does not indicate what is salient and how such a process works. During the last year a research groups at the Netherlands Institute of Advanced Studies (NIAS) has worked on a new model for survey response processes to clarify how these effects could be explained. The model is based on knowledge about neural processes and information processing. After a study of different cognitive models, the ART model of Grossberg (1982) has been selected as the most promising to develop a simulation model for information processing and response behavior. Using this basic model a simulation program Survey ART has been developed consisting of different basic modules of the ART system. In our presentation we will introduce the model and we will show how this model can generate some of the well known response effects mentioned in the literature. Direction of Response Scales, Change of Numerical Values and Respondents’ Behavior Bettina Langfeldt1, Dagmar Krebs2 and Juergen H.P. Hoffmeyer-Zlotnik3 1 University of Giessen, Germany, [email protected] 2 University of Giessen, Germany, [email protected] 3 Center for Survey Research and Methodology (ZUMA), Mannheim, Germany, [email protected] Based on a split-ballot design, the paper approaches the question whether response behavior differs depending on the direction of the response scale as well as on the values attached to the extreme points of the response scale. Therefore respondents were asked to answer different attitude questions by using an 8-point scale offering increasing (split 1) or decreasing (split 2) intensity of how much an item applies to them. Only the extreme points of the 8-point scale were labeled. Data collection was conducted with three different versions of the questionnaire: First, all persons got the same questionnaire with a response scale ranging from “does not at all apply” (value 1) to “applies completely” (value 8), which acts as the baseline measurement. Second, the same individuals were asked again but this time they were assigned to different split versions. For one group of respondents the response scale was labeled as “applies completely” to “does not at all apply” with values ranging from 1 to 8 (split 1). The second group got a questionnaire with a changed response scale ranging from 8 to 1 where the value 8 was labeled with “applies completely” and the value 1 was labeled with “does not at all apply” (split 2). With this design the effect of the direction of the response scale can be assessed as well as the effect of different values attached to the extreme points. Furthermore, item content is supposed to be an intervening variable. If items contain rather unspecific statements or describe rather general “states of mind” as f. e dimensions of motivation, it is expected 222 that distributions (as well as means) differ between the baseline measurement and split versions 1 and 2. With unspecific items respondents might check answering categories at random because it does not matter to a high degree and maybe respondents are less attentive. If, however, items contain specific statements, respondents might feel a stronger appeal to clarify their position by checking the appropriate answering category. Thus, the direction of answering categories is expected to produce no difference in response behavior resulting in comparable distributions (as well as comparable means) between the baseline measurement and the two spilt versions. Respondents are students of economy, social science and education. Analyses is based on 700 questionnaires altogether. Measurement models of the baseline measurement and the two split versions are compared. Keywords: response scales, measurement quality, response behavior, scaling effect. Cognitive and Motivational Determinants for the Strength and Direction of Social Desirability Bias in Racial Attitude Surveys Volker Stocké, University of Mannheim, Germany, [email protected] Several factors have been assumed to determine the strength and direction of social desirability bias. One factor, predicted from personality psychology, is the subjects’ need for social approval. Here, the strength of social desirability bias is expected to increase with the respondents’ situational invariant but individually variable motivation to behave according to social norms. In other approaches it is predicted that different features of the interview situation influences respondents’ beliefs about whether the response behavior may have consequences or not. For instance, whether the questions are self administered or administered by an interviewer affects the privacy of the response situation and how likely subjects expect approval or disapproval from the side of others, present during the interview. Another important determinant for the strength and in particular for the direction of social desirability bias are the respondents’ desirability beliefs. With this beliefs subjects anticipate whether certain response options are positive, negative or neutral evaluated and to select them as an answer is therefore likely to cause approval, disapproval or a neutral reactions from others. The ability of all three factors, the need for social approval, the response privacy and subjects’ desirability beliefs, to predict social desirability bias has been tested empirically. However, nearly all studies have concentrated on just one of these factors and have not taken the possibility into account, that their explanatory power may be interdependent. In the framework of rational-choice theory of response behavior such an interdependence of cognitive and motivational determinants of social desirability bias is predicted. Here, each factor is assumed to be a necessary but not sufficient condition for social desirability bias. Since this implies a non compensatory relation between the determinants, the theory predicts a three-way interaction between these factors. This hypothesis is tested using data from two studies about the response behavior in racial attitude surveys. Keywords: Mode of administration; need for social approval; racial attitudes; rational choice theory; trait desirability SESSION: INTERACTION ANALYSIS AND DATA QUALITY OF SURVEY INTERVIEWS Organizer(s): Johannes van der Zouwen and Johannes H. Smit Contact: Vrije Universiteit, Amsterdam, The Netherlands, [email protected] Abstract A detailed analysis of sequences of coded behaviors of respondents and interviewers in survey interviews (interaction analysis for short) can improve data quality, i.e., the completeness, validity and reliability of the collected data. First, it may offer valuable information about the adequacy of the questionnaire and the wording of the questions and the set of response alternatives, by indicating what questions give difficulties with 223 asking, lead to many inadequate responses and misunderstandings, and require much 'repair' activities of the interviewers. Second, the protocols representing the question-answer sequences indicate which interviewers are less competent (e.g., by deviating from the questionnaire, suggestive probing, or omitting necessary 'repair' activities). Third, the analysis may point at specific categories of respondents showing difficulties with fulfilling the task as respondent, e.g., by giving many inadequate answers or misunderstandings of the questions. The interactions studied may concern closed or open questions; therefor papers using a more quantitative sequence analysis as well as papers using a more qualitative conversation analysis of the interactions are welcome. Papers in this session may stress the effect of characteristics of the questionnaire, interviewers or respondents on characteristics of the interaction, or of the interaction on data quality. Most welcome are papers in which both topics are combined. Conversational and Formal Questions in Survey Interviews Yfke Ongena, Wil Dijkstra and Stasja Draisma, Vrije Universiteit, Amsterdam, The Netherlands, [email protected] Interaction analysis of Question-Answer sequences from a telephone survey shows that the respondent is usually the first to produce problematic behaviors, requiring corrective actions of the interviewer. The occurrence of problematic respondent behavior is affected by the questionnaire design, whereas inadequate interviewer behavior is affected by respondent behavior, rather than directly by the questionnaire design. Especially so-called mismatch answers, i.e. answers that do not correspond to the required answering format, appear to be the most frequently occurring problematic deviation from a ‘paradigmatic’ interaction sequence. Explanations for the occurrence of mismatch answers concern cognitive, conversational and task-related factors. Survey questions with words and phrases common in ordinary conversations (i.e. ‘What is your age?’) can be characterized as conversational. In contrast, survey questions with non-common words and phrases (i.e. ‘What is your year of birth?’) can be characterized as formal. Similarly, a distinction can be made between formal and conversational response alternatives, depending on their frequency of occurrence in common conversations. It is hypothesized that conversational questions with formal response alternatives yield the most mismatch answers, whereas formal questions with conversational response alternatives yield the least mismatch answers. An experiment was conducted in which formal and conversational questions and formal and conversational response alternatives were varied systematically. Most of the questions were derived from existing questionnaires, used in several official surveys concerning health issues. Twelve interviewers interviewed 610 respondents who were randomly assigned to the different versions of the questionnaire. A CATI-program was used that also enabled digital recordings of the interviews (i.e. CARI, Computer Audio Recorded Interviewing). In addition, the experiment was designed to test how mismatch answers can be converted into adequate ones, by systematically varying corrective actions performed by the interviewer. Keywords: Interaction analysis, Question wording, Behavior Coding, Telephone Survey Interviews The Effectiveness of Repair Strategies of Interviewers in Survey Interviews Johannes van der Zouwen and Johannes H. Smit, Faculty of Social Sciences, Vrije Universiteit Amsterdam, The Netherlands, [email protected], [email protected] In interviewer administered questionnaires using closed questions, ideally, respondents only mention their choice of the set of presented response alternatives. Deviations from this ideal can occur because the respondent refuses to answer, or says "I don't know." Or the respondent gives an irrelevant answer 224 or an answer that is relevant and normal within the context of a conversation, but that is not equal to one of the response options presented (a 'mismatch answer') and thus inadequate. Then the interviewer is instructed to intervene and try to get as yet a proper response. The present paper aims at getting insight into the different types of deviations by respondents from the ideal of the standardized survey question, the frequency by which these types occur, the ways in which interviewers attempt at repair, and the strategies that seem to be the most effective. To answer these research questions we analyzed parts of 201 interviews conducted in the Longitudinal Aging Study Amsterdam (LASA), a study of antecedents and consequences of changes in autonomy and well-being of older adults. The fragments were related to four closed questions regarding income (sources, amount, satisfaction, expectations). Transcripts of the question-answer sequences following these four questions were coded for the type of (in)adequate initial answer, and the occurrence, correctness and effectiveness of repair behavior of the interviewer. In 41% of the 804 question-answer sequences the initial response is inadequate; mostly by giving a mismatch answer or responding with "don't know." The distribution over the types of inadequate initial responses is very uneven, and strongly related to the topic and the format of the questions. In 74% of the sequences in need of repair, the interviewer makes an attempt at repair. Part of the repair strategies can be qualified as 'correct' because they will not bias the answers, while other strategies, like probing in a suggestive manner, are incorrect, because they may lead to biased answers. Of all repair attempts, 75% are performed correctly. Most of the repair attempts are effective, i.e., result in an adequate final response. Correct attempts at repair have a significantly higher probability of being successful than incorrect ones. The result of the repair activities of the interviewers is that the number of sequences ending with an adequate response is increased by 20%. The paper contains a list of repair strategies that appear to be both correct and effective. Keywords: interviewer behavior; repair strategy; effectiveness of repair; closed question; adequacy of response; mismatch answer Opening the Black Box in Survey Research. Data Collection as an Ongoing Process of Total Quality Management. John Lievens, Els Bundervoet and Hans Waege, Ghent University, Ghent-Social Science Research and Methodology, and Cultural Policy Research Centre ’Re-Creatief Vlaanderen’, [email protected] The design and implementation of survey research is increasingly guided by principles of Total Quality Management (TQM). These efforts, however, are often solely concentrated on the stages of the preparation of the data collection (such as questionnaire design, selection and training of interviewers) and the evaluation of the data quality once the data collection is completed (coverage, non-response, interviewer effects, etc.). The intermediate stage in which the actual field work of the data collection takes place then remains a black box, often left to the discretion of external data collection agencies. In this stage, however, several processes are in play that have important effects on the quality of the obtained data, such as the capacity of interviewers to convince respondents to cooperate, the capacity of interviewers to fulfil their interviewing task properly, timeliness, respect for contact procedures, and the capacity of the overall organisational structure to uphold the highest standards of methodological quality throughout the entire field work. Effects of these are too often only discovered when the full dataset is available and opportunities for remediation are limited or nonexistent. We propose an integrated TQM-approach for a continuous and largely automated follow-up of the data collection stage. In the paper we elaborate readily applicable measurement instruments and quality indicators that allow to continuously monitor the actual developments on the field and provide the necessary information for timely adjustments. Central elements are indicators for actual progress, response, interviewer effects, and data quality; as well for the total field work as for interviewers and 225 geographical units separately. The input comes from a variety of sources, such as temporary datasets from CAPI-interviews, contact sheets and a database containing the actual status for each address. The integrated procedure is applied to a large scale survey in the Flemish population on cultural participation. From October 2003 until June 2004 approximately 3.000 individuals are interviewed using face-to-face CAPI. PAPI-questionnaires are left for other household members. A complex contact procedure and refusal conversion are used. The data collection is carried out by a market research agency. We demonstrate the use and surplus-value of our integrated TQM-approach, and provide information on its costs. Departing from real-life situations we show how the monitoring system allows to take concrete adjustive actions and discuss to what extent these actions have an effect on the total data quality. Keywords: survey research, data collection, total quality management, data quality indicators SESSION: APPROACHES TO USING MIXED METHODS Organizors: Angela Dale1 and Lynn Jamieson2 Contact: 1University of Manchester, [email protected] 2 University of Edinburgh, [email protected] Keywords: mixed methods; qualitative and quantitative; epistemology Abstract Mixed methods are increasingly seen as an innovative approach to social research. However, they raise many important issues, some epistemological, some related to the nature of the research question and some more practically oriented. What they do not do is provide a simple basis for ‘triangulation’. Through a combination of invited papers and an open-call, the session will include papers that are empirical examples of a mixed method approach to a research question or topic and those which mix methods drawn from within or between paradigms. More specifically invited papers will deal with an investigation of a single topic based on five methodologically discrete but linked projects which use: quantitative secondary data, qualitative interviews (including online), visual analysis, focus groups, and diaries. The process of integrating this research, and the issues that need to be resolved, form an important part of the data. A second potential paper will address the theoretical, methodological and empirical implications of undertaking ethnographic research that exploits contemporary information and communication technologies (ICT) and thus involves a wide range of data sources. A third potential paper will be based on combining evidence from qualitative and quantitative studies in systematic reviews. This will examine methods of using narrative approaches to the synthesis of evidence from multiple studies. Integrating Quantitative and Qualitative Research: How is it done? Alan Bryman, Department [email protected] of Social Sciences, Loughborough University, England, There appears to have been relatively little investigation of the ways in which quantitative and qualitative research are combined, referred to by Bryman (2004) as multi-strategy research, in spite of the fact that: there seems to have been a growing tendency for social researchers to employ both quantitative and qualitative research in their investigations; there have been several accounts of the advantages of such an approach and some of the processes involved (e.g. Bryman, 1988; Creswell, 1995; Tashakkori and Teddlie, 1998, 2003); and the ‘paradigm wars’ of the past, with their emphasis on epistemological and ontological issues have abated (though not disappeared). This paper will entail a preliminary report of the findings of a content analysis of articles in refereed journals that have 226 employed a multi-strategy research approach at the levels of data collection, data analysis, or both. The disciplines covered by the investigation are: sociology, social psychology, geography, management and organizational behaviour, and media and cultural studies. The paper examines a variety of issues. It explores the different types of data collection method and research design that are involved in investigations using a multi-strategy research approach. In doing so, it uncovers some confusions surrounding precisely what is typically taken to be multi-strategy research in the research community. It examines the rationales that are given for employing a multi-strategy research approach and contrasts these with how quantitative and qualitative research are in fact employed in the studies concerned. The contrast between the claims that are made about why both quantitative and qualitative research were employed and the actual use of such an approach provides some interesting contrasts. Use is made of Greene, Caracelli and Graham’s (1989) five-fold classification of approaches to combining quantitative and qualitative research in evaluation research, but a more fine-grained approach was also devised and employed in the analysis. The findings also strongly suggest that the discussions associated with the paradigm wars are largely ignored by researchers who, at the level of research practice, typically push such discussions to the sidelines of their concerns. The paper considers some of the implications of the findings for the practice of multi-strategy research, such as some of the ways in which the frequently-used term ‘triangulation’ is employed in actual research that claims to employ it and how far quantitative and qualitative research are employed to answer different explicitly formulated research questions. Keywords: multi-strategy research; mixed-methods research; quantitative research; qualitative research Multi-Methods, Triangulation, and Integration – Different Goals or Just Different Terms? Jo Moran-Ellis, Victoria Alexander, Ann Cronin, Mary Dickinson, Jane Fielding, Judith Sleney, and Hilary Thomas, Department of Sociology, University of Surrey, Guildford, Surrey GU2 7XH, UK, [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected] There is a long standing interest in social science research in the idea of using more than one method to investigate a research question or topic in order to gain better understandings and insights, better measurements, or to open up alternative senses of ‘the truth’ of the matter. The notion of ‘knowing more’ through different approaches has been variously captured in ideas about triangulation, multimethods, integration, combining methods, complementarity, and mixing methods. However, a survey of the literature in this area reveals that these terms are used with little consistency, and that there are many unresolved debates concerning the use of multi-methods, primarily articulated around epistemological or ontological dilemmas. This paper is based on a project which empirically examined how the notion of ‘vulnerability’ differs when viewed through a variety of epistemological lenses (quantitative, qualitative and visual), and across a range of populations (whole households, people living alone, and people who are homeless) in the same geographical locality. The empirical focus was on how people experience vulnerabilities in everyday life, and how they manage them. The methodological aim was to make visible the methodological and practical issues of undertaking an integrated approach to a research question, and examine how strategies of integrating data impact on each stage of the research process. In the paper we explore current conceptualisations of approaches that use more than one method or methodological approach, and then focus on approaches which specifically aim to integrate data or findings. In the course of this, we raise questions of ‘what is integration and what, if anything, is distinct about it as a methodological goal?’. Drawing on our empirical work, the paper demonstrates that integration as a goal carries implications for the research process from conceptualisation of the research question through to interpretation of findings. These implications are both methodological and pragmatic in nature. We examine what conditions may be necessary to be able to make a distinct 227 claim of integration in approach, and consider the limitations of other conceptualisations of what it is to use more than one method. Finally, we argue that integration of approaches may not always be possible for a variety of reasons, and indeed that it should be taken up with appropriate caution and hesitation rather than simply embraced as a panacea for the limitations of individual methods. Keywords: integration, mixed methods, methodology Qualitative Methodology as a Parenthesis for Quantitative Methodology Mihai Pascaru and Carmen Costea-Bezerita, University 1 Decembrie 1918 Alba Iulia, Romania, [email protected] The proposed paper starts from the premise that a good questionnaire, for a good sociological inquiry, means responding two fundamental conditions: 1) a strong theoretical basis 2) accurate knowledge of the daily life of the space to be studied. The existence of the second condition necessarily implies the use of the qualitative methodology and of some of its instruments: the participative observation, the semi-structured interview, the group interview etc. Yet, there is no perfect sociological inquiry. The limits of the methodological realm, of time or money influence the results of a sociological inquiry to reflect only partially the reality of the area proposed for study. However, even if many things can be explained, not many can be understood. Some of these limits can be, in our opinion, diminished if the results of the sociological inquiry are restored to the questioned. Data from the sociological inquiry will be presented to the respondents and these are invited to interpret them, to identify motivations of their own actions etc. the restoration of the results can be made by a semi-structured interview or by a group-interview, anyway, by an instrument specific to the qualitative methodology. From this reason we support the syntagm << the qualitative methodology as a parenthesis for the quantitative methodology>>, and we propose a working schemata qualitative-quantitative-qualitative. This working schema was also used in some investigations made by us and by our collaborators about the problem raised by the Post-Communist transition in Romania. We intend to present a part of these results in the paper we are proposing. Keywords: qualitative methodology, quantitative methodology, restoration of the results Multi-Semiotic Ethnography, Bella Dicks, School of Social Sciences, Cardiff University, UK, [email protected] Ethnographers, like other researchers, currently have a broad range of media at their disposal both for conducting fieldwork and representing their interpretations and analysis. These include digital media such as photographs, video film, audio-recordings, graphics, and others besides. Through the computer 'writing space'of hypermedia (Jay Bolter), these media can be integrated together, alongside more conventional written interpretation. However, integration poses a number of potential problems. In particular, different media can be seen to 'afford' different kinds of meaning (Kress and Van Leeuwen). Further, organising and structuring a hypermedia ethnography poses challenges relating to readability and 'navigability'. Our paper explores the implications for ethnography of integrating different media in ethnographic research. It will explore the challenges of 'writing' ethnography in a multi-media platform. Rather than seeing these media forms as discrete, it outlines an approach to ethnographic work which sees meaning as emerging from the fusion of differently-mediated forms into new, 'multi-semiotic' modes. It therefore goes beyond the current interest in visual methods, suggesting ways of developing the project of a multi-modal ethnography. 228 A Delicate Dialectic: Mixing Qualitative and Quantitative Methods in Two Studies for US Federal Government Agencies Susan G. Berkowitz and Cynthia S. Robins, Westat, Rockville, MD USA, [email protected], [email protected] In the "benign abstract," the idea of mixing qualitative and quantitative methods in the same study design has gained increasing acceptance. However, it can still be extremely challenging to come to terms with the realities of what actually happens---and what it means--- when divergent sets of epistemological assumptions and different methodological approaches are brought to bear on a single problem or research question. This paper will discuss the challenges and rewards of two applied research studies conducted for US Federal government agencies. Both started as straightforward population surveys, but generated more in-depth qualitative "spin-offs," whose methods and findings came to challenge many of the assumptions underlying the quantitative paradigm. Both studies have generated an ongoing dialogue across methodological boundaries with substantive as well as methodological implications. One series of studies involved in-depth telephone interviews with purposively selected subsets of the respondents from a larger, population-based survey of youth. The interviews probed the respondents' approaches toward career decision-making as well as their views and images of military versus civilian lifestyles. The fine-grained narrative format of the interviews, allowing the respondents' to speak in their own voices, revealed the deeply contextual nature of their decision-making processes. The findings illustrated the fallacy of giving undue weight to one-time survey responses and attributing "inscribed" propensities to youth whose life circumstances were fluid. They also revealed the deep influence of social class (over racial and ethnic) differences. These results helped to inform and temper the interpretation of the survey findings and catalyzed changes to the survey instrument. The second study is exploring how community context affects elderly Medicare managed care beneficiaries’ ability to obtain needed services. Four study sites were selected for their “average-ness” across several key demographic variables (e.g., median household income). Using rapid ethnographic assessment techniques (e.g., participant observation, in-depth interviews, focus groups), we are examining the array of services in each site; the extent to which beneficiaries are aware of and make use of services; and community factors (e.g., cultural geography) that impede or facilitate efforts to receive care. Initial findings suggest not only considerable cross-site variation, but also important cultural differences within a given site. These findings challenge tendencies to rely too heavily on statistical averages by showing the differences that can underlie apparently similar “average” patterns. Further findings will stimulate ongoing dialogue with the survey findings and methods. Keywords: mixed methods, combining qualitative and quantitative approaches, working across the methodological divide SESSION: ESTIMATING INFORMED OPINION: ADVANCES IN THE THEORY AND MEASUREMENT Organizer: Patrick Sturgis Contact: University of Surrey, UK, [email protected] Keywords: Informed Opinion; Survey; Non-attitude; Deliberation Abstract Survey researchers and scholars of public opinion have long been concerned with the quality of the attitudinal data they collect. ‘Rationally ignorant’ citizens are too little engaged in and knowledgeable about public affairs to provide anything more than top-of-head ‘nonattitudes’ in survey interviews (Converse, 1964, 2000). Public opinion, as measured in surveys amounts, thereby, to little more than an ‘echo chamber’ of elite discourse played out through the media (Key 1961; Fishkin 1995). The weight of empirical evidence supporting this dismal characterisation (see Delli-Carpini and Keeter 1996) presents a troubling problematic for normative theories of democracy and raises serious 229 concerns about the validity and reliability of attitudinal data more generally. Various methods have been mooted as solutions to the nonattitude problem, all of which, by various means, aim at eliciting what has been termed ‘informed opinion’ (Price and Niejens, 1997; 1998). Some of the more prominent examples in this regard are Deliberative Polling (Fishkin 1995; Fishkin and Luskin, 2002); Planning Cells (Dienel 1978; Renn et al 1984); Choice Questionnaires (Neijens 1987); and simulation modeling (Althaus, 1997; Sturgis 2003). This session invites papers which contribute to this important and growing area of social research methodology. Papers making empirical comparisons between different methods and those which focus on the validity and reliability of estimates of informed opinion will be especially welcome. CAPI-based Information Intervention (CIi): A New Way of Estimating Informed Opinion Patrick Sturgis1 and Helen Cooper2 1 Department of Sociology, University of Surrey, UK 2 Chris Fife-Schaw, Department of Psychology, University of Surrey, UK Scholars of public opinion are justifiably reticent about the use of survey measures of issue preferences to represent the ‘will of the people’. Evidence abounds that citizens in modern capitalist democracies have a shallow grasp of matters politic and provide labile and ephemeral attitudinal responses as a result. Various methods have been proposed as solutions to this ‘nonattitude’ problem, all of which, by various means, aim at eliciting what has been termed ‘informed opinion’ (Price and Niejens, 1997; 1998). Some of the more prominent examples in this regard are Deliberative Polling (Fishkin 1997); Planning Cells (Dienel 1978); Choice Questionnaires (Neijens 1987); and simulation modeling (Althaus, 1998). In this paper we report on a study using a new approach to estimating informed opinion; the CAPI-based Information intervention (CIi). The basic design of CIi is as follows: a probability sample is drawn from the target population and respondents administered a questionnaire via Computer Assisted Personal Interview (CAPI). Approximately six months later respondents are re-contacted and randomly allocated to treatment and control groups. Treatment groups are presented with factual information in the form of a short film as part of the CAPI interview, the control group receiving no information. All respondents are then readministered the original questionnaire, with the causal effect of the information being equivalent to the average of the differences in opinion between the two groups. A third wave of data collection, via Computer Assisted Telephone Interview (CATI), is then conducted approximately three months after second interview to examine the stability of observed information effects. While CIi has clear limitations in the depth of information that can be communicated relative to other methods of estimating informed opinion, its potential strengths lie in its use of random sampling and the high penetration of information provision across the sample as a whole. This paper reports results from the first two waves of a CIi study on British attitudes to genomics conducted between June 2003 and April 2004. We evaluate the average causal effect of information provision across a range of attitudes toward modern genetic science. We also examine respondents’ own appraisals of the accuracy and relevance of the information provided. We conclude by considering the general utility of the CIi approach for estimating informed opinion. Toward Informed Opinions by the Means of Deliberation Kasper M. Hansen, University of Southern Denmark, Department of Political Science, Campusvej 55, DK-5230 Odense M, Denmark, [email protected], www.sam.sdu.dk/staff/kmh. Since George Gallup introduced the first public opinion poll based on random sampling procedures, the number of polls conducted have increased exponentially. Today the polls are part of our everyday lives not to mention the daily news flow. Even though the traditional opinion polls dominate, literature has been giving increasing attention as to how to measure public opinion more sophisticatedly that the usual top-of-the-head responses given in traditional opinion polls. 230 The increasing attention to more sophisticated ways of measuring public opinion builds on arguments from at least two overlapping research traditions. Research within the public opinion tradition has shown that over time the changes in public opinion tend to fit a random pattern, because large parts of the public simply lack opinions. That is, they only possess nonattitudes. The prevalence of nonattitudes may be due to a number of factors such as the complexity of the issues, how often the public is confronted with the issues, the public awareness of the issues, and how much thought the public has given the issues before expressing their opinions. The social choice tradition is concerned with the problem that arises when individual opinions are aggregated in order to find the winning alternative. Different rules of aggregating might provide different winners and as the choice of aggregating rule is rather arbitrary or at least normatively contested, aggregating individual opinions become a general problem in any democratic society. Furthermore, individuals might try to influence the aggregation by behaving strategically when expressing their opinions. Strategic behavior and decision about which aggregating rule should be used questions the entire idea of measuring public opinion and providing aggregated results - not to mention democracy more generally. One way of confronting the challenges of these two research traditions is through deliberation. By given the public the opportunity to deliberate and providing the public with information before they express their post-deliberative opinions, it is possible to confront the nonattitude-challenge to public opinion as well as the problem of picking the winning alternative. Deliberation increases the public awareness and increases the level of information causing the number of nonattitude-answers to decrease. Furthermore, deliberation helps the individual to provide a consistent ranking of the different options at stake, which makes the winning alternative less dependent on the rule of aggregation. The deliberative solution to the challenges of the two research traditions can be empirical verified simply by comparing pre-deliberative with post-deliberative opinions. This was done through a quasiexperiment of four deliberative hearings that were held across Denmark in the fall of 2003. A total of 168 self-selected from a random sample participated in the hearings. The issue was a new regional structure in Denmark. The participants were recruited through a telephone interview. In the telephone interview the participants were asked to rank the three options for a new regional structure on four different dimensions. These questions were repeated to the participants by the end of the events. Before the events the participants received written information on the issue and during the events the participants debated in small groups with each other lead by a neutral moderator and in plenary session with top regional politicians, experts and other key interests represented. Each of the four events lasted 5 hours. The analyses of the pre- and post-deliberative opinions show that information and deliberation decrease the number of “don’t know” answers and cause many participants to change their opinions. Two out of five participants change their first priority for another of the two options comparing the pre- with the post-deliberative opinions. Furthermore, post-deliberative opinions reflect more informed opinions measured on the level of knowledge, consistency, singled-peakedness and absence of cycles. Hence the post-deliberative opinions are more transitive than pre-deliberative opinions. Finally it proves that whereas pre-deliberative opinions are affected by rule of aggregation post-deliberative are not. Keywords: Deliberative Poll, informed opinions, opinion consistency, opinion transitively, singledpeakedness, cycles, deliberation, deliberative democracy, nonattitudes. Gauging Informed Opinion amongst Young Citizens Marcus Longley1, Christian Thomas2, Rachel Iredale1, Anita Shaw2 and Angela Burgess2 1 University of Glamorgan, UK, 2Wales Gene Park, Cardiff, UK Citizens’ Juries are fast becoming a familiar feature of the on-going attempt to engage citizens in the policy-making process in the UK and elsewhere. They derive from the jury model used in criminal trials, and are predicated on the assumption that the ‘verdict’ of a (more or less) randomly-selected group of citizens, properly informed with all the relevant evidence, has high validity, and can help to reduce the ‘democratic deficit’ which characterises many representative democracies. They are designed to put the citizen in charge – allowing them to decide the topic for investigation, to determine 231 the agenda, to call and question witnesses, to deliberate, and to reach a conclusion or ‘verdict’. The growing experience with such juries has highlighted various methodological and philosophical problems with the approach, some of which have been successfully addressed, and others remain problematical. This paper will provide a brief overview of the more significant of these issues, and how they have been addressed, before describing a novel approach currently being undertaken in South Wales, UK. This involves a ‘jury’ of people aged 16-19, broadly representative of their age group, to be held in September 2004, and focusing on aspects of genetics and reproductive decision-making. Both the age of the jurors and the topic make this experiment interesting. These parameters have helped to shape the method, which is designed to enable young people to specify as many as possible of the elements of the method, including the nature of the interaction with the witnesses, the role of facilitation, the approach to deliberation, and the nature of the verdict. Wales is particularly fortunate in having a relatively well-developed infrastructure for the involvement of children and young people in policymaking at the local level, and a national policy which lays emphasis on the implementation of the UN Charter on the Rights of the Child, and the UK’s first Children’s Commissioner. The paper will describe how the jury approach fits with such an infrastructure and history, and how the various methodological and epistemological challenges are being defined and addressed. It will conclude with some reflections from the author and from participants on how successfully the approach addresses these challenges, and how it might be developed and used in other circumstances. Keywords: Informed opinion, citizens jury, genetics, young people Do People really care about GM Food Risks? A latent State-Trait Model to differentiate ‘Attitudes’ from ‘Opinions’ amongst the UK Public Nick Allum, Department of Sociology, University of Surrey, Guildford GU2 7XH, England, [email protected] In the past few years, much has been made of public opposition, in the UK and elsewhere in Europe, to the commercialisation of genetically modified (GM) food products. However, some recent research has indicated that the issue of GM food has little salience for large sections of the general public (Gaskell, 1997; Gaskell, Allum et al., 2003). This begs the question: do people really hold stable, enduring attitudes about the issue, or do they simply generate an ill-considered ‘top-of-the-head’ opinion in an interview situation? This paper explores this question by conceptualising attitudes as ‘traits’ and opinions as ‘states’. In psychological research, states and traits have often been presented as mutually exclusive entities, for instance mood is a state while introversion is a trait. However, it is more realistic to assume that most psychological constructs vary along a continuum of stability or what Kenny and Zautra call ‘traitness’ (Kenny & Zautra, 2001). The paper presents analyses from a UK panel survey using latent state-trait (LST) structural equation models (Steyer, Schmitt, & Eid, 1999) to examine the stability of public perceptions of GM food risk over time. The LST model partitions variance in responses to survey items about GM food risk into trait and state components that correspond to ‘attitude’ and ‘opinion’ respectively. Results suggest the existence of relatively stable views amongst the UK public, particularly in relation to the affective rather than cognitive aspects of attitudes to GM food risk. Keywords: risk perception; GM food; latent state-trait model; attitude stability Self-reported Familiarity and Coherence of Opinions on Biotechnology Sally Stares, Research student, Social Psychology Department, London School of Economics and Political Science, UK, [email protected] The question of whether ‘informed’ opinion can be distinguished from ‘non-attitudes’ is a longstanding concern in survey research. It has been widely discussed in recent research into public perceptions of biotechnology. Eurobarometer survey data suggest that for about half of the respondents biotechnology is an unfamiliar topic: this is evidenced by respondents’ own reports of 232 whether they have heard of various applications of biotechnology, and of whether they engage with the subject matter via the press, informal conversation, etc. It is perhaps also reflected in high rates of ‘don’t know’ responses (over 30% on some questions) to items asking for opinions on biotechnology. However, we cannot rely solely on such measures to indicate whether or not respondents hold informed opinions or not: self-reported familiarity with the topic does not necessarily equate with informed opinion, and frequency of ‘don’t know’ responses does not necessarily equate with lack of it. A ‘don’t know’ response may be the product of a considered and nuanced opinion, just as a non-‘don’t know’ response may be instantaneously given and substantively empty. This paper attempts to identify informed opinion on biotechnology by assessing the coherence of sets of responses to multiple questions. Latent trait and latent class analyses are used to take a wide-angled view of response profiles for batteries of items, and a comparison of the methods is presented. Sets of items are taken from the Eurobarometer data of 2002, including some which call for judgements of utility, riskiness and moral acceptability for various applications of biotechnology, and a number of items covering self-reported levels of familiarity with the topic area. These self-report items and ‘don’t know’ responses are interpreted with reference to connotations of coherence of opinion. The paper concludes with a discussion of the implications of the findings, both methodologically and substantively – that is, how confident we can be of saying who really thinks what – in terms of reporting levels and distribution of public opinion, and in terms of using latent class allocations or latent trait scores in further multivariate analyses. Keywords: survey, latent class models, latent trait models, self-reported familiarity, biotechnology Measuring Nonattitudes through Response Latencies Piet Sellke, University of Stuttgart, Institute for Social Sciences, Department of Sociology and Empirical Social Research, Keplerstrasse 17 II, D – 70174 Stuttgart, Germany, Tel: +49 (0)711/ 1213577, Fax: +49 (0)711/ 121-2768, [email protected] It is investigated whether nonattitudes can be measured through response latencies. Nonattitudes are directly linked to the concept of attitude strength. The definition of attitude strength, however, is controversial. Krosnick and Petty (1995) define attitude strength as a latent psychological construct, represented in memory through specific attitude propositions. Persistence and impactfulness (e.g. on behavior) are hereby attitude strength’s consequences. However, it remains unclear what specifically a strong attitude is, that is to say if defined with the characteristics of persistence and impactfulness the definition is tautological. Another definition relies on attitude strength as a heuristical label (Krosnick/Petty 1995). The more persistent and impactfull, the stronger the attitude is. Persistence and impactfulness work in this definition as causal indicators of attitude strength. However, high consistency of an attitude in this definition is caused by a strong attitude, which means in turn that the attitude has an high impact. Thus, this definition is tautological, too. Due to these problems in the definition of attitude strength, research has focused on indicators of attitude strength. Some of the most investigated indicators are certainty, importance, vested interest, extremity, and accessibility. Moreover, it was proposed to investigate the dimensions of attitude strength through factor analysis. One of the most promising results in this matter relies on the differentiation between meta-attitudinal and operative measures of attitude strength (Bassili 1993). Whereas meta-attitudinal measures rely on respondent’s self-reports (i.e., judgements over judgements), operative measures rely on the process of information processing itself. These operative measures are indicated by attitude accessibility (measured with response latency), extremity, and ambivalence. Especially attitude accessibility is based directly on cognitive processes underlying attitudes. The idea of attitude accessibility is based on Fazio’s research: Attitudes vary in their attitude strength based on the strength of an associative link between attitude object and evaluation (Fazio 1990). The stronger this link, the higher the likelihood of attitude activation when confronted with the attitude object. Further, the stronger the association, the faster the respondent should be able to express the attitude. Thus, response latency should be used as an indicator of attitude’s associative strength. Fazio thought herby of an continuum between nonattitude on the one side and strong attitudes on the other side. Is no attitude accessible, it’s a nonattitude and respondents have to generated an attitude ‘on the spot’. Is an attitude highly 233 accessible, it is supposed to be a strong attitude. Thus, response latency now can be used as an indicator of nonattitude. However, response latency is multidimensional. It does not only represent the associative strength, but also the mode of processing (e.g. spontaneous versus thoughtful). For empirical applications response latency’s biases have to be controlled. In this paper the idea of response latency as an indicator of accessibility shall be explored theoretically. Further, statistical procedures of controlling biases of response latency will be presented. Findings suggest that response latency can indeed be used to measure nonattitudes, however not without controlling the possible biases and also not without theoretical underpinnings. Keywords: Nonattitude, Response Latency, Survey Methodology References Bassili, J.N. (1993): Response latency versus vertainty as indexes of the strength of voting intentions in a CATI survey. Public Opinion Quaterly, 57, 54-61. Fazio, R.H: (1990): Multiple Process by which Attitudes guide Behavior: the MODE Model as an integrative framework. Advances in experimental Psychology, 23, 75-109. Krosnick, J.A. / Petty, R.E. (1995): Attiude Strength: An Overview. In: Petty, R.E. / Krosnick, J.A. (Eds.): Attitude Strenght: Antecedents and consequences. Hillsdale: Lawrence Erlbaum. 1- 24. SESSION: TELEPHONE SURVEY METHODOLOGY IN THE 21ST CENTURY Organizor: Edith D. de Leeuw Contact: Utrecht University, Utrecht, The Netherlands, [email protected] Keywords: telephone survey methodology, call screening, non-contacts, non-response, cell phones, mobile phones, mixed mode, data quality, interviewers. Abstract Telephone surveys have been one of the main data collections methods in the past 30 years. It was at its highlight of importance in the eighties and nineties, but became under pressure again at the end of the 20th century. Societal and technological changes make that our standard good practice is no longer optimal and force us to adapt the methodology and redesign telephone surveys. Among the societal changes are the increased individualization, which influences contact, and increased ‘selling’ by phone influencing willingness to participate. Among the technological changes are the increased use of screening devices, such as telephone answering machines, voice mail, and caller-ID, and the growing use of mobile (cell) phones as second or even only telephone connection. Central in this session will be methods to increase response and data quality in telephone surveys. Abstracts are invited on all aspects of telephone survey methodology. Main topics are: (1) coverage and sampling, (2) reducing nonresponse, (3) reducing measurement error, (4) interviewer training, (5) mixed-mode strategies. Papers discussing the role of modern technology (e.g., mobile/cell phones) are of special interest, but all methodological papers aiming to raise the quality of modern telephone methodology are welcome indeed! Part I: New technology and its Influence on Coverage, Sampling, and Nonresponse Finding (and Listing) the Unlisted: A Strategy for Achieving a Listed Sample’s Cost Savings without Sacrificing Coverage Anthony Salvanto and Kathleen Frankovic, CBS News, New York, USA, [email protected] In pre-election polling, the lure of sampling from lists of registered voters instead of random-digit dialing is obvious: it offers more productive interviewing, shorter field times, and lower costs. Across the United States, many states and municipalities are now modernizing their voter systems and voter 234 databases – making these lists more widely available, and perhaps even more alluring to the survey researcher. However, despite the potential gains in efficiency, using only list-based samples for surveys can result in substantial non-coverage error, because lists can vary greatly in completeness, coverage and quality due to a variety of factors. CBS News developed a strategy for avoiding non-coverage when calling from lists, while retaining many of the operational advantages of using them. The paper will report on several tests of the strategy, which uses a combination of both list-based samples and tailored RDD samples. It presents findings from CBS News pre-election polls in Senate races and a 2004 Presidential primary. The paper examines the specific biases found in the lists and their sources. For example: which groups and voter types were excluded, and how variations between the local agencies that compile lists can introduce measurable geographic biases. It offers suggestions, based on our results, for how these potential biases can be measured and avoided, and also discusses the cost savings and improvements in efficiency that can result from the method. The analysis also examines the survey results to see if information on the lists matches respondents’ reported demographics and behavior. Survey data is matched back to the list file after the election to compare reported and actual participation, and whether demographic information on the lists is verified by the respondent. Keywords: telephone sampling, telephone survey methodology, data collection, data quality Where Can I Call You? The “Mobile Revolution” and its Impact on Survey Research and Coverage Error: discussing the Italian Case Mario Callegaro1 and Teresio Poggio2 1 Program in Survey Research and Methodology, University of Nebraska, Lincoln, [email protected] 2 Department of Sociology and Social Research, University of Trento, Italy, [email protected] Italy has one of the highest mobile phone penetration rates among European countries. The rapid “mobile revolution” taking place since the late 1990s is changing the frame for landline telephone surveys, with an increasing impact on their coverage error. In 1993, the annual face-to-face Multipurpose Household Survey of the Italian National Institute of Statistics (ISTAT) began collecting data on the presence of fixed telephones in Italian homes. Within the same survey, ISTAT has also been collecting data on mobile phone ownership since 1997, in order to account for a quick rise of mobile phone only households (MPO) within Italian society: family units with no fixed telephone but owning mobile phones were only 1.8% of all households in 1997, while the same figure rose to 13.1% in 2002. This paper briefly describes the main traits of the Italian phone market, the significant changes that are taking place and the rising impact of MPO households. Some comparisons with other European countries are also provided. Then, it discusses the main results of research carried out on coverage error in landline CATI surveys, using Multipurpose Household Survey data. A telephone survey that uses a landline frame excludes the households with no phone at all (3.9% of all family units) and the MPO households, creating a coverage error of 17% (13.1% + 3.9%). Moreover, the chances of not owning a fixed telephone, and hence the likelihood of being excluded from the sample of such a survey, differs dramatically among households by region, according to the type of family unit and by its phase in the life course. Level of education and social class also count. Family units living in Southern Italy and in the islands are more likely to be excluded from these kind of surveys: they have more chances both to be MPO households and not to own a phone device at all. Singles, younger family units, those with medium level of education, and working class households are then more likely to be MPO units. While older households, those with lower education, and farmers are more likely not to own any phone device. Taken together, these differences clearly introduce a not-negligible bias in landline telephone surveys. Possible solutions to this problem are discussed from a methodological point of view. 235 Keywords: Telephone survey methodology, mobile (cellular) phone surveys, mobile only households (MPO), coverage bias, Italian Multipurpose Household Survey. Mobile Phone Surveys in Hong Kong: Methodological Issues and Comparisons with Conventional Phone Surveys Liam K P Lau, Social [email protected] Sciences Research Centre, The University of Hong Kong, In Hong Kong, in 2003, there were households which were not equipped with conventional phones for the first time. On the other hand, at the end of 2003, mobile penetration rate was 100%. The mobile penetration in Hong Kong is very high which makes a mobile phone survey generally feasible and also, some households nowadays are not equipped with conventional phones which makes a mobile phone survey important to see if it can capture this sub-population. The current research aimed at examining and comparing methodological issues and survey results of a mobile survey and its corresponding conventional component. The sampling frame was the Hong Kong general public aged 12 or above. Conventional numbers were selected systematically from the residential telephone directory with their last digits randomized. Mobile numbers were generated by random digit dialing stratified by mobile networks and operators. Call receivers were invited to participate in a sociolinguistic survey of language use in Hong Kong. Telephone interviews were conducted with a computer-assisted telephone interviewing system from 18:30 to 22:30 during weekdays. Each working number was tried for three times on different days. The response rates of the mobile and conventional components were 50% and 47% respectively. The mean interview lengths per question of the two components were not significantly different. The mobile component included higher percentages of respondent who were working, males, never married, and respondents with higher educational level while the conventional component included higher percentages of students, homemakers, the retired, females, now married, and respondents with lower educational level. Respondents of the conventional component were significantly older than that of the mobile component. For the mobile component, 82% of the respondents who completed the whole interview were willing to talk with a mobile phone without diverting the call to a fixed line telephone for about 18 minutes on average although they have to pay for it. 16% of them did not have a fixed line telephone at home. This ‘mobile only’ group included higher percentages of the unemployed and widowed/divorced/separated than the conventional component. Mobile surveys are generally feasibly in Hong Kong because of high mobile penetration rate and satisfactory response rate. The two components of the current study have covered respondents with different demographic characteristics and the ‘mobile only’ group also has some distinct characteristics when compared with the conventional component. Conventional telephone survey results may not represent some sub-populations such as the unemployed very well. Keywords: mobile phones, coverage, telephone survey methodology Occurrence and Consequences of Multiple Telephone Lines, Answering Machines, and Voice Mail Chris Barnes and Alicia Burrington, Center for Survey Research and Analysis, University of Connecticut, USA, [email protected] Scientific telephone surveys have traditionally weighted households with multiple telephone lines to adjust for disproportionate probabilities of selection. This method relies on the assumption that all telephone lines in a household are used in the same way and answered with equal frequency. As an increasing number of households have more than one telephone line, the impact of adjusting for multiple telephone lines increases substantially. If telephone lines are not answered with equal frequency, adjusting a household’s probability of selection by the number of telephone lines has the potential to increase rather than reduce bias. 236 This paper, using new data from a national (U.S.) telephone survey expands on a survey of Connecticut residents from October 2003. It will shed light on this issue in two ways. First, it will explore the distribution of the number of telephone lines, answering machines and voicemail systems, and privacy managing devices per household; including the frequency of use and the purpose for each telephone line. Second, the paper will demonstrate that telephone lines in households with more than one telephone line are often used for different purposes and are not answered at the same frequency as the primary telephone line. The implications of the bias created by the use of the current weighting techniques are also discussed. Keywords: multiple phone lines, weighting, bias, mobile phones Calculating Response Rates for Mobile Phone Surveys. A Proposal of a Modified AAPOR Standard and its Application to three Case Studies Mario Callegaro1, Trent D. Buskirk2, Linda Piekarski3, Vesa Kuusela4, Vasja Vehovar5 and Charlotte Steeh6 1 University of Nebraska, Lincoln, USA, [email protected] 2 Eastern Virginia Medical School, USA, [email protected] 3 Survey Sampling Inc., USA, [email protected] 4 Statistics Finland, [email protected] 5 University of Ljubljana, Slovenia, [email protected] 6 Georgia State University, USA, [email protected] Since 2000, surveys calling mobile phones are becoming more common in European research firms. A survey on mobiles phone poses new challenges to the survey researcher including the calculation of response rates. New call dispositions codes are possible and some dispositions of RDD landline telephone surveys are no longer applicable to mobile phones, or have a different meaning. The present paper analyzes the AAPOR Standard Definitions and proposes mobile phone final dispositions of case codes for mobile phone surveys. Each case is discussed reporting data on mobile phone usage. An application of the new codes with a calculation of response rate, cooperation, refusal, and contact rates is done using the data from three mobile phone surveys. The first one was carried out by Statistics Finland, in a nation with one of the highest mobile phone penetration rates and mobile phone only household rates. The second one was carried out by the RIS in Slovenia, a nation that mirrors the penetration rates and mobile phone only household percentages of many European nations, such as England, for example. The last study was conducted by the Survey Research Center of the Georgia State University and highlights the differences between the U.S. mobile phone system and the European one. Keywords: Mobile (cellular) phone survey, RDD landline telephone survey, response rates, AAPOR standards, sampling design The Influence of Advance Contacts on Response in Telephone Surveys: A MetaAnalysis Edith de Leeuw, Joop Hox, Elly Korendijk and Gerty Lensvelt-Mulders, Department of Methodology and Statistics, Utrecht University, The Netherlands, [email protected] Recently, the leading position of telephone surveys as the major mode of data collection has been challenged. Telephone surveys suffer from a growing nonresponse, partly due to the general nonresponse trend for all surveys, partly due to changes in technology specifically influencing contactibility by phone. One way to counteract the increasing nonresponse is to increase the number of contacts. In mail and face-to-face surveys advance letters have been shown to be an effective extra contact and raise response. This inspired researchers to perform experiments with advance letters in telephone surveys. Like an advance letter, leaving a message on an answering machine can be seen as a precontact and opportunity to influence survey participation. Therefore, in this meta-analysis we present a quantitative summary of empirical studies on the effectiveness of two types of precontact on the response in telephone surveys. In the first part we summarize the effect of advance letters on response 237 in telephone surveys. In the second part we look into the effectiveness of a second type of precontact: leaving a message on answering machines. Keywords: telephone interview, advance letter, prenotification, (telephone) answering machine, response, refusal Part II: Challenges to Data Collection and Data Quality Interviewer Training: Responding to Changes in Telephone Survey Methodology Lisa R. Carley-Baxter, RTI International, USA, [email protected] Social and technological changes continue to affect telephone survey methodology by providing new challenges and improved efficiency. Telephone surveys are facing new challenges in terms of sampling, contacting, response rate, and the need for utilizing a more dynamic mixed mode approach to reach sample members. Technological changes have given rise to some of these challenges, including the increased use of cell phones and call screening devices (answering machines, privacy managers). In addition, declining response rates and advances in new technology as a means to contact/interview sample members have increased the utilization of mixed mode surveys for studies that were once only conducted by telephone. Social changes, such as the increased pace and individualization of society, the changing nature of our relationship with respondents (some argue moving away from social exchange towards economic exchange), the proliferation of telemarketing followed by the recent formation of state and national “do not call” lists, have also affected telephone survey methodology. Both social and technological changes, as well as the expansion of studies that were traditionally conducted only by telephone to include other modes, require modernization of interviewer training techniques. Social and technological changes have resulted in adapting interviewer training by changing the strategies used for avoiding and converting refusals, developing strategies for distinguishing survey research from telemarketing, and responding to the “do not call” list challenge. The addition of other modes to telephone studies, in particular the addition of web self-administered questionnaires, has created new challenges for training interviewers when a single instrument is used for both purposes. This resulted in expanding traditional telephone interviewer training, both in terms of administration guidelines and guidelines for responding to technical issues from sample members who are having difficulty completing the web instrument. This paper focuses on the changes in telephone survey methodology, how those changes impact interviewer training, and strategies for responding to those challenges in interviewer training. Keywords: interviewer training, telephone survey methodology, mixed mode An Experiment in Cognitive Training of Telephone Survey Interviewers Claire Durand, Marie-Eve Gagnon and Christine Doucet, Department of Sociology, Université de Montréal, [email protected] Telephone survey interviewers must communicate with people and convince them to cooperate in a context of having few cues about the characteristics of the household they are calling, and of having to adapt quickly to situations they can imagine only with difficulty. Training should therefore take this situation into account. This paper will present the results of an experiment to test our hypothesis that a better sense of self-efficacy improves interviewer performance. The experiment employed a training complement aimed at improving interviewers’ sense of self-efficacy by better informing them as to the reasons they are asked to do what they do and the situations they face. The setup was a telephone survey of alcohol consumption and drug addiction conducted by a private polling firm on behalf of Health Canada. The survey polled 14,000 respondents over more than three months, allowing for a multi-phase study. First, all interviewers completed a questionnaire on their attitudes, behaviours and characteristics (see Lemay & Durand, 2002; Hox & de Leeuw, 2002) after twenty hours of interviewing. Second, interviewer performance was monitored daily using the Weighted Unbiased 238 Performance Index (WUPI, see Durand, submitted) and analyzed using trajectory analysis (Nagin, 1999). This latter analysis allowed us to identify two groups of interviewers, one with greater difficulty in convincing respondents to cooperate than the other. However, since performance improves with time on the project, it was decided to include newly recruited interviewers in the training. The training was devised to be mostly cognitive and comprised two parts: a) information on sampling and bias in selection using a modified version of the M&M experiment (Auster, 2000), with a number of add-ons pertaining to the impact of refusals and of bad selections; and b) information on why people refuse or agree to cooperate (Goyder, 1987), together with successful strategies for taking control of the situation and dealing with refusals, as suggested by the best interviewers. The training complement was conducted after one-and-a-half month of fieldwork. It lasted one hour and was given to 18 interviewers, identified as either underachieving or newly recruited, divided into three groups. The paper will present the results of this experiment at both the qualitative and quantitative levels. It will first present the assessment of the training itself. Second, using trajectory and multilevel analyses, it will identify the impact of the training on performance. The paper will conclude with recommendations for the training of interviewers conducting telephone surveys. Keywords: telephone surveys, training, interviewers, performance, cooperation rate Sociolinguistic Parameters of Cross-cultural Variation in Telephone Surveys: Brian Kleiner1 and Yuling Pan2 1 Westat, USA, [email protected] 2 U.S. Census Bureau, [email protected] There has long been recognition that surveys, and telephone surveys in particular, constitute a specific type of linguistic interaction or discourse. A current and growing body of research is exploring the relationship between norms of language use and survey research. Participation in telephone survey interviews involves specific norms of language use that vary across cultures. For example, the actions of opening and closing interviews, initiating and shifting topics, asking and responding to questions, and repairing problems (among others) are all norm-driven components of survey interactions that may vary in important ways from one cultural/linguistic group to the next. It will be demonstrated that variation along these lines is not insignificant and may affect measurement error and data quality in crucial ways. Such variation in the norms of language use across cultures may conflict with the requirement of standardization (i.e., the principle that scripted survey questions should be administered consistently and word-for-word across respondents). However identical survey wording (including word-for-word translations) across cultural/linguistic groups may also lead to various kinds of measurement error. There is, therefore, a paradox uniquely involved in conducting international surveys: given variation in norms of language use, strict standardization may be more likely to lead to some kinds of measurement error than variable survey wording. We argue that the need for standardization must be balanced with the need to tailor surveys and questions to different cultural/linguistic groups with varying norms of language use. Thus, there is a need to identify the sociolinguistic parameters of cross-cultural variation in telephone surveys, and to document the specific cultural/linguistic norms that may impact data quality. This paper will outline these parameters and discuss ways in which discursive variation may be overcome in international telephone surveys by loosening the constraints of strict standardization where needed. We propose a program of study to determine the extent to which cross-cultural differences in norms for language use in telephone surveys affect data quality along the parameters outlined, in order to assess the areas in which constraints of standardization should be loosened. One result of this research program will be to document actual differences in telephone survey norms and language use that occur in particular cultures. Ultimately, the research program may provide international survey researchers with practical guidelines for avoiding language-related pitfalls associated with conducting surveys across cultural/linguistic borders. Keywords: Telephone surveys, cross-cultural variation, norms of language use, sociolinguistic parameters 239 Data Quality of Surveys of Minority-Majority Attitudes in Divided Societies: A Comparison between Telephone and Face-to-Face Surveys among Arabs in Israel Galit Gordoni, Department of Sociology and Anthropology, University of Haifa, Israel, [email protected] It is still an unanswered question in methodological literature whether telephone surveys amongst hard to reach populations, such as dissident minorities, and traditional societies, produce data, similar in quality, to face-to-face surveys. The paper will present comparisons of sensitive attitudes concerning coexistence, expressed by minority members, in three representative face-to-face, and telephone surveys in Israel. A model derived from the reasoned action theory serves as the main framework for testing the processes leading to the differences in data quality and survey response intentions between the modes. An initial test of the model is based on findings from a face-to-face representative survey of the Arab minority (N=688) conducted in 2002. In this survey, a coexistence index was validated using confirmatory factor analysis. The results support the usefulness of the reasoned action theory, in explaining survey response intentions and response error. The privacy and confidentiality concerns and attitudes towards surveys expressed by respondents were found to have a small, but unique effect, on the highly sensitive attitudes reported. Data analysis of the additional surveys will includes SEM (structural equation modeling) analysis, for testing two models: one for predicting attitudes toward coexistence and another for predicting participation intentions. Multiple group comparisons will be used for comparing scale reliability and factor loadings between the two modes and between high/low sensitive questions. Mode comparison of mean differences in the coexistence index questions will be performed, using ancova analysis. Analysis of the additional surveys: face-to-face (N=701) and telephone (N=500) surveys conducted in 2003 is still in progress. Keywords: non-response - theoretical approaches, data collection modes, data quality, marginalized groups, sensitive questions Can Multimode Approaches Salvage RDD for Public Health Surveillance? Michael W. Link, and Ali Mokdad, Centers for Disease Control and Prevention, Behavioral Surveillance Branch, [email protected], [email protected] Can alternative modes such as web and mail surveys extend the reach of random digit dial (RDD) surveys by encouraging participation across the general population? The decline in RDD response rates in recent years threatens the validity and reliability of their data. Researchers have looked at a variety of options to address this problem, one of which is the use of multiple modes, in particular web and more traditional mail surveys as complements to telephone data collection. Studies have shown that some sample members actually prefer certain modes over others. Yet, different modes have also been shown to produce varying results even when questions are asked of the same sample members. Thus, alternative modes may increase response rates, but can also increase measurement differences. As the largest ongoing, state-based RDD telephone survey, the Behavioral Risk Factor Surveillance System (BRFSS) is confronted with the issue of declining response rates and data reliability and validity. Use of multiple modes is one possible means of addressing this problem. For the BRFSS and other RDD surveys that collect similar data to continue to meet the public health data needs of local, state, and national researchers and policymakers, alternative means of collecting these data need to be examined to determine if the use of multiple modes of data collection can increase response rates and ensure data quality. This research provides the results of a set of experiments conducted in four states in the United States, testing the effectiveness of web and mail surveys when used in conjunction with telephone follow-up of nonrespondents as a means of increasing BRFSS survey participation. The research attempts to 240 answer several key questions regarding the use of mixed-mode for the BRFSS instead of a telephoneonly approach: Can alternative modes help to increase BRFSS response rates? What types of individuals in the general public respond to web and mail surveys? And, what effect, if any, does combining self-administered and interviewer-administered modes of data collection have on the survey estimates generated by the BRFSS? These experiments were conducted in parallel with the regular monthly BRFSS data collections in these states, providing a baseline for comparison with the experiment results. Keywords: health surveillance, random digit dialing, web surveys, mail surveys Using Interactive Voice Response Technology to Capture Census Data Sarah E. Brady, Robin A. Pennington and James B. Treat, U.S. Census Bureau, Washington, DC, USA, [email protected] In 2003, the United States Census Bureau conducted the 2003 National Census Test to examine the impact on self response of offering respondents the opportunity to provide their data using an Interactive Voice Response (IVR) application in addition to a paper questionnaire. The IVR application was an automated system that enabled respondents to complete their questionnaire by calling a toll free telephone number. The IVR application used voice recognition technology to capture the respondent’s answers to the census questions. For the IVR application there was the constraint that the question wording and flow needed to mirror the paper questionnaire. This paper presents how this constraint was addressed in the design and development of the IVR application. That is, the paper will present how questions were presented on the paper questionnaire and how this translated into the questions heard in the IVR application. The paper will discuss the level of self response for those who were offered a choice to respond by IVR and paper and those who were only offered paper. In addition, the paper will discuss how the IVR design affected data quality for the IVR application as measured by item nonresponse rates and errors in reporting demographics. Comparisons will be made between the item nonresponse rates for the paper questionnaire and the IVR application. Comparisons will also be made between the demographic characteristics for those responding using the paper questionnaire and those responding using the IVR application. The paper will also discuss system measures indicating where respondents experienced difficulty with the IVR application and measures of customer satisfaction with the system. Keywords: Self Response, Questionnaire Design, Item Nonresponse, Demographic Data, Customer Satisfaction SESSION: RESPONSE LATENCIES IN SURVEY RESEARCH Organizors: Volker Stocké1, Jochen Mayerl2 and Stasja Draisma3 1 University of Mannheim, D-68131 Mannheim, Germany, Tel: +49(0)621-181-3432, Fax: +49(0)621181-3451, [email protected] 2 University of Stuttgart, Institute for Social Sciences, Department of Sociology and Empirical Social Research, Keplerstr. 17 II, D-70174 Stuttgart, Germany, Tel: +49(0)711-121-3577, Fax: +49(0)711121-2768, [email protected] 3 Vrije Universiteit, Faculteit der Sociale Wetenschappen, Afdeling Methoden en Technieken, De Boelelaan 1081-c, 1081 HV Amsterdam, The Netherlands, [email protected] Keywords: Attitude strength, attitude stability, attitude-behavior consistency, information accessibility, mode of information processing, response effects, response latency. Abstract The time necessary to answer survey questions is an innovative and cost effective measure to enrich survey-data. Response latencies are an objective and unobtrusive operationalization for the cognitive availability of information addressed by the particular survey question. However, it is uncertain 241 whether and under which conditions they are better than self reported meta-cognitive measures. Response latencies are used in survey research to judge the quality of respondents’ answers in three respects: a) their susceptibility to response effects, b) the degree of attitude-strength and attitudebehavior consistency and c) the extent of temporal stability and persistence of attitudes. Unfortunately, the results on the response latencies’ predictive power in these areas are rather mixed. At least three factors may influence the validity of the latency-measure: a) measurement issues (e.g. coding, or not coding of invalid response times during the interview), b) multi-dimensionality (response latencies may capture any mixture of information accessibility, mode of information processing, calculations due to social desirability) and c) issues of preparing latency data for analysis (e.g. the way of transformation/standardization, outlier definition and treatment). In order to improve survey methodology, more research into the factors affecting the predictive validity of response times is needed. We would like to invite contributions relating to the topics addressed above, or to any other topics, which would improve the effective utilization of latency data in survey research. Response Latency as an Indicator of Optimizing. A Study Comparing Job Applicants and Job Incumbents’ Response Time on a Web Survey Mario Callegaro1, Yongwei Yang2, Dennison S. Bhola3 and Don A. Dillman4 1 University of Nebraska, 722 Y St., Lincoln, NE 68508-1165, USA, Tel. 1 402 304 3770, Fax. 1 402 458 2038, [email protected] 2 The Gallup Organization, USA, [email protected] 3 Psychometric Solutions, Inc., Lincoln, Nebraska, USA, [email protected] 4 Washington State University, Pullman, USA, [email protected] Latency measures can be used in survey research as a method of understanding the cognitive processes people employ in answering a question. There is a general consensus that a respondent performs four steps in answering a close-ended question: comprehending the question, retrieving from memory the information necessary to form an answer, judging the information retrieved, and selecting the best answer among the response options. Respondents who carefully employ all four steps are called optimizers. However, not every respondent is an optimizer, and a survey question can be answered by shifting response strategy using satisficing schemes. In weak satisficing, respondents still go through the four steps but in a less thorough manner. In strong satisficing, respondents skip some steps in order to quickly form a plausible or, sometimes, don’t know answer. Longer response times usually indicate more complex and engaged processing. Respondents who are highly motivated are more likely to become optimizers and thus spend more time responding. On the contrary, less motivated respondents tend to be satisficers and would spend less time. In this study we compared response latencies of job incumbents versus job applicants on a 133-item personality assessment administered via the web. Subjects responded to the statements using a 5-point Likert scale. Job incumbents were informed that their responses would be kept confidential. On the other hand, the job applicants were seeking employment and had higher stakes and likely higher motivation to complete the assessment. We hypothesized that such higher levels of motivation would cause the applicants to be more deliberate in responding. We also hypothesized that job applicants would be more likely to engage in impression management. In both cases, applicants (who had higher motivation) would be more likely to employ optimizing strategies. We hypothesized that this more engaged and deliberate response process would be indicated by longer response times. Our analysis supported this hypothesis. The median item response time of job applicants was consistently greater than that of the job incumbents. Our study adds further evidence to the satisficing paradigm from a different point of view. If satisficers had been identified using their level of education as a proxy to respondents’ ability, task difficulty can be rated with external judges and cognitive interviews, and motivation can be measured directly during the questionnaire, time latency is a seamless technique that can provide further evidence of the cognitive processes respondents employ in answering a questionnaire. Keywords: Response latency, Satisficing, Optimizing, Job Applicants, Job Incumbents, Web Surveys. 242 Question Wording Effects and Processes of Question Answering. A Different Way to Measure Response Latencies Bregje Holleman, Utrecht University, Utrecht Institute for Linguistics OTS, Trans 10, 3512 JK Utrecht, The Netherlands, Tel, +31 30 2536029, Fax: +31 30 2536000, [email protected] Response latencies provide a way to obtain insight in the cognitive processes underlying survey answers: they are an operationalisation for the difficulty to answer a particular survey question. It is more difficult to answer a question if the information asked for is not available, if conflicting information has to be integrated in order to come to a judgment, if the judgment cannot easily be translated easily to one of the answering options, etc. Usually, computer-assisted telephone interviews are conducted and response latencies are defined by measuring the time lapse between the off-set of the question and the on-set of an answer. Various frameworks have been proposed to describe the processes of answering attitude questions (e.g. Tourangeau, Rips and Rasinski 2000). Previous research (Holleman 1999) demonstrates that it is very revealing to distinguish only two stages; one stage of comprehension and retrieval (or attitude formation), and one stage of translating the judgment into the response options. If the research goal is to investigate the causes of response effects, this distinction into two stages is quite relevant. Response latencies can be related to these two stages if self-administered computer assisted questionnaires are used. Respondents can be asked to push a button to get a question visualized on the screen, and to push another button in order to get the answering categories. The time spent on the first stage of question answering (reading, attitude retrieval and judgment) then starts by the button push to get the question on screen and end when a button is pressed to get the answering options. The time spent on the second stage of question answering would start as soon as the answering options appear on the screen, and stop as soon as an answer is given. This paper presents an experiment in which response latencies were measured in this way. A questionnaire was constructed about ethical medical issues and administered to 90 respondents through a laptop. Reading and answering times were measured using the computer program E-prime. Based on the data obtained, it will be discussed how these reading and answering times are related to the stages of question answering and to each other. In the experiment, the question wording of some questions was manipulated in order to investigate the relation between a question wording effect (the well-known forbid/allow asymmetry) and the response latencies. It will be demonstrated that differences in question wording can be related to differences in answering times and not to differences in reading times. Also some preliminary results will be presented concerning the relation between reading times, answering times and various meta-attitudinal measures of attitude strength (such as elaboration and certainty). Keywords: response effects, wording effects, response latencies, survey methodology, question answering process, attitude strength Measuring Information Accessibility and Predicting Response Effects: The Relative Validity of Differently Transformed Response Latencies Volker Stocké, Project Program SFB 504, "Rationality Concepts, Decision Making and Economic Modeling", L13,15, University of Mannheim, D-68131 Mannheim, Germany, Tel: +49(0)621-1813432, Fax:+49(0)621-181-3451, [email protected] Survey respondents’ answers are frequently influenced by seemingly irrelevant features of the question wording, the presentation of response options and different aspects or the response context. One explanation for these response effects is the insufficient accessibility of the requested information in the respondents’ memory. Using different indicators for information accessibility, researcher have tested whether respondents’ susceptibility to response effects decreases when the cognitive 243 accessibility of the requested information increases. Some studies have utilized response latencies, which can be regarded as a innovative accessibility measure in survey research. However, the results about the predictive power of this indicator for response effects is rather mixed: in some studies response latencies proved to be a significant moderator variable, in others they do not have predictive power at all. One possible reason for these inconsistent results is the very heterogeneous way latency data is treated before utilized in data analysis. In order to reduce the characteristic skewness and the impact of outliers, response latencies are often transformed using different statistical functions. Furthermore, different ways of standardizing response latencies have been suggested, to control for measurement error caused by individually stable differences in the base line speed and differences in its variability. Whether these different treatments affects the response latencies’ validity for predicting response effects has not yet been studied systematically. In our study we compare the predictive power of raw response latencies for response effects with those of three differently transformed and three differently standardized versions of response times. The analyzed types of transformations are the logarithm, the square root and the reciprocal. Difference-, ratio- and z-scores are included into the study, representing different ways of standardization. The validation criterion is, how well the different measures predicts the effect of using either high- or lowfrequency response scales on the respondents’ reports about their daily television consumption. Data from a local random probability sample shows in a first step that raw response latencies are a significant predictor for whether different presentations of response options affects the answers: faster responses are found to be significantly less affected. Second, the different transformations are found to have neither a positive nor a negative effect on the response latencies’ predictive power. This is however the case for the standardized versions. Here, z-scores, computed using the mean and standard deviation of the individual respondents’ response times in the complete interview, proved to be the best predictor, which nearly doubles the raw response latencies’ ability to predict the analyzed type of response effects. Key words: Data transformation, information accessibility, response effects, response latencies Controlling the Baseline Speed of Respondents: An Empirical Evaluation of Data Treatment Methods of Response Latencies Jochen Mayerl, University of Stuttgart, Institute for Social Sciences, Department of Sociology and Empirical Social Research, Keplerstr. 17 II, D-70174 Stuttgart, Germany, Tel.: +49(0)711-121-3577, Fax: +49(0)711-121-2768, [email protected] In attitude theory, response latencies to attitude questions are regarded as a measure of chronic attitude accessibility, interpreted as an operative indicator of attitude strength. There is few but growing research in data treatment of response latency measurements to minimize systematic effects which can lead to a biased or wrong interpretation. Depending on the theoretical interest, several determinants of response latencies can be treated as bias effects (e.g. the individual baseline speed of respondents, effects of the measurement instrument, situational effects). Individual baseline speed reflects a constant individual characteristic of the mental speed of information processing. Although there are already some further developments for statistical data treatment methods to control baseline speed, these methods were not empirically evaluated and compared in a systematic way, yet. In the present study using response latency data of a nation-wide German survey (CATI), four statistical transformation methods to control the individual baseline speed are empirically evaluated and compared: Z-Score, Difference Score, Ratio Index, and RateAmount-Index. Additionally, these transformation indices are controlled for other systematic bias effects (question order, effect of extremity). Six evaluation criteria are used to compare the power of the different indices to control baseline speed: (1) correlations between the different response latency indices, (2) correlations of these operative indices with meta-judgemental indicators of attitude strength, (3) reduction of the covariance with baseline speed compared to “raw” reaction times, (4) power to moderate attitude-behavior relations (interpreting response latency as an indicator of attitude 244 strength), (5) power to moderate attitude-behavioral intention relations, and (6) power to moderate the persistence of an attitude. The empirical findings support the assumption of an increased data quality due to transforming “raw” reaction times into indices controlling the baseline speed (e.g. the indices’ moderating power of attitude-behavior-relations is significant compared to “raw” reaction times with non-significant moderating effects). Furthermore, the data quality increases if additional systematic bias effects are controlled (here: question order, effect of extremity). Comparing the different transformation methods according to the evaluation criteria, the Z-Score seems to be the best choice, followed by the rateamount index and difference score. It is recommended to use at least two different transformation indices to control empirical results of response latency applications. Keywords: response latency, baseline speed, data treatment, attitude strength, CATI Expressed Uncertainty in Question Answering: Response Latencies and (Para)Linguistic Expressions in Factual and Attitude Questions Stasja Draisma, Department of Social Research Methodology, Faculty of Social Sciences, Vrije universiteit, Amsterdam, [email protected], [email protected] Draisma and Dijkstra (2004) investigated the relationships between several indicators of respondent uncertainty and response error in a CATI survey. They found that the more difficult a survey question is, the more uncertain a respondent becomes, which is expressed in: 1. longer latencies between offset of the question posed and the onset of an adequate answer, 2. more verbal expressions of doubt (e.g. hedge words) such as “I think”, “I believe”, “I might”, 3. higher judgments of question difficulty as assessed by evaluations of individual questions by respondents in a debriefing interview, 4. more paralinguistic expressions of uncertainty, such as filled pauses (“ehm”), switches in chosen answer alternatives and number of words used in answers. The questions used in the Draisma and Dijkstra study were part of a so called validity study, in which individual true scores of respondents were known for simple binary factual questions. In this way, measurement error could be calculated, and the relations between the four above mentioned indicators and measurement errors studied. It was found that the best indicators for measurement error were verbal expressions of doubt (hedge words) followed by response latencies. In the present -follow up- study, these results are extended to attitude questions. In the same CATI study, some attitude questions were posed about ‘environmental issues’. Obviously, for these questions, no information concerning measurement error is available. Nevertheless, response latencies and verbal expressions of doubt may provide insight in the strength or accessibility of the expressed attitudes. The latencies for the attitude questions and the patterns of hedge words are compared to individual latencies and error patterns (within subjects) of other factual questions in the survey. Latencies and (para) linguistic expressions of the attitude questions are compared to factual questions as well (between subjects). Moreover, specific order effects and response latencies are considered: a subsample of respondents provided an estimated attitude of a reference group (“how important do you think environmental issues are for ‘an average Dutch person’) immediately before a question was posed about their own attitude (“how important do you yourself judge environmental issues”). Finally, effects of different transformation procedures on the latencies, such a z-score and logarithmic (see the abstracts of Stocké and Mayerl) will be investigated for the data from the validity study as well as for the attitude questions. Keywords: response latency, task uncertainty, measurement error, attitude strength, CATI 245 STREAM: COMPLEX ANALYSIS AND LONGITUDINAL DATA SESSION: CLASSIFICATION AND STRUCTURE: RECENT DEVELOPMENTS IN LATENT CLASS AND LATENT TRAJECTORY ANALYSIS AND NEW DEVELOPMENTS IN CLUSTER ANALYSIS Organizor: Johann Bacher1, Jost Reinecke2 and Christian Tarnai3 Contact: 1Lehrstuhl für Soziologie, Universität Erlangen-Nürnberg, Findelgasse 7-9, D-90402 Nürnberg, Germany, Tel: +49-(0)911-5302-679, Fax: +49-(0)911-5302-660, [email protected], www.soziologie.uni-erlangen.de 2 Empirische Sozialforschung, Abteilung Soziologie, Universitaet Trier, 54286 Trier, Tel: (+49) (0)651-201-2655/ Fax: (+49) (0)651-201-2645, [email protected] 3 Fakultät für Pädagogik, Universität der Bundeswehr München, Werner-Heisenberg-Weg 39, D-85577 Neubiberg, Tel: (089) 6004-3006,-2058/ Fax: (089) 6004-2057, [email protected] Keywords: cluster analysis, validation, data mining, meta cluster analysis, two-step-clustering, latent class analysis, latent trajectory analysis, mixture modeling Abstract Statistical tools for classification analyses has been widely developed in the last 15 years although they are hardly incorporated in statistical software packages. Beside cluster and correspondence analysis developments in latent class and latent trajectory models are discussed in the literature. Single computer programs like WINMIRA (von Davier 2001) and Latent Gold (Vermunt 2003) are used in various applications. Connections between classification and structural equation modelling have been proposed by Muthen (2001) and the program MPLUS (Muthen/Muthen 2002) under the label “Mixture Modeling”. Part I: Comparision and validiation of cluster solutions Methods for Comparing Latent Class Cluster Solutions Michael D. Larsen, Department of Statistics, Iowa State University, USA, [email protected] Latent class models for contingency table data are useful when the population is thought of as arising from distinct but unidentified subpopulations. In latent class analysis, the observed table is modeled as a composite of sub-tables each described by its own set of parameter values. Maximum likelihood estimates of latent class model parameters can be produced using the EM algorithm. The model estimates can be used to cluster the observations into latent classes. The U.S. National Science Foundation's SESTAT database is created from biennial surveys that are representative of U.S. scientists and engineers. The surveys provide detailed information about the employment, educational, and demographic characteristics of U.S. scientists and engineers. Several sets of qualitative questions focus on motivations for career decisions, work activities, and, for doctoral recipients, PhD preparation. Latent class analysis when applied separately to distinct sets of variables produces alternative cluster solutions of the survey respondents. Various methods can be used to describe the quality of clustering and for comparing the different cluster solutions produced for various sets of variables. Besides latent class analysis there are other strategies, such as monothetic analysis, for clustering categorical data. Methods of comparing cluster solutions will be compared across sets of variables and clustering algorithms. The work on this example is exploratory in nature, but aims to suggest principles for efforts of this type. Keywords: Adjusted Rand Index. Mixture Models. Model Selection. Number of classes. Stability of Clusters. Variable Selection. 246 L∞-consensus for Dendrograms : an Alternative to the Average Consensus Procedure Guy Cucumel, Département des sciences comptables, Ecole des sciences de la gestion, UQAM, CP 8888, Succ. Centre-Ville, Montréal (Québec), H3C 3P8, Canada, Tél.: (514) 987-3000 poste 4701, Fax : (514) 987-6629, [email protected] In the last decades many methods have been proposed to combine hierarchies (dendrograms) defined on a common set of n objects (Leclerc, 1998). Given a profile P = (H1, …, Hi, …, Hk) of k hierarchies, the average consensus (Lapointe and Cucumel, 1997) allows to obtain a consensus solution that is representative of the entire profile P by minimizing the sum of squared distances between the initial profile and the consensus (L2-consensus). This problem is equivalent to fit a least squares ultrametric to dissimilarity data and is NP-complete. It can only be solved by using a branch and bound algorithm. When the data set is large, one has to use heuristics to resolve it. As a result, the uniqueness and optimality of the solution is not certain. Moreover, the convergence toward a solution is sometimes difficult. Chepoi and Fichet (2000) have developed an algorithm that yields to a universal L∞consensus in a maximum of n2 steps. The two methods as well as a heuristic will be presented and compared on some numerical examples. Keywords: Meta Cluster Analysis, Average Consensus, L∞- consensus References Chepoi, V. and Fichet, B. (2000). “L∞-Approximation via Subdominants”. Journal of Mathematical Psychology, 44, 600-616. Lapointe, F.-J. and Cucumel, G. (1997). “The Average Consensus Procedure: Combination of Weighted Trees Containing Identical or Overlapping Sets of Objects”. Systematic Biology, 46, 306-312. Leclerc, B. (1998). “Consensus of Classifications : the Case of Trees”, In : A. Rizzi, M. Vichi and H.-H. Bock (Eds.) : Advances in Data Science and Classification, Springer-Verlag, Berlin, 81-90. Classifying Adolescents Using Value Orientations – a Comparison of Cluster Analysis and Latent Class Analysis Andreas Pöge, Universität Trier, FB IV – Soziologie, Universitätsring 15, 54286 Trier, Germany Westfälische Wilhelms-Universität Münster, Institut für Kriminalwissenschaften, Abteilung IV – Kriminologie, Bispinghof 24/25, 48143 Münster, Germany, Tel.: 0251-83-22749, Fax: 0251-8322376, [email protected] Our empirical data is part of a continuing criminological and sociological panel study of deviant and delinquent behaviour of adolescents. With the help of questionnaires, pupils in Duisburg and Münster are annually interviewed. The sample, which forms the basis of the given analyses, consists of data of Year 9 pupils in Duisburg. By means of LCA (Rost, 1997) and of cluster analyses (Bacher, 1996), classifications of adolescents are presented in view of value orientations. These classifictaions have been conducted with the help of the EDP-programmes ALMO (Holm, 1997), WinMira and LatentGold (Vermunt, Magidson, 2000). They are further examined for patterns of delinquent behaviour. Moreover, differences and the comparability of the results of the different classificatory methods shall be paid attention and discussed. The focus will be on the comparison of latent class models and cluster analysis. References BACHER, Johann: Clusteranalyse, Anwendungsorientierte Einführung, München, Wien 1996. HOLM, Kurt: ALMO Statistik System, Linz 1997. ROST, Jürgen: Lehrbuch Testtheorie, Testkonstruktion, Bern u. a. 1997. VERMUNT, Jeroen K.; MAGIDSON, Jay: LatentGOLD Users’s Guide, Belmont 2000. 247 Detecting Aberrant Patterns of Change for the Dimensions of a “Quality of Life” Questionnaire by means of Linear Mixed Models Herbert Matschinger, University of Leipzig, Department of Psychiatry, Johannisallee 20/1, D- 04317 Leipzig, Germany, Tel: ++49 341 97 24533/ Fax: ++49 341 97 24539, [email protected] In the framework of longitudinal studies quite often only very little change is observed for a dependent variable. If models for linear or higher order trajectories are employed the parameters are characterized by big standard errors which seem to forbid a further explanation of the change by individual characteristics of the observations. Nevertheless it must be assumed that the sample is heterogeneous with respect to a change over time, which results in remarkable variance of the trajectory parameters if the observations are treated as 1st level and the individuals as 2nd level. Unfortunately in many instances the individual characteristics responsible for this heterogeneity are not known in advance, so a latent class analysis seems to be adequate to investigate the latent heterogeneity. The aim of this investigation is to identify subgroups of observations which can be characterized by very particular patterns of change, differing considerably from the rest of the sample. The division of the sample in latent classes of observations with different forms of trajectories then facilitates the interpretation of the heterogeneity observed and provides the necessary information for further investigation. Data are used from a cost-effectiveness study on 307 psychiatric patients, which had been observed 5 times over the range of 2½ years. The endogenous variables of interest are the 8 dimensions of the socalled SF36 questionnaire measuring different aspect of quality of life. The SF36 is a multi-purpose, short-form health survey with only 36 questions. It yields an 8-scale profile of functional health and well-being scores as well as psychometrically based physical and mental health summary measures. It is a generic measure that does not target a specific population (Ware, Snow and Kosinski 2000). It is assumed that these variables are subject to change in the course of the clinical treatment. In a first step the intercept, the linear and quadric terms and their respective variances are estimated by means of a model developed by Muthén and others (Muthén 1997; Muthén and Curran 1997; Muthén and Khoo 1998). The parameters of the model are treated as random effect growth factors. In a second step the variance of the trajectory is decomposed in several latent classes of observation in order to explain the latent heterogeneity. Results clearly show, that only a few people comprised in one latent class are characterized by a change of certain aspects of their quality of life. Additionally it is shown that merely looking on the parsimony of a solution will be misleading, since then the aberrant classes will not be detected. References Muthén, B.O. (1997). Latent Variable Modeling of Longitudinal and Multilevel Data. In A. E. Raftery (Ed.), Sociological Methodology. (pp. 453-480). Boston: Blackwell. Muthén, B.O. and Curran, P.J. (1997). General Longitudinal Modeling of Individual Differences in Experimental Designs: A Latent Variable Framework for Analysis and Power Estimation. Psychological Methods, 2(4), 371-402. Muthén, B.O. and Khoo, S.-T. (1998). Longitudinal Studies of Achievement Growth Using Latent Variable Modeling. Learning and individual differences, 10(2), 73-101. Ware, J.E., Snow, K.K. and Kosinski, M. (2000). SF-36 Health Survey: Manual and Interpretation Guide. Lincoln, RI: QualityMetric Inc. Part II: Software, pracitical requirements and empirical validation SPSS TwoStep Clustering – A First Evaluation Johann Bacher, Knut Wenzig and Melanie Vogler, Lehrstuhl für Soziologie, Universität ErlangenNürnberg, Findelgasse 7-9, D-90402 Nürnberg, Germnay, Tel: +49-(0)911-5302-679/Fax: +49(0)911-5302-660, [email protected], www.soziologie.uni-erlangen.de SPSS Version 11.5 and later versions (SPSS 2002, Chiu et al. 2001) offer a new cluster analysis feature, called TwoStep-Cluster. TwoStep-Cluster was developed to cluster large data sets. The procedure consists of two steps. The first step computes a new data matrix, that contains cluster centres as rows. Cluster centres are defined as dense regions in a multivariate space. In the second 248 step, a probabilistic hierarchical cluster analysis is performed. The procedure allows to handle mixed type attributes and offers test statistics to determine the number of clusters. These two characteristics are attractive for the social sciences. However, TwoStep is rarely used in the social sciences. The presentation analyses the question whether it is useful to apply TwoStep in the social sciences. TwoStep will be compared with LatentGold (Vermunt/Magidson 2000). Examples from social stratification research will be used. Keywords: TwoStep-Cluster, evaluation, mixed type attributes, model based clustering References Chiu, T., et al. 2001: A Robust and Scalable Clustering Algorithm for Mixed Type Attributes in Large Database Environment. Proceedings of the 7th ACM SIGKDDD International Conference on Knowledge Discovery and Data Mining 2001, S. 263-268. Vermunt, J. K., Magidson, J., 2000: Latent Gold. User's Guide. Belmont. SPSS Inc., 2002: SPSS TwoStep Cluster Component. (www.spssetd.com, 04.02.2004) Clustering Messy Social Data with ClustanGraphics David Wishart, Department of Management, University of St. Andrews, St. Katharine's West, The Scores, St. Andrews KY16 9AL, Scotland, [email protected], www.clustan.com Clustering with messy data regularly occurs in the behavioural sciences, for example in social surveys. We discuss the practical issues involved in designing cluster analysis software for such data in ClustanGraphics. k-means analysis is extended to handle mixed variable types, differential case and variable weighting, and missing values, suitable for use with very large datasets such as arise in data mining and survey analysis. Unlike other popular software, an exact assignment test guarantees that the procedure will converge. The detection of outliers and mapping of complex cluster shapes by tessellations of spherical clusters are unique features, and a summary tree is obtained for the resulting k-cluster model. The talk will be illustrated by a demonstration of ClustanGraphics7. Cluster Analysis Software: Proposals of Tools for Validation Michael Wiedenbeck and Cornelia Zuell, ZUMA, P.O. Box 12 21 55, D-68072 Mannheim, Germany, Tel: +49 6211246147, Fax: +49 6211246100, [email protected], [email protected], http://www.gesis.org/zuma Application of cluster analysis software as it is implemented in packages like SPSS implies two obstacles for the ordinary user: large data sets and the lack of decision rules which support the choice of an outcome as an appropriate solution of a classification problem. The last problem makes classifications sometimes rather arbitrary unless the findings have a plausible interpretation. This problem can be overcome in case of small data set by means of theoretical reasoning. It gets often worse when using large data sets and the software is designed to the analysis of small sets of data. We will discuss these problems considering typical examples, show how far the problems are approached by SPSS and ClustanGraphics by tests and preprocessing the data, and specify how tools for validation of clusters should be designed which, for example, indicate compactness and separation of clusters together with a substantive description. Empirical Validation of “Whisky Classified” David Wishart, Department of Management, University of St. Andrews, St. Katharine's West, The Scores, St. Andrews KY16 9AL, Scotland, [email protected], www.clustan.com A thousand tasting notes were reviewed for 86 single malt whiskies, from which a vocabulary of 500 aromatic and taste descriptors was compiled and a standardised flavour profile of 12 sensory features developed. The principal malt whiskies of all the Scottish distilleries were then profiled and classified 249 into ten groups according to flavour, and 3 industry surveys were conducted to test the new classification which is now fully endorsed by the producers. “Whisky Classified: Choosing Single Malts by Flavour” was the first book to classify single malt whiskies by flavour. It is a consumer–friendly guide that takes the confusion and guesswork out of whisky–buying, and aims to help the novice or present–buyer make the right purchase of Scotland’s national drink. The talk concludes with an empirical validation of Aberlour, Bruichladdich, Glen Grant, Glenmorangie, Laphroaig and Macallan. SESSION: MULTILEVEL ANALYSIS Organizor: Tom A.B. Snijders Contact: Department of Sociology / ICS, Grote Rozenstraat 31, 9712 TG Groningen, The Netherlands, [email protected] Keywords: multilevel analysis, random coefficients, hierarchical linear modeling. Abstract Multi-level analysis, also called hierarchical (non)-linear modeling, has provided a set of methods for data that have a natural hierarchical structure, and these methods now are routinely used in the analysis of such data. A frequently occuring nesting structure is individuals nested within groups, but these methods have been applied also, e.g., to longitudinal and panel data, growth curve modeling, and meta-analysis. The model has been extended to cover dichotomous and count data, multivariate outcomes, and data structures that include crossed as well as nested factors. Recently much work has been done on non-linear models; Markov Chain Monte Carlo algorithms have been instrumental to extending the scope of multilevel modeling to data structures that previously were too complicated. For this session, papers treating new statistical developments are welcome, but also review papers, as well as papers about applications that pose special methodological problems because of the complicated data structure, interpretations, the choice of a well-fitting model, etc. Frequentist MCMC Estimation Methods for Multilevel Logistic Regression Carlos Coimbra and Tom A.B. Snijders, Department of Sociology, University of Groningen, Grote Rozenstraat 31, 9712 TG Groningen, The Netherlands, [email protected], [email protected] Various estimation procedures are available for multilevel models with non-normally distributed outcome variables. Maximum likelihood estimators can be implemented using numerical integration, but such algorithms do not always work perfectly. Other frequentist methods are in use that are based on approximations, but these approximations are known to lead sometimes to biased estimators. Good Bayesian Markov chain Monte Carlo (MCMC) methods have been developed, but the Bayesian paradigm is not always the first choice of approach. Thus, there still is discussion about what is the most appropriate estimation procedure for these models. In this presentation, a MCMC algorithm for calculating maximum likelihood estimators will be presented, which can be used in multilevel models for non-normal outcome variables, i.e., hierarchical generalized linear models. The algorithm is based on stochastic approximation (generalizations of the Robbins-Monro algorithm) for solving the likelihood equation. The principle of the algorithm will be explained and experience with its use in various examples will be presented. Marginal maximum likelihood estimation in multilevel item response models is an example. Keywords: Hierarchical Non-linear Models, MCMC estimation, Maximum Likelihood, Stochastic Approximation, Item Response Theory. 250 The Use of Internal Pilot Studies to Derive Powerful and Cost-Efficient Designs for Studies with Nested Data M. Moerbeek, Department of Methodology and Statistics, Utrecht University, The Netherlands, [email protected] Many data in social science research have a nested structure with persons nested within clusters. Examples are children nested within schools, members nested within households and patients nested within therapy groups. To design a cost-efficient study to test the effect of an independent variable on the outcome at a desired power level, the optimal number of clusters and the optimal cluster size need to be calculated. Optimal sample size formulae can be shown to depend on the intra-class correlation coefficient (ICC). An initial estimate of the ICC may be obtained from similar studies in the past or from subject-matter knowledge. An incorrect estimate, however, may result in an under- or overpowered study. The use of an internal pilot is one of the approaches that has been proposed to avoid designs that are too expensive or have too low power. Designs with internal pilots are wellknown in the biomedical sciences, while hardly any attention has been paid to them in the social science literature. The basic idea is to calculate the optimal sample sizes based on an initial estimate of the ICC. Then a pre-selected proportion of the optimal number of clusters is sampled and a sub-sample of persons equal to the optimal cluster size is taken from each cluster. This is the internal pilot or the first stage of the study. The ICC is estimated based on the data collected until this interim time point and the optimal sample sizes for the second stage of the study are calculated based on the estimated ICC. After having collected the data in the second stage, the analysis is done using data from both stages. Since data of the pilot are also used in the final analysis, it is called an internal pilot. This procedure is known to have an inflated type I error rate since the total sample size of the study depends on the estimate of the ICC at the interim time point. The aim of this presentation is to explore the usefulness of internal pilots for studies in the social sciences with nested data. Designs with different interim time points are compared on basis of their type I error rates, power levels and median total costs. The comparison is done using a simulation study with various degrees of bias in the initial estimate of the ICC. Keywords: intra-class correlation, re-estimation, multilevel model, interim time point Performance of Likelihood-Based Estimation Methods for Multilevel Binary Regression Models Marc Callens and Christophe [email protected] Croux, Katholieke Universiteit Leuven, Belgium, By means of a fractional factorial simulation experiment, we compare the performance of Penalised Quasi-Likelihood, Non-Adaptive Gaussian Quadrature and Adaptive Gaussian Quadrature in estimating parameters for multi-level logistic regression models. The comparison is done in terms of bias, mean squared error, numerical convergence, and computational efficiency. It turns out that, in terms of Mean Squared Error, standard versions of the Quadrature methods perform relatively poor in comparison with Penalized Quasi-Likelihood. Keywords: Binary Regression, Fractional Factorial Experiment, Gaussian Quadra-ture, Monte Carlo Simulation, Multilevel Analysis, Penalised Quasi-Likelihood Outliers and Multilevel Models John F. Bell and Eva Malacova, University of Cambridge Local Examinations Syndicate, UK, [email protected] This paper will consider some problems relating to outliers in multilevel models drawing on research into various data sets including progress in secondary and higher education and university admissions. 251 In particular, it will be demonstrated that when outliers are generated by a contaminating process then they can lead to unnecessarily complex models. In these cases, it is necessary to carefully consider the nature of the data and the processes that are generating. For example, do variables measure the same thing for different sub-populations in the data set. In one example, it will be demonstrated that such process leads to an extremely complex model that may completely mask the true relationship. The paper will also consider the importance of issues such as reproducibility and the role of subjectivity and background knowledge in dealing with outliers. Keywords: multilevel models, outliers, educational data On the Relative Efficiency of Unequal Cluster Sizes in Multilevel Intervention Studies L. Kotova, G.J.P. van Breukelen, M.J.J.M. Candel and M.P.F. Berger, Department of Methodology and Statistics, Faculty of Health Sciences, University of Maastricht, The Netherlands, [email protected] In this paper, optimality of equal cluster sizes and loss of efficiency for unequal cluster sizes in multilevel intervention studies are examined. A few theorems concerning the D-optimality of equal cluster sizes when the randomisation is done at the cluster or person level are proven. The relative efficiency of designs with unequal cluster sizes is computed for different values of the intraclass correlation coefficient and a variety of frequency distributions of the cluster size. It is concluded that although designs with equal cluster sizes are preferred above designs with unequal cluster sizes for their optimality properties, the actual loss of efficiency by using unequal cluster sizes is not so large and can be compensated by increasing the number of clusters to a small extent. Keywords: Multilevel intervention studies; Varying cluster sizes; D-optimality; Relative efficiency. Multilevel Multiprocess Modelling of Partnership Transitions and Fertility in Britain Fiona Steele, Constantinos Kallis, Harvey Goldstein and Heather Joshi, Bedford Group for Lifecourse and Statistical Studies, Institute of Education, University of London, 20 Bedford Way, London WC1H 0AL, UK, [email protected] In this paper, we describe a general framework for the analysis of correlated event histories. The research is motivated by a substantive research question: the relationship between partnership transitions and fertility among British women. Previous research has treated outcomes of the fertility process (e.g. age and number of children) as covariates in models of partnership dissolution and transitions from (unmarried) cohabitation to marriage (e.g. Berrington and Diamond, 1999). However, the causal direction of the relationship between the processes of partnership transitions and fertility is complex. Research in the USA has found that marital stability (as measured by the risk of divorce) may also influence decisions about childbearing (Lillard, 1993), and it is likely that the stability of a cohabiting partnership would have a similar effect on childbearing decisions. In other words, partnership transitions and childbearing may be jointly determined; the two processes may be influenced by a common set of factors, some of which are likely to be unobserved. In this paper, we describe an extension of Lillard’s approach in which marital dissolution is modelled jointly with transitions from cohabitation using a multilevel multistate competing risks model (Steele et al., 2004). A multilevel model is used to allow for repeated partnerships for the same woman, while a multistate model is used to distinguish between two partnership states: marriage and cohabitation. Competing risks arise when considering transitions from the cohabitation state, since a period of cohabitation may end in separation or marriage to the same partner. Using a simultaneous equation model, partnership transitions and fertility are modelled jointly. By modelling the two processes simultaneously, we allow for the possibility that partnership transitions may depend on outcomes of the fertility process and partnership stability (as measured by the hazard of a partnership transition) might affect fertility decisions. 252 The research is based on data from the National Child Development Study, a longitudinal study of a cohort of British men and women born in 1958. Keywords: multilevel modelling, event history analysis, simultaneous equation model, multiprocess model, partnership transitions, childbearing References Berrington, A. and Diamond, I. (1999). “Marital dissolution among the 1958 British birth cohort: The role of cohabitation.” Population Studies, 53: 19-38. Lillard, L.A. (1993) “Simultaneous equations for hazards: Marriage duration and fertility timing.” Journal of Econometrics, 56: 189-217. Steele, F., Goldstein, H. and Browne, W. (2004) “A general multilevel multistate competing risks model for event history data, with an application to a study of contraceptive use dynamics.” Journal of Statistical Modelling , 4 (to appear). The Concept of ‘Social Level’ and how to assess it Pieter van den Eeden, Department of social research methodology, Vrije Universiteit, Amsterdam, The Netherlands, [email protected] The concept of “social level” has been become more and more important in social investigation and theory formation in social sciences. The assessment of levels seems to be without any doubt in cases where various units of investigation of a same class are so clearly distinguished to be assumed independent from each other, and where units of different classes appear to be susceptible for multistage sampling. A well-known example of clear multiple social levels is: students within classes within schools. However, several serious problems can be met in practical investigation. In reality independent comparable units are seldom met, and in social research multi-stage sampling ask mostly hard decisions. And, is not multilevel methodology in contrast to classic sociological notions about the interconnectedness of comparable social units? How to think about ‘relative autonomy’ of levels? Can ‘social levels’ be viewed as classes of ‘social units’ that are mutually related in a specific way? What is the exact nature of this relatedness? Starting from these questions, it seems to be worthwhile to define carefully the concept of ‘social level’ and to develop instrumental procedures of its empirical assessment. Specifically, the paper pretends to show that the assessment of social levels needs highly qualified measurement of level-specific variables and a careful analysis of their interrelations. In connection to this, the paper offers an overview of types of general interpretations. This will lead to the conclusion that, although a priori notions of social levels indeed can ake sense for establishing multilevel relations, also these notions needs a posteriori empirical reference. Keywords: multilevel analysis measurement concept-formation The Hausman Test of Random Effects Specifications A. Fielding, University of Birmingham, UK, A. [email protected] In the econometrics literature for panel data modelling the Hausman test statistic is routinely reported as a diagnostic for assessing whether panel unit effect should be treated as fixed or random. The multilevel setting is repeated measures within panel units. It has been suggested that in more general multilevel modeling , this is one way of deciding whether higher level effects should be treated as random or fixed. Of course this decision is often made on substantive grounds and apart from the desire to generalise beyond the specific data units, there are design questions that should influence this decision. The paper will examine the Hausman test and show that it is really a specification test of whether proposed random effects are correlated with covariates, thus making standard estimation procedures poor. The decision to use fixed effects is then really a pragmatic one that avoids this problem due to conditioning of the effects on other effects. By simulation of such random effects models the paper will examine why routine use of the procedure may cause us to reject a random effects approach contrary to true situations. We will also explore an example where substantive 253 conclusions may be flawed by uncessary use of fixed effects. We propose alternative estimation procedures using Conditional Iterative Generalised Least Squares, MCMC , and Instrumental variables which directly deal with random effects correlated with covariates. We also make suggestions as to how the real question of deciding between fixed and random effects. The paper proposes that this is a real unsolved question in multlevel analysis which is ripe for advanced research. The Problem of Time Dependent Explanatory Variables at the Context Level in Discrete Time Multilevel Event History Analysis. A Comparison of Models Considering Mobility Between Local Labour Markets as an Example Michael Windzio, EMPAS, University of Bremen, Germany, [email protected] This paper deals with the question of how to include time dependent explanatory variables at the context level in multilevel event history models. In general, context level variables in multilevel models are assumed to be constant over time. Only time constant context level variables can perform the task of reducing error variance at the context level. Thus, a proposal is made to extend the analysis to a three level model, where time periods of persons form the units at level one, historical time periods form the level two, and the spatial units form the level three on their own part – within which, in turn, the units of level two are clustered. In this paper it will be argued that the complex character of contexts in the analysis of processes embedded in spatial units is only taken into account by the three-level model. If characteristics of spatial context units change over time then the context won’t be the spatial unit itself, but rather the situation within a particular spatial unit during a certain (historical) period of time. Consequently, the social structure differs not only spatially but also temporarily. For this reason, time periods which are part of spatial units should be regarded as the correct context level units. But this implies a problem, namely the violation of the assumption that units of level two are independent. Thus, in order to satisfy this assumption, spatial units are considered as the third level whereby the context level time units are clustered within the context level spatial units. Considering the event of mobility between local labour markets as an example, four different ways of modelling time varying context level explanatory variables are compared. The result is that the proposed three-level model leads to the most conservative findings. In other words, the three-level model leads to the highest upward correction of standard errors which is desired if data is clustered. Keywords: multilevel analysis, event history analysis, logistic regression, spatial mobility SESSION: MEASUREMENT (IN)VARIANCE IN LONGITUDINAL RESEARCH Organizers: Han Oud1 and Reinoud Stoel2 Contact: 1Nijmegen University,the Netherlands, [email protected] 2 Vrije Universiteit Amsterdam, The Netherlands, [email protected] Keywords: measurement invariance, factorial invariance, longitudinal research Abstract Measurement invariance means that measurement across occasions originate from the same measurement scale. The important aspect of measurement (in)variance is whether, under different conditions of observing and studying phenomena, measures of the same attribute are obtained. Although measurement invariance plays a role in cross-sectional studies, it becomes especially important when a study adopts a longitudinal design in which, by definition, the same measurement instruments are applied at more than one occasion. Violation of the assumption of measurement invariance may hinder the assessment of change, because change will be confounded with change in 254 the meaning of the construct across time. Within the context of the common factor model measurement invariance is often investigated under the name of factorial invariance. The requirement of factorial invariance, and measurement invariance in general, is highly likely to be violated in its strict forms in practice. In this session the focus will be on issues regarding measurement (in)variance in longitudinal studies. Contributions may focus on the conceptual comparison of degrees of measurement invariance, from strict factorial invariance to partial measurement invariance. We also welcome more applied contributions, using and assessing different degrees of longitudinal measurement invariance. Practical Issues in the Assessment of Measurement Invariance and why Invariance does not Imply Measurement Reinoud D. Stoel, Department of Biological psychology Vrije Universiteit Amsterdam, NL-1081 BT Amsterdam, The Netherlands, [email protected] Measurement invariance, or factorial invariance, across time means that the numerical values across measurement occasions of a given concept are obtained from the same measurement scale. In practice, measurement invariance is likely to be violated, and researchers have been investigating whether the requirements of strict factorial variance can be relaxed. In the first part of this paper it will be demonstrated that measurement variance, mostly referred to as partial measurement invariance, may present some problems in practice if the latent variable structure is scaled using a reference indicator. Pros and cons of identifying the model by means of restrictions on the measurement part of the model will be discussed. In the second part of the paper it will be argued that a pure confirmatory approach to the assessment of measurement invariance may sometimes be problematic. It may well occur that a common factor model under full measurement invariance provides a good fit to the data, while the same model without measurement invariance does not fit. This intuitively contradictory issue can be solved by acknowledging the possibility that a certain factor model may be wrong, but consistently wrong across occasions or groups (i.e. while the model does not fit the data, its parameters may be invariant). This issue will be illustrated using a simulated dataset, and its implications will be discussed. Keywords: measurement invariance, factorial invariance, partial measurement invariance Measurement (In)Variance in SEM State Space Modelling of Panel Data Johan H.L. Oud, Institute of Family and Child Care Studies, University of Nijmegen, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands, [email protected] The topic of measurement variance and invariance became popular in latent growth curve modeling. First SEM state space modelling of panel data (Oud, 2004) will be compared to the latent growth curve model. In contrast to the descriptive nature of the latent growth curve model, having time as the independent “explanatory” variable, the state space model has a Laplacian causal starting point: it explains the current state of a system in terms of its previous state and the input over the intervening period. Nevertheless, latent growth curve modelling is easily incorporated into the measurement part of the state space model as shown by several authors over the last 25 years (Goodrich & Caines in 1979; Jones in 1993; Shumway & Stoffer in 2000). However, the random subject effects and parameters of the latent growth curve model, defined by means of basis functions in the measurement part of the model, cause serious interpretation problems, especially as a result of their strong dependence on the time scale origin (zero point). This is avoided in a state-trait model. Traits are also random subject effects, but defined as dynamic components in the latent structural part of the state space model in the form of unobserved heterogeneity between subjects. It is argued that measurement invariance should preferably induce time invariance for the overall model, meaning that its parameter 255 values do not depend on the time scale origin but only on the time interval, and are scale free in the sense of simple transformations between different time scale units. Keywords: measurement invariance, SEM state space modeling, time scale origin, unobserved heterogeneity References Oud, J. H. L. (2004). SEM state space modeling of panel data in discrete and continuous time and its relationship to traditional state space modeling. In K. van Montfort, J. Oud & A. Satorra (Eds.), Recent developments on structural equations models: Theory and applications (pp. 13-40). Dordrecht: Kluwer Academic Publishers. Types of Change in Self-Report Data: Definition, Interpretation and Operationalization Frans J. Oort, Medical Psychology (J4-410), AMC/University of Amsterdam, Meibergdreef 15, 1105 AZ Amsterdam, The Netherlands, [email protected] The assessment of change in self-reported attributes is hindered by the fact that there are different types of change, among which different types of response shift. Different typologies of change have been introduced, in different fields of research. Here we consider a psychometric approach to assessing change, based on factor analysis and item response theory. This approach results in a new change typology, with clear definitions, interpretations, and operationalizations. If it turns out possible to satisfactorily link the new psychometric change typology to existing response shift typologies, then structural equation modeling can be used for the detection of response shifts. Here we submit a proposal for such a linkage, and offer a procedure for the detection of recalibration, reprioritization, and reconceptualization response shifts and for the measurement of “true” change. Merits and demerits of this four-step procedure are discussed. Keywords: measurement of change, response shift, structural equation modeling Impact of Direction of Answering Categories on Response Behavior Dagmar Krebs, Institut fuer Soziologie, Universitaet Giessen, Karl-Gloeckner-Str. 21 E, D-35394 Giessen, [email protected] A response scale offering answering categories is usually directed from left to right. Does it, however, make a difference in response behavior whether the scale starts on the left side with the positive answering category or the negative answering category? This question is addressed from different perspectives. First, response behavior is observed in a split-ballot design. Respondents (students) were asked to evaluate different job characteristics of a later job with respect to their subjective importance (job motivation). Judgements were given on an 8-point scale offering increasing (split 1) or decreasing (split 2) importance to the respondent. Only the extreme points of the 8-point scale were labeled. Thus, for one group of respondents answering categories were labeled as “not at all important” to “very important” (split 1), for the other group answering categories ranged from “very important” to “not at all important” (split 2). It is postulated that direction of answering categories does make a difference depending on the direction of reading from left to right. Thus, if answering categories range from “not at all important” to “very important”, the proportion of negative answers is expected to be higher than under the condition where answering categories range from “very important” to “not at all important” where proportions of positive answers are expected to be higher. Based on the split ballot experiment this hypothesis is tested. Second, measurement of job motivation using the above mentioned two scales was repeated in a panel design. Thus, it is possible to observe if there are individual changes in response behavior depending on the reversing of answering categories. The advantage of this second step is that measurement stability can be observed. In this part the hypothesis of (in)variance of measurement will be tested. Additionally, a method effect can be introduced (as a latent variable) into the model and its effect can be assessed. 256 Third, attention is given to the dimensional stability of indicators. If response behavior changes with the direction of answering categories, as supposed in point one above, it is necessary to check the effect this difference might have on the dimensional structure of the indicators. Measured with one type of scale, indicators of job motivation are ordered along the dimensions of “intrinsic motivation”, “extrinsic motivation” and “leisure motivation”. In this part the hypothesis of dimensional stability across measurement with the above described different scales is tested. Keywords: auto-regression, measurement stability / invariance, panel. Testing Measurement Invariance with Respect to Time, Group, and Latent Gitta Lubke, Department of Psychiatry/VIPBG, Virginia Commonwealth University 800 East Leigh St. Suite 1-132, Richmond VA 23219-1534, USA University of Notre Dame, USA, [email protected] A measurement model relates observed scores (e.g., item or subscale scores) to one or more underlying latent variables. Measurement invariance is defined with respect to a selection variable, and refers to the invariance of a measurement model across selected groups. If the selection variable is time, the selected groups correspond to different time points. If the measurement model is invariant, groups can be compared with respect to the latent variables. Absence of measurement invariance indicates that the observed scores measure a different set of constructs across selected groups. Measurement invariance can be tested by comparing measurement models that are restricted to be equal across the selected groups to measurement models that are less restrictive. The measurement model in a univariate linear growth model relates a single observed score measured at several time points to an underlying intercept factor and a slope factor. Due to inherent model restrictions only partial tests of measurement invariance can be carried out. These tests may reveal whether differences in intercepts and variances of the observed variable across time points can be explained by variation in intercept and slope of the individual growth curves. It is argued that full tests of measurement invariance, which may reveal whether the same theoretical construct is measured across time, can only be conducted using multivariate data collected at each time point. Possible confoundings in the univariate case are illustrated. The linear growth model can be used simultaneously in several groups. If the additional selection variable is observed (e.g., gender), tests of measurement invariance can be easily conducted across time as well as across the additional selection variable. In case of invariance, average growth can be compared across the observed groups. In growth mixture models, the additional selection variable is a latent class variable. It is shown that simultaneous tests of measurement invariance across time and latent class are complicated by the fact that the number of classes and model restrictions within class are interdependent. Keywords: measurement invariance, multivariate analysis, linear growth model, growth mixture model Individual Differences in Factor Loadings and the Fit of the Standard Factor Model: Consequences for Quantitative Behavioral Genetics Research Henk Kelderman1 and Peter C. M. Molenaar2 1 Vrije Universiteit Amsterdam, Faculteit der Psychologie en Pedagogiek, Afdeling Arbeids- en Organisatiepsychologie, Amsterdam,The Netherlands, [email protected] 2 University of Amsterdam, The Netherlands In quantitative behavioral genetics, there are pervasive substantive arguments in favor of individual differences in factor loadings. As a result factor loadings are not invariant over subjects. In this paper, it is shown that, under liberal assumptions, data from a heterogeneous 1-factor model yield the same population covariance structure as the corresponding standard 1-factor model. Furthermore, it is shown that the sample covariance matrices and the standard goodness of fit statistics are the asymptotically same. A small simulation study yields similar results for realistic sample sizes. The 257 consequences of the results for quantitative behavioral genetics research is discussed. It is shown that a seemingly measurement invariant factor model in the population may lack measurement invariance on the individual level. Keywords: quantitative behavioral genetics, factor loadings, individual differences, invariance over subjects SESSION: METHODS OF SOCIAL NETWORK ANALYSIS Organizors: Peter J. Carrington1 and Anuška Ferligoj2 Contact: 1Department of Sociology, University of Waterloo, Canada, [email protected] 2 Faculty of Social Sciences, University of Ljubljana, Slovenia, [email protected] Abstract Social network analysis is the application to sociology of the concept of the network — a set of entities, or nodes, connected by relationships, or ties. Conceptualisation of social structures as social networks has been fruitful in many areas of social research, and has, indeed, facilitated recognition of common substantive patterns and analytic problems in diverse sub-fields of sociology and anthropology, and of similarities to problems in many other sciences. One of the most lively areas of social network analysis has been the development of suitable methods for applying the network concept in social research. These methods address the three main issues in empirical research: sampling, measurement, and data analysis. In each of these areas, the problems faced by network researchers are considerably, though not entirely, different from those encountered by conventional attribute-based research. This session will provide a forum for presentation of new developments in research methods for social network analysis. These papers may be theoretical, concerning epistemological problems in the use of the concept of the social network; methodological, concerning technical developments in sampling, measurement, or data analysis; or empirical, demonstrating novel applications of social network analytic methods in actual research. Part I Playing God on a Budget: Understanding Large-Scale Social Structure from Sample Data and Random Graph Models of Interdependent Local Processes Rick Grannis, Department of Sociology, University of California, Los Angeles, U.S.A., [email protected] Social scientists aspire to understand large-scale social structures, systems, and processes, but typically they only have access to limited samples of individual-level data. Random graph models have proved useful for understanding how local processes translate into global systems and structures. The distribution of vertex degrees in many real-world social networks, however, is often quite different from a “normal” random graph, exhibiting such features as transitivity and local clustering, and such a graph would poorly model their features. In this paper, I use local information to exactly calculate several important global properties for graphs with non-independent edge probabilities. I then calculate the propagation of sample standard errors and show that one can reasonably estimate many useful global properties for social networks with non-independent relationships using only relatively small samples. Keywords: sampling, complex networks, large-scale properties, interdependence 258 New Specifications for Exponential Random Graph Models Tom A.B. Snijders1, Philippa E. Pattison2, Garry L. Robins3 and Mark S. Handcock4 1 Department of Sociology, University of Groningen, Grote Rozenstraat 31, 9712 TG Groningen, The Netherlands, [email protected] 2 University of Melbourne, Australia, [email protected] 3 University of Melbourne, Australia, [email protected] 4 University of Washington, U.S.A., [email protected] The most promising class of statistical models for expressing structural properties of social networks is the class of Exponential Random Graph Models (ERGMs), also known as the p* model. ERGMs are families of probability distributions for graphs and directed graphs (digraphs). The strong point of these models is that they can represent structural tendencies, such as transitivity, that define complicated dependence patterns not easily modeled by more basic probability models. Pseudolikelihood estimation procedures have previously been proposed for these models but these are of doubtful value, and more recently, MCMC algorithms have been developed which produce approximate Maximum Likelihood estimators. These algorithms also are fraught with problems for many data sets, however, which is shown in convergence difficulties. These difficulties are not inherent in the algorithms as such, but can be traced back to shortcomings of the class of probability distributions defined by the usual specifications of the ERGM – where, e.g., transitivity is represented by the number of transitive triplets in the (di)graph. The shortcomings can be summarized by (1) the existence of regions in the natural parameter space where the expected values of the sufficient statistics are extremely steep (i.e., nearly discontinuous) functions of the parameters, which is associated with very large variances and bimodally shaped distributions; (2) the existence of large regions in the natural parameter space where the probability distribution is nearly concentrated on a relatively small set of nearly complete (di)graphs (near degeneracy); these regions often including, or being close to, the maximum likelihood estimate for data sets as found in practice. This paper proposes new specifications of the ERGM. These specifications also represent structural properties such as transitivity and heterogeneity of degrees, albeit by more complicated statistics. Three kinds of statistic are proposed: alternating stars, alternating k-triangles, and alternating independent two-paths. These are linear combinations of related subgraph counts, where the coefficients have alternating signs. The alternation of signs has the effect of a check on the formation of a too large number of the type of subgraph concerned, which goes against the two shortcomings mentioned. Examples are presented both of modeling graphs and digraphs, in which the new specifications lead to much better results than the earlier existing specifications of the ERGM. It is concluded that the new specifications make the ERGM a practically useful tool for the statistical analysis of social networks. Keywords: statistical modeling, graphs, transitivity, maximum likelihood, MCMC, p* model. New Developments in Estimation Methods for the P2 Mode Bonne J.H. Zijlstra and Marijtje A.J. van Duijn, University of Groningen, The Netherlands, [email protected], [email protected] The p2 model is a model for the analysis of social network data. The dependent variable is a (binary) network, or directed graph, on a given set of actors, or nodes; there can be explanatory variables on the levels of the actor and the ordered pair of actors. In the p2 model, differences between actors in attractiveness and productivity (outgoingness) are modeled by random effects. The model includes variance parameters and a covariance for attractiveness and productivity, and parameters for density (log-odds of existence of a tie) and reciprocity; these parameters can be related to explanatory variables. Estimation of the p2 model is not straightforward. Since an IGLS algorithm produced biased estimates for the p2 model, we developed an algorithm to estimate model parameters with a Gibbs sampler. 259 However, the conditional distributions for the random effects and the model parameters cannot easily be sampled from. One way to overcome this problem is using Metropolis steps with random walk proposals to approximate draws from these conditional distributions. Unfortunately, results from a Gibbs sampler with such Metropolis updating steps proved to suffer from the dependence between the covariance for attractiveness and productivity and the reciprocity parameter. Another possibility is to approximate the conditional distributions of the random effects and the model parameters using a normal approximation for the binary network data. Above approaches to overcome estimation problems for the p2 model will soon become available within StOCNET, an open software system for the statistical analysis of network data. We will elaborate on the different estimation methods. These will be illustrated using empirical network data. Keywords: binary network MCMC estimation Part II Actors Invisibility: About the Necessity to Re-Think Internet Research Maren Lübcke and Rasco Perschke, Department of Technology Assessment, Technical University of Hamburg-Harburg, Germany, [email protected], [email protected] To keep up with the rapid development of the Internet is one of the greatest challenges in social research nowadays. Uncountable emails are sent and received, websites are visited by growing numbers of users in search for information and masses of contributions are published in Internet forums. If we look at the Internet as a unique phenomenon we see large scale communication processes with millions of messages passed every day which can be stored and used for future analysis. But something is mostly invisible: the people behind those messages. They disappear behind email addresses and nicknames, often using more than one virtual identity to participate. Communication crosses the border of physical presence, it is substituted through the use of communicative addresses which are merely loosely connected to real people. (Stichweh 2000). The famous cartoon “On the Internet, nobody knows that you’re a dog” by Peter Steiner is an ironical description of this situation, which leads to the well known problems of online-research with sampling methods and cyberethics. Therefore, we suggest to re-think Internet research by leaving actors aside and concentrating on the visible and empirically observable objects. What can be really perceived while glancing at the Internet is sociality based on single utterances. Those messages are referring to each other building a dense network of communication, which is the only visible outcome of users communicative operations. Based on the main assumptions of Niklas Luhmanns social system theory, we are developing a set of concepts which allows us to describe and distinguish different kinds of online communication with a simple framework. Our approach focuses on messages, references, and their social visibility in time. One can use those main ideas to differentiate Internet-based communication like weblogs from online auction or from newsgroup discussions. Methodologically, our approach has the following implication: a switch-over from attribute-based research (users qualities) towards relational research (web of communication) and from static to dynamic analysis. We will present the first results of a tool for simulation and analysis (COMTE) especially designed for large scale communication. This may help to shed light on the specific qualities of online communication and to sensitize researchers who are engaged in using the Internet as a tool of research. Keywords: online communication, methodology, weblogs, newsgroups, online auction 260 Using Social Networks Methods to Measure Changes in Value Systems in CrossCultural Perspective: the Case of Nordic-Baltic Relationship in 1991-2002 Indrek Tart, Centre for Contemporary Cultural Studies, Tallinn Pedagogical University, Tallinn, Estonia, [email protected] In different cultural communities values and their systems are measured by survey instruments like Rokeach or Schwartz value surveys, World/European Value Survey (1990, 1996, 2000) or European Social Survey (2003) questionnaires. I propose to analyse these results using methodology and software of the social networks analysis framework (Stephen P. Borgatti , Martin G. Everett. Network analysis of 2-mode data. Social Networks 19 (1997) 243-269). This makes possible to dive into the data matrices of relations instead of the attributes, even allow to convert individual variables into relational (adjacency) matrices. Its strong part is good visualization of the results. More problematical is choosing of the partition threshold. Several possible constructs based on correlation and shared value amount matrices of data are used and visualized by Ucinet program. Additionally are used clustering and multidimensional scaling techniques as well as CONCOR facility. An empirical research is based on Balticom Program results from 1991 and 1995, Finnish national value research and more recent surveys in Estonia and Sweden. Used data are adjusted to 25-item Rokeach value scale consisting mostly of terminal values. A special analyses of constellations of values equality and freedom was done. The research reveals quite stable value systems for Swedes and Finns and rapid change in Estonians community while other Baltic nations and Russians in Estonia are keeping more their older value orders. It may be explained in the context of axial moments approach (A. Szakolczai, L. Füstös 1998. “Value Systems in Axial Moments”, European Sociological Review 14(3), 211-229) with Estonians turning back to the Swedish kind (axial) pre-1940 value system. Same technical frame may be used in analyzing other data sets of values in cross-cultural contexts. Keywords: social network analysis, values, Nordic and Baltic countries FI 1999 FI 2001 RE 2002 FI 1991 FI 1994 RE 1995 EE 1995 FI 1995 FI 1991A SE 2002 LAT 1995 EE 2002 SE 1995 RE 1991 SE 1991 EE 1991 LIT 1995 FI- Finns, SE – Swedes, EE – Estonians, RE – Russians in Estonia, LAT – Latvians, LIT - Lithuanians 261 Assessing Influence and Selection in Network-Behavioral Co-evolution, with an Application to Collective Action and Informal Sanctioning Christian Steglich, Tom Snijders and Michael Schweinberger, ICS / Department of Sociology, University of Groningen, The Netherlands, [email protected], [email protected], [email protected] From the social science theory perspective, it is fundamental to distinguish between processes by which the network neighborhood of an actor affects his behavior (influence or contagion processes) and processes by which, vice versa, the different actors’ behavior affects network formation (selection processes). Success and failure of intervention programs can crucially depend on the correct assessment of the relative strength of these mechanisms. The opposing causal directions of influence and selection processes, together with the already complex interdependence structure that characterizes social networks, pose a challenge to the statistical analysis of the dynamics of network and behavior. While the collection of longitudinal data on (complete) networks and behavior in principle allows for separating influence from selection on empirical grounds, there are as yet few analytical tools that would enable applied researchers for indeed doing so. In this paper, we introduce such a statistical toolbox that extends earlier methodology for the analysis of pure network dynamics. The model is a continuous-time Markov model for the co-evolution of networks and behavior. Both network change and change of behavior are modeled as resulting from individual decision making. By fitting our model to panel data of networks-with-behavior, we can assess the degree to which influence and selection processes can account for the observed dynamics. Statistical methods for parameter estimation in such models are discussed. In addition, we investigate the ability of the new method to indeed separate influence processes from selection processes by means of a couple of simulation studies. E.g., we simulate data of an evolving network-behavior pattern in which only influence plays a role, and estimate it under the (wrong) assumption that only selection plays a role. By combing such deliberate mis-specification with variations in panel density, we explore the sensitivity of our method to variations in panel design. As empirical example, we analyze a data set on the co-evolution of cooperation and informal sanctioning in a work team. Keywords: network analysis, longitudinal data, stochastic processes, collective action, informal sanctioning Software for Statistical Analyses of Social Networks Mark Huisman and Marijtje A.J. van Duijn, University of Groningen, The Netherlands, [email protected], [email protected] This paper gives a state-of-the-art overview of available software for the statistical analysis of social networks as of spring 2004. It reviews and compares six programs with respect to their statistical procedures, illustrating their functionality with example data. The programs are UCINET, Pajek, NetMiner, Structure, MultiNet, and StOCNET. The choice of routines that were inspected is restricted to procedures for statistical modeling based on probability distributions (e.g., exponential random graph models, QAP correlation, statistical analysis of longitudinal network data). This definition of analysis routines excludes the extensive review of procedure-based routines based on more complex (iterative) algorithms like cluster analysis or eigen decompositions. Also, some other (special-purpose) software packages and software routine packages for general statistical software are reviewed briefly. The paper concludes with some recommendations. Keywords: statistical model, software comparison, exponential graph models, actor characteristics 262 SESSION: PANAL DATA ANALYSIS Organizor: Uwe Engel, Dept. of Sociology, University of Bremen, Germany, [email protected] Abstract In the face of a growing need to analyse complex panel data, the session is primarily devoted to advanced modelling techniques of data coming from multiwave panel surveys and (e.g., household) panel surveys collecting data at the individual and some aggregate level(s) as well as data coming from complex, e.g. multistage, sampling designs. Welcome are contributions to statistical theory and applications of advanced modelling techniques addressing relevant topics in this area. Especially welcome are presentations that shed light on capabilities and possible limitations of growth curve and structural equation modelling techniques as applied to panel data with three and more waves, as are papers on multilevel modelling techniques for panel data. Further topics include missing value treatment of item and unit nonresponse in panel surveys, causal analysis, latent class analysis and latent variable mixture modelling of panel data. Estimation of a two Equations Panel Model with Mixed Continuous and Ordered Categorical Response Variables in the Presence of Missing Data Martin Spiess, German Institute for Economic Research, Berlin, Germany In this paper, the dynamics of effects of the highest level of vocational degree received simultaneously on income and worries about job over a nine year interval in former West Germany is estimated. The model considered is a two equations panel model with one continuous and one ordered categorical variable (with three categories), and several covariates (e.g. age, number of children living in household, marital status, nationality, experience), including highest level of vocational degree received. It is estimated using a GEE-type approach, where the parameters of the systematic part of the model are estimated based on the GEE-approach as proposed by Liang and Zeger (1986) and the parameters of the covariance structure are estimated based on a pseudo-score equations approach (Spiess, 1998; Spiess and Keller, 1999). Parameters of the systematic and the covariance part are estimated simultaneously. The data are taken from the German Socio-Economic Panel (GSOEP) which is a long running household panel. The time points considered are 1991, 1995 and 1999. Since the data set suffers from missing data, two compensation strategies are considered: To compensate for unit nonresponse until 1991, weights are used. Compensation for attrition from 1991 to 1999 and for item nonresponse is based on multiple imputations. The weights are functions of components delivered with the GSOEP. The imputations are generated based on the multiple imputation theory developed by Rubin (1987) using IVEware, a program written by Raghunathan et al. (2001). Analysis results are based on five imputations for each missing value. To assess the explanatory power of the covariates and the modelled covariance structure, a simple general pseudo-R2 measure as proposed by Spiess and Tutz (to appear) is calculated. The estimation results do not imply that the effect of the variable ‘highest level of vocational degree received’ changes over time. Furthermore, they suggest that there are substantial correlations over time for both equations, but the hypothesis of no correlation between the response variables of both equations – given the covariates – can not be rejected. The results based on pseudo-R2 suggest that the explanatory power of both, the systematic and the covariance part, is substantial. Keywords: GEE-approach; panel data; two equations panel model; missing data Comparing Fixed Effects and Covariance Structure Estimators Mette Ejrnaes and Anders Holm, University of Copenhagen, Denmark In this paper we compare the traditional econometric fixed effect/first difference estimator with the maximum likelihood estimator implied by covariance structure models for panel data. Our findings are 263 that the maximum likelihood estimator is remarkable robust to mis-specifications, however in general the fixed estimator is preferable in small samples. Furthermore, we argue that we can use the Hausman test as a test of consistency of the maximum likelihood estimator. Finally we show that the covariance structure model is not identified in the case of time-invariant independent variables. Analyzing Individual and Collective Change of Attitudes Using Panel Data. The Example of Attitudes to Social Security in Germany Nina Baur and Christian Lahusen, Universität Bamberg, Bamberg, Germany Using changing attitudes to social security as a topic, we want to demonstrate possibilities and advantages of adding a temporal component to classical factor and reliability analysis. In order to do so, we differentiate between three survey design types: the singlewave design, the trend design and the panel design. The datasets we use as an example are: the 2002 wave of the German Socio-Economic Panel (SOEP) as an example for single wave data; the 1984, 1991 and 2000 ALLBUS waves as an example for multiwave trend design data; the 1987, 1992, 1997 and 2002 SOEP waves as an example for a multiwave panel design. We explicitly exclude from our discussion well-known problems when using survey data and doing factor analysis, as these are common to all three of the above design types or at least common to all analyses done with one design type. Among these are: sampling problems; wording of indicators; choosing indicators for scale construction; handling ordinal scaled data. Instead, we concentrate on showing which steps have to be taken in order to extend dimensional surveys to multiwave panels: Identifying Dimensions for each wave: The first step of dimensional analysis is identifying (for each wave) the number and the reliability of dimensions as well as the variables for measuring single dimensions. All three design types have this step in common. Factor analysis using SOEP data resulted in two dimensions: satisfaction with social security and preference of state vs. private solutions. This is where factor analysis for single wave designs ends – leading to the main theoretical problem of scale construction with this design type: Scales measuring attitudes or behavioural patterns theoretically imply that similar reactions in many different situations (which can be partly measured in a single survey wave) are stable over time (which can only be measured when analyzing several survey waves). With both trend and panel designs, one can investigate stability over time – if the same items were measured in at least two waves and if the items wording was never changed. Assessing Stability of Dimensions over Time: Dimensions can evolve and dissolve over time. For example, it is plausible that environmental consciousness only became a typical attitudinal pattern in Germany in the 1980s as a result of social movements. Additionally, the importance of single variables measuring a dimension can change over time. For example, although racism always existed in Germany, the word “Neger” (“negro”) only got racist connotations in the 1970s or 1980s. Thus, if dimensions are stable over time, in each wave (a) the number of factors has to be the same in each wave; (b) the same items have to measure the dimension in each wave; (c) reliability has to be stable. Both for ALLBUS and SOEP data, this is true for all waves. Analyzing Collective Change of Attitudes: If dimensions are stable over time, individual values on dimensions can be calculated for each wave. In order to be able to compare the dimensions of different waves, we added the standardized values of single items instead of using regression estimation techniques. Now, analyzing collective changes of attitudes is possible by analyzing time series of aggregated data. For example, SOEP data show that in Germany both satisfaction with social security and preference of state vs. private solutions is remarkably stable: About half the population is neither satisfied nor unsatisfied with the German social security system. Only about 5 % prefer private solutions over mixed or state solutions. At this point, factor analysis using multiwave trend designs ends. In contrast to trend data, panel data allow to trace not only collective change but also individual change and change on the household level, thus avoiding ecological fallacy. Analyzing Individual Change of Attitudes: Multiwave panel data allow tracing changes on the individual level using correlation coefficients and the coefficient of change as indicators. Here too, attitudes measured by SOEP are remarkably stable: Over five year periods, between 40 to 60 % of 264 survey participants did not change their attitudes at all (depending on dimension and period). Even after fifteen years, about 40 % still had the same opinion. Strong opinion changes (of over 30 % on the scale) seldomly occurred. Analyzing Individual Change Patterns: If more than two panel waves exist, one additionally can analyze patterns of change. For example, on the dimension “satisfaction with social security”, seven patterns can be distinguished: (a) no change at all; (b/c) continual rise in satisfaction or dissatisfaction; (d/e) first rise in satisfaction, then in dissatisfaction (or vice versa); and (f/g) first rise in satisfaction, then rise in dissatisfaction, then again rise in satisfaction (or vice versa). One can also make out the period when opinion changes occurred. The paper concludes with the ascertion that panel data analysis can strongly benefit from a sound identification of dimensions and dimensional change patterns, for which we propose a multifacetted modelling technique. In our case, dimensional analysis has come up with an intriguing dependent variable, namely the strong stability of dimensions on the individual and collective level, which can be investigated then according to various techniques of causal analysis. Keywords: Quantitative Research, Panel Data, Survey Data, Time Series Analysis, Longitudinal Research, Factor Analysis, Reliability Analysis, Individual Change, Collective Change, SOEP SESSION: COMBINING DATA FROM DIFFERENT SOURCE Organizer: Susanne Rässler Contact: Institute for Employment Research of the Federal Employment Agency, Regensburger Str. 104, 90478 Nürnberg, Germany, Tel. +49 911 179 3084, Fax. +49 911 179 3297, [email protected] And: Department of Statistics and Econometrics, University of Erlangen-Nuremberg, Lange Gasse 20 D-90403 Nuremberg, Germany, [email protected] Keywords: Statistical matching, data fusion, mass imputation, survey data, record linkage. Abstract The main focus of this session is the integration of registers or, more general, large databases, which contain partial information on a larger amount of the population, and sample surveys, which contain detailed information on a carefully selected sample of the population. The combination should allow an optimal use of the individual data and minimize respondents burden as well as survey costs. A new and rapidly adaptable basis for extracting statistical information shall be created. Methods and tools for a reliable combination of data from different sources are the topic of this session. The methods discussed may either concern record linkage or statistical matching procedures. With record linkage the two samples contain identical individuals that have to be found for the match according to, e.g., the individual's name, address, or social security number. Therefore, it is often necessary to deal with problems of errors in these link variables. Statistical matching techniques typically aim to achieve a complete data file from different sources which do not contain the same units. Traditionally, statistical matching is done on the basis of variables common to all files establishing conditional independence of the (specific) variables not jointly observed given these common variables. Alternative approaches may also be discussed based on multiple or mass imputation. 265 Part I Estimation with Population-Level Constraints on Two Survey Datasets: An Application to first Birth Probabilities by Education in Italy Alessandra DeRose1, Paola DiGiulio2, , Mark S. Handcock3,, Filomena Racioppi2, Michael S. Rendall4, 1 University of Rome, Latina, Italy 2 University of Rome, ‘La Sapienza’, Italy 3 Statistics, University of Washington, Seattle, USA, and Center for Statistics in the Social Sciences, [email protected] 4 RAND, Labor and Population Program; and Pennsylvania State University, Population Research Institute, USA In non-experimental research, data on a population may be collected simultaneously by more than one instrument. The present paper demonstrates the very large reductions in sampling error that are possible when data from more than one instrument are combined with population-level information to infer a multivariate population relationship. Two surveys conducted three years apart, each with retrospective questions on birth histories and contemporary observation of education level, are first pooled in a logistic regression model. The two surveys are shown to sample from the same population during their overlapping retrospective years. Sampling errors in estimates of probabilities of first birth by education level, age, and year of birth are thus reduced by magnitudes equivalent to those achievable by increasing the sample size of one survey by the sample size of the other. Handcock, Huovilainen, and Rendall (2000) demonstrated the potential of a constrained maximum likelihood estimator (CMLE) to combine sample survey data with birth registration data in the estimation of a multivariate model of fertility. Large gains in efficiency were achieved through the intercept term of their single-covariate logistic regression equation. The variance about the intercept parameter was halved when the General Fertility Rate constraint was introduced. In this paper we supplement the survey information with population-level data on women’s first birth probabilities by age and year of birth (but not education) over the same period. By using the CMLE within logistic regression, further large reductions in standard errors about the first birth probabilities by education level, age, and year of birth are achieved. The introduction of population data into the estimation also allows for the evaluation and correction of bias due to non-sampling error in the two survey datasets. Keywords: Combining data; likelihood-based inference; constrained estimation Combining Information from Multiple Surveys in the Analysis of Coarsened Longitudinal Data T. E. Raghunathan, University of Michigan, Ann Arbor, MI 48106, USA, [email protected] Many surveys elicit response to questions about income, wealth and other sensitive items by offering response categories as options. In a longitudinal setting, these response categories often change over time. Longitudinal analysis of such data poses many challenges. For example, the Alameda County Study (ACS) is a cohort consisting of a probability sample of approximately 7,000 subjects recruited in 1968. The subjects were then followed in 1974, 1983,1994 and 1999. In each wave, income was collected as one of many categories and these categories changed over time. Income and changes in income are used as outcome as well as predictors in many analyses. The naive approach of assigning midpoints and analyze them as continuous variable may bias the results due to inherent misclassification. One strategy is to combine information from this survey and another survey that asks exact income. Fortunately, the Current Population Survey (CPS) collects exact income information from a nationally representative sample. This paper develops a multiple imputation approach that combines information from ACS and CPS. A parametric model with income as the independent variable (observed in CPS and interval-censored in ACS) and common variables in the 266 two surveys as dependent variables is used to multiply impute the censored variables. This strategy is evaluated using the actual and simulated data sets. Keywords: Multiple imputation; Interval-censored income A Comparison Record Linkage and Statistical Matching and Their Impact on Linear Regression Analysis Michael D. Larsen, Department of Statistics, Iowa State University, USA, [email protected] It is often the case that one desires to conduct an analysis but the information needed for the computations is not contained in a single source. One such scenario occurs when a large file containing records on a population is available in one agency or in one database, but other important variables are located in a different agency or a separate database. If there are unique identification numbers for exactly linking records on people between files and the linkage is allowed by both agencies, then composite records can be formed and the analysis can be accomplished. If joining the databases is allowed but unique identification numbers are not available on one file or the other, then there are two options. One option is to use probabilistic record linkage to try to identify exact matching records on the two files and then conduct the analysis on the merged files. A second option is to create a matched sample through statistical matching using propensity score methods and then studying the relationship between variables of interest. Both methods introduce errors and uncertainty. Errors in linkage occur in record linkage. Variability is introduced through statistical matching. Using a simulated data set from a large scale survey, the two methods are compared where the goal is to estimate a linear regression relationship. Methods of propagating errors of linkage in the first method and variability due to statistical matching in the second method are presented and compared. Keywords: Fellegi-Sunter procedure, Latent class models, Measurement error, Mixture models, Propagation of error, Propensity scores Combining Separate Datasets in the OPUS Project J. Polak1 and Ch.D.R. Lindveld2, 1 Imperial College London, UK, [email protected] 2 Imperial College London, UK, and Delft University of Technology, The Netherlands The OPUS project is a three-year project, funded under the European Commission 5th framework IST research programme, which has been set up to develop, demonstrate, and apply a statistical framework for combining separate datasets in the estimation of a single latent variable. The setting for which the framework is developed is one in which individual datasets reflect (or partially reflect) observable aspects of a latent variable, and that a certain amount of theory of the problem domain exists and can be operationalised in quantitative mathematical / statistical models. Such models, called General Apriori Models (GAPM’s) within the project, are translated into graphical models on basis of their conditional independence structure. In building a graphical model, a graph is constructed in which variables are represented as vertices, and conditional independence relationships as edges. No assumption is made regarding linearity or non-linearity of the relationships encoded, or the distributional forms used (unlike e.g. Structural Equation Modelling). In addition, particularities of the observation process can be explicitly encoded into the model. We will assume that such a representation is possible, but as the class of graphical models is a very wide one this assumption does not seem unduly restrictive. The advantage of using the language of graphical models is that in the past 8 years software has been developed to operationalise such models. The most notable application is where observational evidence is available on only part of the nodes, and the joint probability density of all nodes in the graph is expressed and calculated using non-informative apriori distributions for variables corresponding to unobserved nodes and repeated application of Bayes rule. 267 The essence of the approach is that disparate datasets can be linked, on basis of domain-specific theory, into a single graphical model that comprises all observed variables plus the latent variable(s) of interest. Constructing the joint probability distribution for the graphical model, and determining the probability distribution of the latent variable are can then (in principle) be treated as a matter of routine. The OPUS project aims to demonstrate this approach in two subject areas: transportation (indicators of urban mobility, and estimation of regional origin-destination flows) and health (health impacts of socio-economic factors and environmental hazards). This was done in order to demonstrate the feasibility of specialising the methodology to a particular domain and applying them, and to investigate the difficulties that arise in doing so. The OPUS project specifically includes, as one of its worktasks, discussions with interested parties in the scientific community, consultancy, and potential beneficiaries of the results. To this effect, the project has created, and actively maintains, a website for public dissemination of appropriate sections of the results and discussions on this subject (see http://www.opus-project.org/). Interested parties are invited to participate in the discussions, or to simply follow them. Modelling the Construction of a Social Accounting Matrix in the Context of Statistical Matching Marcello D’Orazio, Marco Di Zio and Mauro Scanu, ISTAT - Italian National Statistical Institute, [email protected], [email protected], [email protected] In the last years, interest on Statistical Matching problems has increased. This is due to the large amount of data-sets available and, at the same time, to the need of timely and not costly information. Actually, Statistical Matching techniques aim at combining information from different sources. In particular, it is assumed that the two sources (e.g. two samples) do not observe the same set of units, so that neither merging nor record linkage techniques can be applied. In this paper we present some preliminary results on a Statistical Matching application for the estimation of the Social Accounting Matrix (SAM) in ISTAT. A Social Accounting Matrix is a system of statistical information containing economic and social variables in a matrix formatted data framework. The matrix includes economic indicators such as per capita income and economic growth. In Italy, as well as in other countries, an archive containing all the necessary variables is not available. Hence, an attempt to estimate the SAM can be performed by means of the fusion of the Household Balance Survey conducted by the Bank of Italy and the Household Expenditure Survey conducted by the Italian National Statistical Institute. In this paper it will be outlined how this application can be modelled in the Statistical Matching context. Once clarified the statistical issues and the hypotheses that cannot be avoided, the emphasis will be posed on the consequences of the Conditional Independence Assumption (CIA), i.e. the assumption of independence between the variables observed distinctly in the two sources conditionally on the commonly observed variables. Moreover, it will be outlined how this assumption can be substituted by other suitable models based on the economic theory. These models, alternative to the CIA, are under investigation. Some results about this study will be provided. Keywords: Conditional Independence Assumption, Expenditure Data, Income Data Integrating Data for Innovative Analyses: The North Carolina Education Research Data Center Elizabeth Glennie, Duke University, Durham, USA, [email protected] Without ready access to relevant data, researchers cannot conduct high-quality education policy research. Although many state governments collect vast amounts of data on students, teachers, and schools, they do not store the data in formats that researchers can use. North Carolina has not built longitudinal data bases to follow the progress of children over time, nor have they linked teacher and 268 student characteristics. Furthermore, they cannot link agency data to files from other sources, such as other administrative data or surveys. In 2001, the North Carolina Education Research Data Center was established as a partnership between university researchers and government officials to facilitate problem-focused education research. This Data Center encrypts confidential information, creates longitudinal databases, and links district, school, classroom, student, and teacher data files. It also integrates education data to files from other government agencies and university surveys. With files from the mid 1990s to the present, the Data Center can track teachers’ careers over time, track students’ progress over time, link teacher information to student data, and aggregate student or teacher data to a school or district level of analysis. These education files have supported studies on topics including within-school racial segregation, factors that influence teachers’ working satisfaction, and whether the accountability system increases the turnover rate. The Data Center links education records to files from other sources. A researcher surveyed kindergarten students and their parents. Four years later, the Data Center matched the surveys to the children’s third grade test results. To study the effect of welfare reform on children, the Data Center linked welfare records to educational outcomes. To study teacher perceptions of an incentive bonus, the Data Center linked survey results to teacher work histories and school characteristics. For another analysis, the Data Center attached student addresses the student educational information, permitting analyses of neighborhood and classroom effects on education. In each of these studies, the Data Center reduced the response burden on teachers, school administrators, and government officials and lowered study costs. Agencies and researchers use varied means to collect data. Thus, files come in diverse formats and include different combinations of identifying information (name, birth date, social security number, school grade, or ethnicity). Often, those collecting the data do not validate it. The Data Center has developed methods for validating and standardizing identifiers and for using probabilistic matching techniques to integrate these files and permit innovative analyses. Keywords: administrative data, survey data, record linkage Part II Linking Job Episodes from Survey and Register Data: Specific Challenges and Empirical Results Maike Reimer and Ralf Künster, Max-Planck-Institute for Human Development, Center for Sociology and the Study of the Life Course, Lentzeallee 94, 14195 Berlin, Germany, [email protected], [email protected] This paper deals with the linking of longitudinal survey and register data on the basis of events or episodes. Unlike persons, events, episodes or transitions are themselves products of error-prone data generation processes and therefore can differ between the data sets to be linked. The data sets to be linked are retrospective survey reports from the members of the birth cohorts 1964 and 1971 in the West-German “Life History Study”, and register data from the German Social Security agencies about employments held by the respondents between 1990 and 1997. Event history data describes time periods as temporally ordered sequences of events or episodes and transitions between them. Theories and findings from cognitive memory psychology and knowledge about the data generation process in the Social Security database are used to explain how job episodes are constructed and dated in the two datasets and which differences can occur. On this basis, the two data sets are prepared and a match rule that uses starting and ending date of each episode as identifying information is developed, applied and evaluated. It distinguishes between perfect matches, time-liberal matches (that tolerate deviations up to 2 moths) and multiple matches (that match more than one episode from one data set to one episode in the other data set if they cover the same time period). 269 About 70 % of all episodes are thus successfully matched. Employment episodes of women, of the younger birth cohort and of respondents with less work experience are less likely to be matched. The same goes for episodes from the earlier career stages. Qualitative case studies show that linkage failure can be attributed to the differing construction of events/episodes, errors in the recall of the identifying information and in the recall of the preparatory information in the survey reports. In the light of these results, three broader issues are discussed: 1. the question of how a specific time period in the linked data file should be represented, 2. the relationship of manual linkage on the basis of case studies on the one hand and automatic linkage rules on the other, and 3. the implications for the validity of the episodes and dates obtained in the survey. Keywords: record linkage, survey data, retrospective measurement, job spells Non-Exact vs. Exact Matching with Applications to Wages Statistics Seppo Laaksonen, University of Helsinki and Statistics, Finland, [email protected] Administrative data are typically derived from different organisations. In this case, a national statistical institute (NSI) compiles every four years the data on wages using a sample survey, the reference period being one month. Moreover, this NSI maintains an employment register with good coverage. The key variables of this register are employment contract and occupation while the basic personal characteristics are also included. This register, unfortunately, has no information about wages and its components that are of main interest for users. Fortunately, the tax register authority has the complete yearly data on taxable wages, but unfortunately, this authority does not want to assign these data with exact identifiers to the NSI due to confidentiality problems. There are however common variables in both data sets but the exact matching of these data are not possible. For the excercise of this paper, we received these identifiers and consequently we are able to test how well different techniques succeed in non-exact matching. The methods used are closely related to those used in imputations. Both parametric and non-parametric matching models are applied. The results with some methods are promising but the overall quality is not maybe reasonable for a demanding user. Combining Datasets to Create a Synthetic Population: Sample Enumeration. Ch.D.R. Lindveld1 and J. Polak2 1 Imperial College London, UK and Delft University of Technology, The Netherlands, [email protected] 2 Imperial College London, UK The Dutch National Model System (NMS) is a model system to forecast mobility of over time horizons of 15 to 20 years with the associated traffic flows. The NMS is used extensively for transportation policy evaluation in The Netherlands. The NMS is based on so-called disaggregate discrete choice models, which are econometric models based on a random utility maximisation framework that model mobility decisions (such as the number of daily tours generated by households, and the choice of travel mode and destination of individuals). Such a model system faces a number of difficulties: a) how to capture choice behaviour at individual level b) how to get from individual level models to population level in order to reproduce current traffic counts in a way which is suitable for application over long time horizons. A further isue is how to obtain the required inputs for such long-term developments. We will restrict ourselves to issues a) and b). The first issue is addressed by carrying out stated preference experiments on a (representative) sample of individuals (the “prototypical sample”), and estimating econometric models on the resulting dataset (which we call called dataset 1). This results in a number of estimated disaggregate choice models that require individual-level characteristics, which are not available on a population basis. 270 In order to apply these disaggregate choice models at population level, say in a certain geographical zone, the original sample is expanded to the level of the population in the target zone, while respecting various distributions of population-level quantities (dataset 2). Examples are: age, gender, license holding, income class, and household composition). The expansion factors are noted, and the models are applied to the original sample with individual-level variables kept as they were, and trip-level variables changed to reflect the new choice situation. The resulting choices are then expanded with the expansion factors obtained earlier. This process is repeated for each geographical zone. This procedure is called sample enumeration. The set of expanded prototypical samples can be seen as a synthetic population. As a matter of interest, the information from datasets 1 and 2 is combined in producing the final result: an estimate of the current situation at population level. Dataset 1 provides the information about individual-level behaviour and dataset 2 provides the information with which the prototypical sample can be matched to the population of a geographical zone. The “glue” which holds this construct together is the assumption that the estimated disaggregate choice models capture individual preferences which can be transferred to each geographical zone. This is a testable hypothesis, which has been found to hold to an acceptable degree. We note that other, much more extensive and elaborate examples exist, of which we mention the ALBATROSS model, and the TRANSSIMS model. Robustifying Imputation Model Estimation Florian Koller, GfK Fernsehforschung, Nordwestring 101, D-90319 Nürnberg, Germany, [email protected] Mass imputation of missing data needs robust imputation models which require a minimum of assumptions from the imputer with regards to modelling. The specification of “tailor-made” models is hardly feasible if the number of variables with missing values is very large. Therefore, robust estimation techniques are needed to minimise biased inferences due to misspecified imputation models. One approach that will be discussed are Generalized Additive Models which can help to find a balanced solution between underspecified models and overspecified models which “explain” the error term. A second aspect deals with the outcome of the model: It will be shown that “rounding to nearest observed value” can introduce bias for categorical data and how this problem can be mitigated. These techniques will be applied to a fictive split questionnaire survey design where it is assumed that the expected response burden is too high for a complete single-source survey. Split questionnaire designs are favoured over a data integration approach, because all distributions considered for analysis remain jointly observed in a small subsample, thus avoiding the identification problem inherent to data fusion. Keywords: Mass Imputation; Multiple Imputation, Sequential Regression; Predictive Mean Matching; Generalized Additive Model; GAM; Data Fusion Linking Public Health Data in Georgia, USA Mohamed G. Qayad and Emily Kahn, Georgia Division of Public Health/MCH Epidemiology Section, Centers for Disease Control and Prevention, USA, [email protected] Background. Linking data from different sources is becoming an increasingly common practice in public health. Data linkage enables public health to perform essential public health functions and provides opportunities to address public health problems that would not be possible without linked data. In Georgia, we have linked Birth data with Medicaid claims, infant deaths, hospital discharge data, and data from governmental programs to provide food assistance and medical assistance to women and children. We use the linked data to fulfill our contract deliverables, to report State and 271 Federal government-mandated performance measures, and to perform administrative and research duties. Despite the opportunities it brings, combining data by matching records requires additional capacities that are not often covered during the formative years of public health professionals. In this paper, we describe the methods to evaluate the accuracy of matched records. Methods. We obtain the databases from the owners, directly or indirectly, through legal means. Using Automatch probabilistic record linkage software, we match record pairs using multiple variables, personal identifiers, demographic information, geographical information, and dates of events. We evaluate the accuracy of the matched and unmatched pairs by manually reviewing sets of record pairs selected randomly from the matched and unmatched records. In addition, we assess the percent agreement between two programmers who independently use the same data and perform the linkage. External sources are used to assess the accuracy of matches when possible. Conclusion. Linking data is technically and theoretically challenging, but the benefits to public health agencies in being able to analyze complex data to solve health problems that would not be addressed otherwise is invaluable. The value of linked data, however, depends on the accuracy of the linked product. The evaluation methods presented here should be undertaken before any use of the final linked data set. Two Stories about Youth Unemployment? Combining Register Data and Self-Reported Data Hans Dietrich, Institute for Employment Research of the Federal Employment Agency, Regensburger Str. 104, 90478 Nürnberg, Germany, [email protected] Unemployment spell are relevant episodes in the school-to-work transition of young persons even if the function of this type of episodes for the individual life course could be rather different in dependence on timing, duration, or outcome of unemployment spells. From this point of view two questions arise: In how far self-reported longitudinal data about individual unemployment experience correspond with process produced data? In the case, these informations do not fit, what are the quantitative and qualitative findings and explanations for the observed discordance? After a descriptive part concerning the data fit between register data and self-reported data, an analytical section in the paper will ask for social explanation of given data-discordance. Database is a multi-wave-study, starting with a representative sample of 4000 young persons, registered unemployed at the Federal Employment Services between 1998 and 1999, who where interviewed in the years 2000, 2001, and 2004 matched with individuals unemployment register date from 1998 to 2004. STREAM: INFORMATION AND SIMULATION SYSTEMS SESSION: NON LINEAR MODELLING Organizor: Cor van Dijkum Contact: Department of Methodology and Statistics, Utrecht University, [email protected] Keywords: Dynamics, human behaviour, non-linear models, computer simulation Abstract In the struggle of the social sciences to describe and explain the dynamics of human behaviour, theories and models play an important role. Social scientists such as Fechner (1889), Coleman (1964), Forrester (1968) and Blalock (1969) made a plea for theories in which behaviour is dynamically understood and as a consequence mathematically modelled in differential equations. 272 Non linear models are another step ahead. Linear differential equations are handled in the natural sciences analytically by using adequate methods of solving. But only lately, made possible by advanced hard- and software, natural scientists are able to simulate and handle the complex dynamics of non linear differential equations. Thereby they demonstrated that many phenomena in nature can only be understood by using these equations. Inspired by this advancement in science, social scientists have discovered that also the dynamics of human behaviour can also be better described and understood with non linear models. Such pioneers in the social sciences (from psychology, sociology, but also economy, medical sciences, etcetera) are requested to present in this session their state of 'the science and art'. They are invited to demonstrate and explain their methodological approach, for example by answering the question: a. In what way their non linear model is inferred from a dynamic theory; b. In what their way their model is confronted (falsified and verified) with empirical facts; c. In what way the simulated output of their model is compared with time-series of empirical facts; d. What statistics is used to calibrate, falsify and verify the model (fitting measures, etc.); e. In what way they solved related methodological and technical problems; f. In what way the model was useful for the practice of their field. Part I Dynamics of Functional Connectivity in Ongoing, No-task EEG. Evidence for Scale-Free Properties and Explanation in Terms of Coupled Limit Cycle Oscillators C.J. Stam, E.A. de Bruin, T. Montez, K. Linkenkaer-Hansen, M. Breakspear, B.W. van Dijk Functional connectivity refers to statistical interdependencies in multi channel EEG recordings which are assumed to reflect functional interactions between brain regions. We show that during a no-task, eyes-closed state the global level of functional connectivity shows fluctuations which have a scale-free character, and determine optimal parameters for detecting this scaling behaviour. We attempt to explain this property of ongoing EEG in terms of a model of weakly coupled non-linear oscillators. Time dependent global levels of synchronization were determined in eyes-closed, no-task EEGs (19 channels, 32 second epochs) of 15 healthy subjects (10 males; mean age 22.5 years) using the synchronization likelihood (SL). Fluctuations in SL over time were analysed with detrended fluctuation analysis (DFA). The influence of EEG filter bandwidth and parameters of SL computation on the detection of scale-free fluctuations in functional connectivity levels was determined. A similar analysis was applied to the output of a model of 64 globally coupled limit cycle oscillators to study the influence of coupling strength and frequency distribution. Evidence was found for significant, scale-free long term correlations in the global level of synchronization, and optimal parameters to detect this could be determined. The slope of the DFA fit was between 1.2 and 0.9, and decreased with increasing bandwidth. In the model the SL showed the theoretically expected sigmoidal dependence on coupling strength. For weak coupling scale-free fluctuations in the level of SL were found, with a slope that depended upon coupling strength as well as bandwidth. Fluctuations in global levels of functional connectivity, as determined with SL, show significant long term correlations and scale-free properties during an eyes-closed, no-task condition. This could be replicated in a model of globally coupled limit cycle oscillators close to a bifurcation point. Long term correlations and scale-free behaviour suggest that spontaneous brain dynamics is close to a critical and optimal state for pattern formation and integration of information between different spatial and temporal scales. Keywords: Electroencephalogram Synchronization Functional connectivity Detrended fluctuation analysis Scale-free Kuramoto’s model 273 Dynamical Complexity. A Measure of Critical Instability During Human Change Processes S. Weihrauch and G.Schiepek Our project is based on the idea that change processes of human mental and behavioural patterns can be described by the terms of self-organizing systems. For description we use the transdisciplinary paradigm of “Synergetics” (founded by Prof. Hermann Haken). Our data were produced by psychosomatic in-patients treated at a hospital ward (average stay: 66 days). The success of treatment was measured by a set of questionaires filled out before and after treatment. In order to receive process data we used a Psychotherapy Process Rating, containing 53 items. concerning mental and interpersonal topics. Intensities were to be rated on a 7-step Likert scale except for 7 items recorded by visual analogue scales. Corresponding to the theory of self-organization, changes in system dynamics are often accompanied by critical instabilities, in that creating an increasing probability of order transitions. The recognition of these instabilities is an important information for the practitioner to modulate the therapeutical intervention. In searching for adequate methods to describe critical fluctuations, we were limited by the resolution of the measurement scale as well as by the length of the time series. The method we developed identifies the amplitude as well as the frequency of measurement changes (fluctuations) within a seven-day gliding window. Additionally the distribution of data in the window is analysed. We call that combined measure “dynamical complexity”. After applying the method to the time series we obtained for each patient 53 index-time-series containig information about the dynamic attributes of the original data. In order to minimize random effects we summarized the index-time-series by the 7 factors of the daily rating scale. Regarding the group of successful patients (e.g. the patients whose problems were reduced by the therapy) we could show a significant correlation between the effect-size of therapy and the highest attained local complexity during therapy. As a consequence of the short length of the section needed to calculate the index, it is possible to get the results close on, e.g. three days after the day it referres to. We use that fact in an computer-based real-time-monitoring system which can deliver hints of the system dynamics during the ongoing process of psychotherapy. Evolutionary Economics and Complexity Carl Henning Reschke, Mainzer Strasse 80, 50678 Cologne +49-221-3406585 Germany, [email protected] University of Witten/Herdecke, Macroeconomic and Institutional Change, 58448 Witten, Germany The aim of this paper is to sketch a strategic map of some of the relevant issues at the intersection of economics, sociology and evolu-tionary theories of development. The unit of evolution is taken from psychological personal construct theory. Methodologically, a success-ful integration of these areas requires a transfer of complexity and non-linear methods and development of further measurement methods in the complexity area. A particular challenge is the integration of emergence phenomena and the measurement of evolutionary change in practical research efforts. I review selected approaches from physics (socio- and econophsy-ics), computer science (John von Neumann’s theory of selfreproducing automata), biology (systemic evolutionary theory, population approach) and economics (Lancaster and Saviotti on characteristics) and dis-cuss how these areas intersect conceptually. This is integrated into a model of social evolution. Open questions revolve particularly around the characterization, modelling and empirical research on evolution-ary change processes. On this basis, I pose challenges to complexity researchers and statisticians. Special attention is given to the tension between ‘reductionism’ and complexity. Each of the disciplines mentioned offers a specific way of dealing with complex phenomena, with distinctive advantages and disadvan-tages. Physics measures potentially relevant characteristics. 274 Computer science aims at algorithmic accounts of how complex systems can be artificially constructed. Evolutionary thinking asks for the adaptive value of specific characteristics. It shifts attention from a generalized standard optimality criterion to a differentiated account of selection criteria and selection conditions in specific circumstances. Hypotheses and models are illustrated with issues from the pharmaceuticals market. Keywords: social evolution, models, complexity, socio-/econophysics Part II The Dynamics of Bilingual Language Development in Immigrant Children Annemarie Peltzer-Karpf and Marion Griessler, Graz University, Austria This paper deals with the dynamics of language development in a multicultural setting. Data collected in a long-term study on bilingual development in immigrant children (N = 106) are given a neurocognitive interpretation which links brain development with the linguistic growth spurts ascertained in four different languages: Bosnian/Croatian/Serbian, Turkish, German and English. The focus is set on the interplay of systems along a four-year trajectory with the control parameters set by monolingual children in Austria, Bosnia and Turkey. Each of the children was followed up from age 6 to 10 in his/her own linguistic career. The individual languages were given a regular psycho- and pragmalinguistic scrutiny (including the frog story), transcribed in CHILDES and then submitted to a dynamic cross-linguistic analysis. The outcome is a vast data-pool which allows for the tracing of fluctuations and turbulences along increasingly intertwined chaotic itineraries with different onsets and developmental conditions. Of particular interest is the question how a lopsided start into a new language influences the ongoing development of the first language. In linguistic terms the prospective temporal asynchrony takes the following stepping stones in the intake and further processing of information: while starting with the search for coherence in the new language immigrant children keep sorting out the more sophisticated morphological and syntactic patterns of their first language, enrich and reorganize their lexical repertoire and work out explicit rules for the management of discourse and textual structures. A more intensified scene-segmentation in the minor language soon creates a state which shows all the signs of cross-linguistic chaos, i.e. turbulences and fluctuations expressed in the over-generalization of rules, excessive morphological marking and free-floating syntax. The investigations carried out so far provide evidence for a changing sensitivity to language cues at different times and structure-specific time scales for system development. Of particular importance is the notion that successive bilingual language development does not take a linear path but rather comes in phases of intermittent turbulence, fluctuations and stability apt to swap linguistic borders in midstream. The data were collected in the course of a research project funded by the Austrian Ministry of Education, Research and Culture Keywords: language development, bilingualism, temporal asynchrony, non-linearity, critical mass, chaotic itinerary, neural development, brain-growth spurts Methodological Issues of Labour Market Analysis for Countries in Transition Mikhail Mikhalevich1 and Ludmilla Koshlai2 1 Ukrainian Academy of Foreign Trade, :[email protected] 2 Institute of Cybernetics, Ukrain, [email protected] The present social situation in post-communist countries is certainly an exceedingly difficult one. There are various grave problems: growing inequalities in incomes both at the stages of decline and growth of business cycle form islands of prosperity in a sea of distress and poverty, capital flights and growing indebtedness. All this leads to criminalization in a society. The extremely low wages are the 275 main reasons of such a situation. Many phenomena, which are observed at the labour market in postcommunist countries, can be explained when considering this segment as a market with imperfect competition. The analysis of such a market needs a special methodology combining qualitative and quantitative methods, modelling and economic interpretation of the obtained results. The presentation applies the dynamical model based on monopoly consumption or monopsony at the labour markets in transition economies. We consider a manufacturer of commodities who is a monopolist-consumer in some segment of a labour market. The analysis of his behavior demonstrates that wages grows when an expected demand does not exceed some value. If a demand is greater than this value, a manufacturer receives the largest profit for constant wages. The mentioned value is determined by the labour-supply function and by the value of expenses that are conditioned by purchasing of a unit of labour by a manufacturer. Thus, the dependence between wages and an expected demand is non linear. The impact of prices on the demand is also described by a non linear function. Hence, we have obtained the non linear model, some analytical and numerical methods can be applied in order to solve this problem. The conditions for availability of the feedbacks of "wage decrease results in decrease of internal demand and to industrial decline and leads back to wage decrease" and "industrial growth without increase of wage leads to non-increasing of internal demand that stops industrial growth" are analysed. This explains such transitional phenomena as a low l evel of official unemployment, decline in productivity of labour and other features of an employer's behavior. The simulated output of the model demonstrated the possibilities of appearance of specific economic cycles. The investigation results are compared with the real data from different branches of the Ukrainian economy. The monopoly mechanisms in the wage-employment relations are typical for approximately half of them, whilst oligopoly competition between branches exists in the remaining sector. The essential aspects of the theory of employment for early stages of transition are developed and discussed as the conclusion. Keywords: methodology, non linear models, computer simulation, human behavior. Anticipation and Codification in Communication Systems: Towards a Simulation Model for Luhmann’s Sociological Theory of Communication Loet Leydesdorff, Science & Technology Dynamics, University of Amsterdam, Amsterdam School of Communications Research (ASCoR), Kloveniersburgwal 48, 1012 CX Amsterdam, The Netherlands, [email protected], http://www.leydesdorff.net A system which contains a model of itself can function in an anticipatory mode. Anticipatory systems will be simulated using a cellular automaton. The anticipatory (sub)system operates on the modeled system with hindsight, and thus an observer can be generated. The anticipation is based on advancing the clock of the modeling (sub)system with one time-step. From this next stage the model looks back, while the ensemble of the system and its model moves historically, that is, with the arrow of time. While the forward movement can be simulated as a recursive routine, the backward subroutine operates incursively, that is, as a feedback against the arrow of time. In this study, I show that under the condition of functional differentiation the social system can be expected to contain two anticipatory mechanisms: (1) meaning is provided with hindsight, i.e., with a time step difference from the reflected operation; (2) the differentiation in the social system generates an asynchronicity (Ät) between the operation of its subsystems at each moment in time. This asynchronicity allows for an anticipatory mechanism other than (orthogonal to) the one over the time axis: the codes of the different subsystems provide them with representations of each other. For example, the state of the art in a technology is reflected in the prices on the market at each moment in time. The historical development of the technology follows a dynamics different from the development of the prices on the market. These two subdynamics, however, can be expected to interact. 276 When two anticipatory mechanisms are left free to operate on each other, a so-called “strongly anticipatory system” can be shaped as the results of a resonance or a coevolution between the two degrees of freedom. While weakly anticipatory systems contain only a model of themselves—which can be used as a prediction—a strongly anticipatory system is also able to construct its next stage. The social system generates endogenously mechanisms for its renewal. For example, the social systems can be redefined in a techno-economic evolution under the condition of functional differentiation. The functional differentiation provides us with the degrees of freedom. The knowledge base of this system and its subsystems (e.g., the economy) can increasingly be globalized. Knowledge further codifies the meaning in the communication. When two subsystems operate upon each other incursively, hyper-incursion can be expected. The formula for the hyperincursion indicates an uncertainty in the decision-rule which can be appreciated as the need for organization at the interface by using Luhmann’s sociological theory of communication. Catastrophe Theory and Social Policy Making: A Nonlinear Model of the Early Pension Insurance Legislation in Western Europe Georg P. Mueller, Department of Social Sciences, University of Fribourg, Av. de l'Europe 20, CH1700 Fribourg, Switzerland, [email protected] This paper aims at modeling the process of the introduction of new social policy laws by using tools and concepts borrowed from mathematical catastrophe theory. The model presented in this paper departs from a societal conflict between the antagonists and the protagonists of a new social policy. In order to ensure its own political survival, government is assumed to minimize this conflict by making appropriate political decisions about the introduction of social reforms. The policy resulting from this conflict minimization depends among others on the power distribution between the two mentioned conflict groups. Historical trends, which change this power distribution, eventually lead to the disappearance of a well established conflict minimum and thus trigger a sudden discontinuous shift to a new social policy, which is associated with an alternative minimum of internal conflicts. The paper analyzes the conditions under which such shifts in social policy occur. The results of this mathematical investigation are used for an empirical test of the model with historical data about the introduction of the first pension insurance laws in Western Europe. This test of the model shows, that the dependent variable, i.e. the percentage of the labor force covered by the first pension insurance laws, is relatively well explained by changes in the power distribution between labor movements and entrepreneurial groups. This finding corroborates the basic structure of the model, which at the end of the paper is further explored with regard to its consequences for the feasibility and the persistence of social reforms. Keywords: Catastrophe theory, model building, social policy, government behavior, discontinuous social change. Part III Application of Fuzzy - based Methods for Urbanization Problem Ludmilla Koshlai1 and Mikhail Mikhalevich2 1 Institute of Cybernetics, Ukrain, [email protected] 2 Ukrainian Academy of Foreign Trade, [email protected] The problems of development relations between large cities and suburbs are very important both for market and post-communist countries. Handling these problems, different economic, ecological, social, cultural and other weakly formalized factors must be taken into account. Therefore the special methodology must be elaborated for this purpose. "Fuzzy" terms are used often in the mentioned problem formulation, so the theory of fuzzy sets and fuzzy logic may be used as the basis for methodology elaboration. The important aspects of fuzzy-based methods application are considered in 277 the presentation as the example of two connected problems: 1) the establishment of borders for urbanized (suburban) area; 2) the search of legislation alternatives to determine financial, administrative and political relations between municipal structures. Several approaches including multicriteria generalization of the "gravity" model and fuzzy-logic-based hierarchical procedures of complex alternative analysis are proposed. They are united using the system analysis. The empirical results will illustrate the proposed approaches from a practical point of view. Keywords: complex societal problems, fuzzy logic, dss. Low-Dimensional Nonliear Dynamical Systems as a Tool for Socio-Economic Modelling Yuri Yegorov, Institute for Advanced Studies, Vienna, Austria, [email protected] The art of mathematical modelling challenges both the complexity of real world and analytical tractability. While nonlinear dynamic systems have been proved to be a proper tool of modelling in physics which captures the essence of the process, their application in social sciences is still very limited. Two methodological extremes, both far away from optimally simple description of reality, seem to dominate. The first can be named „oversimplification“; it relies mostly on linear static relationships across variables. While it has an advantage to be understandable for a broad group of scientists without special mathematical background, it can rarely capture intereseting phenomena, known as „nonlinear science“. The second can be named „excess complexity“; it either uses too many variables or focuses on chaotic solutions, making robust predictions and simple policy recommendations hardly possible. Due to rigidity of some scientific schools in using the concept of rationality, even simple dynamical systems are taken away from subfields of economics, and at the same time very complex functional relationships are analyzed in static form. The interaction between social and natural phenomena clearly escapes rationality trap, as nature cannot be rational. Forrester has suggested the methodology of dynamic modelling of such systems, which has been used successfully in prediction of future processes on the Earth. The trap here is that models typically use too many variables, and the result often depends on the correct choice of parameters and initial conditions. But very often low-dimensional systems can both represent an interesting dynamics and allow for some analytical tractability along with numerical robustness of results. Here two threedimensional models of such type will be presented. The first focuses on interaction between economics and nature and shows that Malthusian and neoclassical concepts can represent two extreme cases of more complex dynamical system which is governed only by three nonlinear differential equations for population, natural resources and technology. The system is derived using quite standard relationships, but taking part of them from economics and other part from biology. The second model is about the link between urbanization and demographic transition. While more than three economic and social variables are used initially, the dynamics is governed by the system of 3 differential equations for capital, total population and rural population. While the first model predicts a possibility of a catastrophy after a period of almost exponential growth, in the second model population stabilization can take place due to low fertility in urban areas. Keywords: nonlinear dynamical system, socio-economic modelling, technology, nature, demography, urbanization. 278 SESSION: TOOLS AND TECHNIQUES FOR AGENT-BASED SOCIAL SIMULATION Organizors: Klaus G. Troitzsch1 and Nigel Gilbert2 Contact: 1Universität Koblenz-Landau, Institut für Wirtschafts- und Verwaltungsinformatik, Germany, [email protected] 2 Department of Sociology, University of Surrey,Guildford, Surrey, UK, [email protected] Abstract Papers in this session should introduce new software for agent-based social simulation and address topics like the unification of different approaches, for instance to combine an equation-based macro model of the natural environment with an agent-oriented micro model of the human beings living in it, the calibration of social simulation models in order to foster their use to solve actual problems, the comparison and integration of different tools for different research and applications objectives (Swarm, RePast, NetLogo, SDML, MAML, etc.), since there is a number of different tools available with restricted capabilities which need to be integrated and to be made more user-friendly, the design of new tools and ways to make them available. Part I: The Applicability of Agent-Based Social Simulation Tracking the Invisible Hand: An A.B.M. Approach to CDA Design with Evolutionary Behaviour M. Posada Calvo, Cesareo Hernández Iglesias and Adolfo López Paredes, Grupo de Ingeniería de Organización, InSiSoc, ETS Ingenieros Industriales de Valladolid, 47011 Valladolid, Spain, [email protected], [email protected], [email protected] The ancillary hypothesis of unbounded rationality has dominated, and still dominates, economic modelling. This extreme assumption has been relaxed in a fast growing literature under different headings: new institutional economics, evolutionary economics and experimental economics (E.E.). In E.E. the eighteenth century Scottish tradition (Hume, D. and Smith, A.) is revealed in the observation of emergent order in numerous studies of existing markets such as the continuous double auction (CDA). This emergence demonstrates that the logic of choice (individual rationality) and the ecology of choice (social rationality) can diverge, (Smith V. 2002). Why not to take a step further and experiment with soft agents that could allow us to engineer the social institutions (Hernández 1999 and López 2002)?. In this paper we examine the emergence of Nash efficient solutions for a CDA market, since it is the dominant institution for the real-world trading of equities, energy, derivatives, etc. and there is a lot of valuable knowledge from E.E. Taking as reference the paper by Walsh et al 2001, we endow the soft-agents with ZI, ZIP, Kaplan and GD ask-bidding strategies. Every class of agents in Walsh, has a fixed strategy. But we allow our softagents to spontaneously change their strategies in an autonomous and evolutionary process. Instead of fixing the agents strategies from the modeller and then calibrating the parameters to achieve Nash equilibrium we look for autonomous emergence of Nash efficiency. Thus extending previous work and introducing evolutionary learning and adaptation. Although our analysis can be programmed in standard object oriented language, we use SDML as programming shell, a cognitive strictly declarative language, since we intend to extend this work to allow for richer learning strategies. At least three mental processes interact in the decision choice: Motivation (the driving force); adaptation; and cognition (reasoning) (Selten 1998). Keywords: Continuous Double Auction. Computational Organization. ABM Social Simulation 279 Development of a Hybrid Multi-Agent Model for Modelling Petrol Price Markets Alison J. Heppenstall, Andrew J. Evans and Mark H. Birkin, School of Geography, University of Leeds, Leeds, West Yorkshire, UK, [email protected] The petrol price market is a highly sensitive and competitive market with many processes combining at different temporal and spatial scales to affect each petrol station’s prices. For example: internal costs (cost of production; fixed costs e.g. staff pay), external influences (price of crude oil and levels of taxation) and effects of locality (rural versus urban areas). Previous models developed to represent the relationship between petrol and a variable are empirical and mathematical. These suffer from a number of problems, chiefly: the parameters are all on the same scale (behaviours executed at the ‘micro’ level are not tied to ‘global’ level variables like oil prices); the parameters are often difficult to estimate and lack realism; very little, if any, account of any geographical effects is taken, and, finally, mathematical models by their nature only consider quantitative parameters and therefore miss out on qualitative, behavioural information. We present a series of three multi-agent and hybrid models that seek to rectify some of these problems by combining the strengths of several techniques. The models vary chiefly in how profitability is assessed. The three systems have an increasing complexity of customer model. An individual-level agent based model of each potential customer would be a vast computational and behavioural modelling effort, so the three systems utilise simplified models of customers. The systems are: A pure Agent based model in which the profitability is decided by simple price comparison. Stations attempt to undercut competitors while maximizing price. A hybrid Spatial Interaction Model (SIM) – Agent system in which customers’ choice of petrol stations is determined by the SIM based on distance to station and price of petrol. Stations aim to maximize their profit based on customer numbers, price, and a fixed volume of fuel that is sold to each customer. A hybrid SIM-Network-Agent system. This is the same as the hybrid SIM-Agent model, except that the customers are initially distributed on the basis of journey to work data along potential routes. This introduces an “intervening opportunities” style component. The performance of each model is evaluated by using idealised simulations and through comparison with the real data. We show that the combination of traditional market-analysis tools, for example SIM’s with the agent-based approach gives a powerful methodology for replicating and analysing the geographical ramifications of such system. Keywords: pricing; multi-agent model; spatial interaction model; networks Dynamics of International Negotiations: A Simulation of EU Intergovernmental Conferences Nicole J. Saam1, Paul W. Thurner2 and Frank Arndt2 1 Allgemeine Soziologie, University of Marburg, Ketzerbach 11, D-35032 Marburg, Germany, [email protected] 2 Mannheimer Zentrum für Europäische Sozialforschung (MZES), Postfach, D-68131 Mannheim, Germany Why do international negotiations fail or succeed? The success story of the constitutionalisation of the European Union seems to be seriously challenged since december 2003 when the head of member states admittedly failed to decide on a proposal on it’s future constitution. Media ventilated the reason to be the obstinate refusal of Poland and Spain to accept a new reweighting of votes. This reweighting would have abolished a compromise of the conference in Nice in 2000. Obviously, previously applied technics of compromising during long negotiations were not successful – or even attempted. However, the public impression of one-shot intergovernmental conference negotiations and the invoking of a single cause of failure are inconclusive. Such negotiations span months of formal meetings and informal coordination between member states. Furthermore, they are not separated from ongoing political business and events - at the international as well as at the domestic level. Therefore, 280 an understanding of negotiation outcomes has to take account of the specific form of the underlying processes and its connectedness to process-relevant determinants. Unfortunately, our knowledge about processes and dynamics of negotiations is scant. Whereas the sophistication of game-theoretic models has increased enormously during the last decade, their model setup remain highly stylized. The identfication of necessary assumptions for a sequential two-player game immediately ending in a unique equilibrium is ingenious and instructive. However, it does not answer our question. Even if there is a convergence of real world negotiations towards the noncooperative result of the Rubinstein model or the cooperative Nash-bargaining solution, we would like to know the actors’ microbehaviour in getting these macro-results. In our lecture, we contribute to a ‘comparative game dynamics’ (Richard 2000). We present the Zeuthen-Harsanyi model of concession behaviour - a predecessor to the Rubinstein model – as a framework for simulating the dynamics of a multilateral, multiple issue, multi-stage and multi-level negotiation system, the EU Intergovernmental Conference of 1996 which led to the Amsterdam treaty. In our model, the 15 EU member states are represented as more or less homogeneous agents who have preferences with respect to 46 issues which are negotiated in parallel processes. The states avoid nperson games in order to economize on transaction costs. Instead, after signalling their preferences, they form coalitions. The negotiations are reconceptualised as Zeuthen-Harsanyi model for two coalitions. In order to evaluate the appropriateness of our simulation model we have calculated several goodness of fit parameters. These parameters are based on different kinds of predicted outcome (simulated or regressed) as well as the empirical outcome - the Amsterdam treaty - which has been operationalised (cf. Thurner/Pappi/Stoiber 2002). Keywords: international negotiations, game theory On a Theorem in the Theory of Guerrilla Warfare Jim Doran, Department [email protected] of Computer Science, University of Essex, Colchester, UK, Agent-based social modelling may be viewed as individual-based modelling of social phenomena using the techniques of artificial intelligence studies. Investigation of guerrilla warfare offensive and defensive strategies is a natural, timely (e.g. ongoing conflicts in Iraq, Nepal), and potentially very important application for this type of modelling involving as it does both social and individual factors, including the cognition of individuals. In this paper I focus in particular upon the following assertion by T.E.Lawrence concerning the conditions under which a guerrilla insurgency is guaranteed to succeed: “Granted mobility, security (in the form of denying targets to the enemy), time and doctrine (the idea to convert every subject to friendliness) victory will rest with the insurgents, for the algebraical factors are in the end decisive, and against them perfections of means and spirit struggle quite in vain.” [T.E.Lawrence, Guerrilla Warfare, Encyclopaedia Britannica, 14th ed., 1929, Vol. 10, p 953] Lawrence’s own “proof” of his assertion is informal and straightforward. Generalising from the circumstances of the 1917/18 Arab Revolt and the Turkish defence of the Hejaz railway, he points to the impossibility of even a large force having the ability to control and defend a very large territory under the stated conditions. In this paper I formalise Lawrence’s assertion, in a simple way, and prove it as a theorem. What may be called the “Supermarket Theorem” is also considered. This asserts that almost certainly the vast majority of people in a supermarket will leave the building within a relatively short time (an hour, say) however diverse may be their beliefs and goals. The relationship between such theorems and agent-based modelling in general is discussed, including linkage to the now established notion of a rule-based “expert system”. A central issue is the extent to which it is possible to discover such theorems directly from the model and without experimental work on the computer. The discussion is illustrated by reference to a programmed generic model of a guerrilla insurgency set on the imaginary island of “Iruba”. Throughout the paper, particular attention is paid to the requisite level of abstraction of a social model, to the “mix” of cognitive and social factors, and to the seemingly crucial role played by belief systems in guerrilla warfare and terrorism. Keywords: guerrilla warfare, asymmetric warfare, agent-based modelling, social simulation 281 Using Multi Agent Systems and Role-Playing Games to Simulate Water Management in Peri-urban Catchments Diana F. Adamatti and Jaime S. Sichman, Laboratório de Técnicas Inteligentes, Escola Politécnica da Universidade de São Paulo Raphaèle Ducrot, Brasil, [email protected], [email protected], [email protected], CIRAD-TERA, Visiting Scientist IEA-USP Peri-urban catchments are characterised by their complexity (hydrological, sociological and geographical processes) as well as their dynamics. This dynamics depends on the multiplicity of stakeholders, with differing and often conflicting land use representations and strategies. To simulate these, it is necessary to have both innovative methods and computer tools to support their coordination and mediation processes, aiming at an improved, more decentralized and integrated natural resources management. Multi Agent Systems (MAS) and Role-Playing Games (RPG) are methods that have been used for some years to represent this dynamics. Computer tools are implemented using MAS and RPG to integrate the conceptual modelling, allowing to represent and to explore functionning and dynamics through testing of different scenarios. An example of peri-urban catchment is the Metropolitan Region of São Paulo, that is studied in the Negowat Project (Negotiation of Land and Water Conflicts in Latin America). We have developed a prototype of a computer tool to integrate water management and the actions of the stakeholders. This prototype, called “JogoMan”, links MAS and RPG, and its implementation is based on Cormas, a Multi-Agent Simulator developed at CIRAD (Agricultural Research Developing Countries of France). JogoMan is a simplification of the real phenomena of interaction between the several actors, and it is used as a means of learning and analysis. The idea of the Negowat Project is to define a complete conceptual model of the area and its implementation in two levels: local and catchment. On the first level, the focus is on the negotiation and interation of stakeholders, and in the second level, the focus is on multi-agent simulations and its effect on the quality of the water. For both levels, the project researchers will need to have strong linking with the local communities, because they can supply much information about the real situation, being crucial for model construction. One of the project objectives is to use representative groups of society better to understand the functioning of the social and political processes, and all research will need a direct interaction with local comunities. In order to, an effect of this approach, some interesting questions related with this modelling and its implementation arises, like how one should model social processes in such an approach? Keywords: Agent Based Modeling; Role Playing Games; Simulation Lagoon, Agents and Kava: A Companion Modelling Experience in the Pacific Pascal Perez1,2, A. Dray1,2, C. LePage1, P. D’Aquino1,3 and I. White2 1 CIRAD, Montpellier, France, [email protected] 2 RSPAS/RMAP, Coombs Bldg, Australian National University, Canberra, ACT 0200, Australia 3 IAC, New Caledonia, France Expanding populations and limited land area in small islands are increasing the pressures on fresh groundwater resources. The dilemma is how to protect shallow groundwater reserves without alienating traditional landowners and without generating costly conflicts. The problem is complex and involves the interaction of hydrologic and technical factors with socio-cultural, economic, policy and institutional factors. Multi Agent Systems (MAS) have been developed to study the interaction between societies and the environment. Here we use MAS in conjunction with a Companion Modelling approach to develop a Negotiation Support System for groundwater management in Tarawa (Rep of Kiribati). In agreement with the complex and dynamic nature of the processes under study, the Companion Modelling (ComMod) approach requires a permanent and iterative confrontation between theories and field circumstances. Therefore, it is based on repetitive back and forth steps between the model and the field situation. 282 The methodology applied in Tarawa relies on three successive stages. First, a Global Targeted Appraisal focus on social group leaders in order to collect different standpoints and their articulated mental models. These collective models are partly validated through Individual Activities Surveys focusing behavioural patterns of individual islanders. Then, these models are merged into a single conceptual one that is further simplified in order to create a role-playing game. This game is played during iterative sessions, generating innovative rules and scenarios. Finally, when the rules become too complex, a computer based version of the game replaces the board version. Stakeholders can explore the possible futures of freshwater management in Tarawa and eventually agree on an equitable collective solution. Keywords: Water management, Multi-Agent System, Companion Modelling, Pacific, Tarawa Agent-Based Simulations backing Use of Role-Playing Games as Dialogue Support Tools: Teaching from Experiments William’s Daré1 and Olivier Barreteau2 1 Cirad Réunion, Cirad Tera/Rev, Equipe GREEN, TA 60/15, 73 rue Jean-François Breton, 34398 Montpellier, Cedex 5, France, Tél: (33) (0)4 67 59 38 49, Fax: (33) (0)4 67 59 38 27, [email protected] 2 Cemagref Montpellier, France, [email protected] A Role-playing game (RPG) was used in a companion modelling approach to facilitate negotiation between stakeholders sharing an irrigated system in the Senegal River valley. This RPG describes a schematic representation of the coordination of farmers to manage water. A Multi-Agent System (MAS), computer version of the RPG, has been used to speed up time of the game and make players started in different conditions of interaction. The aim of this article is to present the issue raised and the limits of the use of a MAS in a such context. After presenting our use of RPG and their place among other simulation games, we analysed their interests and their limits as support tools in negotiation processes. RPG are more powerful than purely computer based Group Decision Support System to support dialogue, since they make stakeholders better handle the simulated impacts of their decision on their system. The first part finishes with a synthesis of issues raised in this context. The second part illustrates how the MAS Shadoc was used to speed up time during RPG sessions. A session is described. Avoid the black-box view of the computer model, by linking its components with the game parameters and introducing the MAS as a response to game limits expressed by the players are some of the conditions necessary to use the MAS in this context. We conclude that (i) the MAS is helpful to the RPG in negotiation processes but its introduction to stakeholders need to be done carefully otherwise models could be delegitimated; (ii) even though RPG and MAS are schematic representation of stakeholders’ behaviours to manage their irrigated system, they do not reduce the complexity of the reality represented. But these models used directly with stakeholders may empower some of them in front of others and the researcher has to assess the impacts of game sessions on the local system. Keywords: Agent-Based simulations, Role-Playing Games, dialogue support tools Actualities of Social Representation: Simulation on Diffusion Processes of SARS Representation Kazuhiko Shibuya, Cyber Assist Research Center, National Institute of Advanced Industrial Science and Technology, 2-41-6 Aomi, Koto ward, Tokyo, 135-0064, Japan, Tel: +81-3-3599-8302; Fax: +813-5530-2067, [email protected] The objective of this paper is to explore social processes on informational representation of severe acute respiratory syndrome (SARS) by means of an multi-agent based model. Especially I attempt to clarify a spatial modelling on a social network. Therefore I will be to explore (a) Social representation in social process, (b) Examine network model of small world, (c) Opinion leader in social networks, 283 (d) Spatial informatics and analysis. The immediate crisis of SARS has already recognized a fact that human population and relationships as network patterns is significant for understanding. I define here that social representation regards as sharing social information and shared cognitive patterns among members in specific opinion groups. Social representation can diffuse and emerge through social communication and social consensual understandings in social network. In this concern, spatial model and simulation have an opportunity to analyze and examine computational process of social informational representation. I would like to show computational results of spatial modeling, and discuss on considerable relative aspects. Keywords: spatial modelling, social network, small world, SARS Part II: Session: Tools and Techniques for Agent-Based Social Simulation MASON: A JAVA Multi-Agent Simulation Library Sean Luke1, Gabriel C. Balan1, Liviu Panait1, Claudio Cioffi-Revilla2, Sean Paus1 1 Department of Computer Science, George Mason University, 2Center for Social Complexity, George Mason University, 237 Robinson Hall MS 3F4, 4400 University Drive, Fairfax, VA 22030 USA, tel (703) 993-1402, fax (703) 993-1399 [email protected] Agent-based modeling (ABM) has transformed social science research by allowing researchers to replicate or generate the emergence of empirically complex social phenomena from a set of relatively simple agent-based rules at the micro-level. Swarm, RePast, Ascape, and others currently provide simulation environments for ABM social science research. After Swarm —arguably the first widely used ABM simulator employed in the social sciences —subsequent simulators have sought to enhance available simulation tools and computational capabilities by providing additional functionalities and formal modeling facilities. Here we present MASON (Multi-Agent Simulator Of Neighborhoods), following in a similar tradition that seeks to enhance the power and diversity of the available scientific toolkit in computational social science. MASON is intended to provide a core of facilities useful not only to social science but to other agent-based modeling fields such as artificial intelligence and robotics. We believe this can foster useful "cross-pollination" between such diverse disciplines, and further that MASON's additional facilities will become increasing important as social complexity simulation matures and grows into new approaches. We illustrate the new MASON simulation library with a replication of HeatBugs and a demonstration of MASON applied to two challenging case studies: ant-like foragers and micro-aerial agents. Other applications are also being developed. The HeatBugs replication and the two new applications provide an idea of MASON's potential for computational social science and artificial societies. Keywords: MASON, agent-based modeling, multi-agent social simulation Simma: A Toolkit for Participatory Business Simulations Xavier Noria1, Nigel Gilbert2, Gilbert Peffer3 1 [email protected] 2 University of Surrey, Guildford, UK, [email protected] 3 [email protected] In the course of a publicly-funded project on business strategy in the on-line news and music sector, a multi-agent based simulation toolkit – Simma – has been developed to aid experts in building market models and to provide end-users with an adaptable and easy-to-use simulation GUI. The toolkit is being designed to support a participatory style of development and implementation using rapid prototyping, design workshops, and user panels involving both modellers and end-users. In an iterative evaluation process, existing commercial and freely distributed tools for implementing multi-agent 284 models were found not to support adequately the divergent needs, objectives, and skills of these two types of Simma toolkit users. In this article, we examine some of the better known MABS toolkits, discussing and comparing their characteristics which are relevant from the vantage point of the Simma users. We then explain the diverse user requirements and how they were translated into design options. The components of the current version of the tool are described and the functionality of the Simma toolkit is illustrated using a simple example. Following the demonstration, we clarify how the original design requirements are met and how the iterative development process was instrumental in this respect. We conclude with an outlook on future plans for improvement, in particular the possibility of a collaborative development of new components for the tool. Keywords: participatory business simulations; multi-agent based simulation, market models The Political Actor Simulator (PAS): The Simulation of a Foreign Policy Crisis during the Peacekeeping Mission in Bosnia 1995 Markus Bresinsky, Projektleiter Human Factors bei der Industrieanlagen-Betriebsgesellschaft mbH (IABG), Einsteinstraße 20, D-85521 Ottobrunn, [email protected] The Political Actor Simulator (PAS) is an agent-based model. Its development is guided by an interdisciplinary approach; the goal is to apply PAS when investigating crisis situations in international politics. This model takes into account both political theory as well as principles of cognitive modelling. Parts of the research project are directed at testing the PAS-model with respect to assumptions derived from political theory. PAS is a moderately complex model for a complex domain and gives an example for the necessity of an interdisciplinary approach in the foreign policy analysis. The research objective of PAS is (1) to characterize the problems arising when modeling short-term decision-making processes, and (2) to provide an example for the modeling of political decisionmaking in a crisis situation. Political science as a discipline has provided a wide range of theories and models about decision-making. However, the influence of psychological science, especially of cognitive science, is rather unusual. The PAS model is an agent-based approach directed at the modelling of political decision-making with a cognitive architecture. PAS is an approach to investigate political decision making by a computer simulation. For this goal the PAS model includes a scenario that refers to events around Srebrenica in the year 1995. The model uses four different agents: Two of them represent the international decision makers on the level of NATO and the UN, another one the level of the local commander of the peacekeepers in Srebrenica, and the fourth one the level of the leader of the Bosnian Serbs. The scenario implemented in the simulation describes a risky political and military situation. A safe area is attacked by the Bosnian Serbs. NATO and UN have to decide about a political and military mandate. If the mandate is decided, the peacekeepers have to implement it. When there is a direct threat to the safe area the peacekeepers can call for air strikes to deter attacks. Whether air strikes will take place or not has to be decided by consensus between NATO and the UN. The results for PAS suggest that a multi-agent based simulation of political decision making is a useful tool for the analysis of crisis development. The experimental data derived from the simulation runs support the position according to which cognitive modeling may enhance the understanding of decision making in political crisis situation. Keywords: multi-agent based simulation, political decision making, cognitive modeling Using GIS and Multi-agent Systems for Interactive Design of Policy in Natural Resource Management Rhodora M. Gonzalez, University of the Philippines, Department of Geodetic Engineering, [email protected] The complex problems of natural resource management (NRM) require correspondingly complex syntheses of insights and tools from the social sciences, natural sciences, and humanities. This 285 includes bringing modern technologies and artifacts closer to people who need them most in alleviating environmental problems. One way to effectively do this is for scientists and laypersons to engage in collectively monitoring change in the environment and making it visible for all concerned to be aware of and hopefully, change their errant ways. With geographic information systems (GIS) and multi-agent systems (MAS) modelling and simulation, the dynamics of the landscape brought about by natural and human activity systems may be formally represented and analyzed interactively. Taking what such technologies can offer into natural resource management processes can lay the groundwork to support interactive design of policy, that is, more direct interplay between science and local perceptions. Possibilities can be better articulated and scrutinized by simulating scenarios before directives are formed; causes and reasons can be identified, and management process and policy can be guided. This contributes a means to devise more effective ways to exchange new and existing knowledge among policy- and decision-makers, and better communication of knowledge to be used by the public. This paper is about an ongoing exploration in harnessing MAS modelling and simulation in analyzing environmental status and designing policy to cope with changes in mountain agriculture. Increased land abandonment, tourism, and invasive tree species are altering the landscape and pose a problem to the community that depends on open grasslands for pasture. A monitoring process using GIS was initially created among scientists and local government to help inhabitants and National Park officers in bringing forth images of the changed environment that they inevitably manage together. MAS modelling techniques and capabilities were further used to better understand how the changes come about and how they are inter-related; simulation results were used in analyzing information about tradeoffs among real alternatives. Lessons are drawn in bringing multi-disciplinary endeavors closer into policy-making. The potentials and weaknesses in harnessing computerized earth observation and simulation technologies for NRM are discussed. Keywords: Multi-agent systems, natural resource management, interactive policy design, GIS, participatory modelling and simulation Multiple Equilibria Regulation Model in Cellular Automata Topology Ioannis D. Katerelos and Andreas Koulouris, Psychology Department, Panteion University, L.Sygrou 136, 17671 Athens, Greece, [email protected], [email protected] In this paper we shall present the Multiple Equilibria Regulation (MER) Model of opinion dynamics in Cellular Automata Topology. As we have stated in our previous article[Katerelos, I. & Koulouris, A. (2004) Seeking Equilibrium leads to Chaos: Multiple Equilibria Regulation Model. Journal of Artificial Societies and Social Simulation, vol. 7, no 2, http://jasss.soc.surrey.ac.uk/7/2/4.html], for certain parameter’s settings the behaviour of the system exhibits transient chaos (unpredictable but rests in a final steady state). In order to approach empirical reality, we introduce a cellular automata topology. One weakness (inherited by the Bounded Confidence Model) of MER model is that every agent is potentially in position to communicate with every other agent if and only if, their opinions differ in a degree less or equal than the bound of confidence. This functional postulate assumes that there is not any geographical (or any other kind of) obstacle excluding the magnitude of bound. Nevertheless, in real life, two (or more) individuals may never communicate just because they will never meet each other. Consequently, in a more realistic scope, we must consider first a topology of the agents in question and then take into account their opinions. Thus, one can hypothesise that a geographical pattern of interaction between agents must be taken into account. Cellular Automata contribution has been proven crucial, since the introduced topology converts the behaviour of our system in a noteworthy way: from transient chaos to “pure” chaos. This is to say that the system is not only unpredictable on the long run but, in addition, it will never rest in a final steady state: prediction is only possible for a short predictability horizon. Keywords: MER Model, simulation, cellular automata, prediction, chaos, opinion dynamics 286 SESSION: INFORMETRIC AND TEXT BASED PROCEDURES FOR INFORMATION RETRIEVAL AND ANALYSIS Organizers: H. Peter Ohly and Maximilian Stempfhuber Contact: Social Science Information Centre; Lennestr. 30, 53113 Bonn, Germany , [email protected], [email protected] Keywords: Citation Analyses, Scientific Productivity and Networks, University Ranking, RetrievalWeighting, Information Mining, Cross-Concordances, Integration of Heterogeneous Data Abstract The aim of this session is to present new insights into formal ranking and linking procedures based on knowledge organization systems or statistical and text analytical algorithms, and to discuss restrictions and possibilities of these approaches in the information field. Of special interest are methods applicable to several of the different types of information which can be found in digital libraries: fulltext documents, bibliographic descriptions, metadata for statistical data, web sites, process produced traces and structural patters of these. Digital libraries with their wealth of different data types and sources rise new challenges when analytical methods for linking or ranking are applied to new types of data or to merge data from heterogeneous sources. In the field of Social Sciences, the integrated retrieval of text-based information, metadata and empirical data plays an important role for the scientific community. How can informetric and text based procedures support this? What standards are necessary to apply them to different data types? How can their results be merged to yield predictable results? Especially for evaluation purposes the quantitative analysis must be enhanced by qualitative decisions and background information. Applications are information retrieval, resp. analyses of literature databases and internet sites, extraction and combining of text information for information purposes, ranking and networking of science actors and knowledge and input-output-analyses of scientific institutions. There should be also space for a critical consideration of indicator quality, reality constraints, and economy of information. Part I: Bibliometric Mining Bibliometric Mining Peter Ohly, Social Science Information Centre, Lennestr. 30, 53113 Bonn, Germany, [email protected] Bibliometrics are known as the analysis of structures and processes in scientific communities by analyzing process-produced library data. Library catalogues, bibliographic online databases, internet pages and more are serving as such a data source. Well-known problems for cross-section analysis are the definition of universe, the definition of a unit, the sampling procedure, and the validity relation. Mining tries to retrieve individual hidden information by selecting, aggregating or linking information. Though largely dependent on thresholds and logic structures the questions of appropriate evaluating, ranking and comparing remain. Content related information - the greatest part in bibliometric considerations - is even more sensitive and must be value-added by semantic and text-statistical relationships. Keywords: bibliometrics, informetrics, information mining, information retrieval Automating the Analysis of Co-Words in Contexts: Measuring the Meaning of “StemCells” and “Frankenfoods” in Different Domains Iina Hellsten and Loet Leydesdorff, University of Amsterdam, ASCoR, Kloveniersburgwal 48, 1012CX Amsterdam, The Netherlands, [email protected], [email protected] Co-words have been considered as carriers of meaning across different domains in studies of science, technology, and society. Words and co-words, however, obtain meaning in sentences, and 287 sentences/messages gain their meaning in contexts of use. At the science/society interface words and co-words can be expected to have different meanings at the different sides of the interface. For example, whereas ‘energy’ is strictly defined in physics, this concept is often confused with ‘energy carriers’ in public debates about energy shortages. In other words, these different sides use different codes of communication for providing meaning to words. Furthermore, meanings may change over time. Given these differences in meaning in different domains and over time, we search for a reflexive mechanism that enables us to study co-words in contexts. The empirical questions are: To which extent can words and co-occurrences of words carry meaning across domains? What could be considered as reflexive mechanisms for the function of the translation? How can translation mechanisms between different codes of communication be operationalized? In the sociology of translation and actor-network theory, the specification of the domains followed the translation (as a construct). In our design, the contexts (domains) are first operationalized independently of the specific words carrying meaning at the interfaces: one can compare the usage of words in a newspaper with the usage of that same word in scholarly discourse. We distinguish between two types of metaphors as reflexive mechanisms for the comparison of co-words across the domains: symbolic metaphors and words as messengers of meaning. “Energy,” for example, functions as a metaphor in the above case although it is a common word. The study is based on our earlier work: one of us has studied symbolic metaphors (e.g. “Frankenfood”); the other author has focused on how distributions of specific control terms (e.g., “stem-cells”) relate textual domains in a sub-symbolic mode. Both are more codified than average words and co-words, and thus they may serve as tools for “translation.” We will analyze two debates (on gm-foods and on stem-cell research) that have recently flourished at the science/society interface. We will first show techniques and routines that we developed to download the data available in webbased sources such as search engines, the Web-of-Science database, electronic archives of newspapers, and journals. Secondly, we will demonstrate our techniques and routines for conducting automated text analysis on this data, and finally, the procedures for visualizing the data as semantic maps, based on the co-occurrences and proximity of the words used. The presentation contributes to the quantitative text analysis tradition, and points to new avenues that relate the study of co-word analysis in context with the sociological quest for the analysis and processing of meaning. Keywords: codification, translation, mapping, metaphor, co-occurrences, interfaces Mining Document Contents in Order to Analyse a Scientific Domain Josiane Mothe, Bernard Dousset and Désiré Kompaore, Institut de Recherche en Informatique de Toulouse, 118 Route de Narbonne, 31062 Toulouse CEDEX 04, France, [email protected] Large-scale analysis to understand the state of the art and the evolution of scientific communities and research topics become possible thanks to the availability of large sources of scientific literature. Number of methods are used to achieve this type of analysis. Co-citation analysis is among the most popular as it allows many different types of highly informative analysis to discovery the structure of a domain. A co-citation can be extracted when two references appear in the same published paper. When Journal co-citation is analysed, information on how two journals are co-cited and the relationships among journals can be extracted. On the other hand, co-authoring analysis can also be used to identify groups of scientists and their inner structure. Free text analysis provides complementary information. Mining the document content and combining it with other meta data such as author, affiliation, date of publication informs on the topics of interest and their evolution. Such domain mapping is possible when using various data analysis methods. Classification and factorial analysis are among the ones we use when analysing a domain. In this paper we will present the approach we developed as well as the system based on this approach (named Tétralogie). It is composed of several modules that implement a knowledge discovery process. Such a process begins by selecting data or documents from one or several sources on a targeted domain. Then the documents are pre-treated in order to extract the useful information (meta data and concepts from the free text). This information is then mined using different data analysis methods 288 including Hierarchical Agglomerative Clustering and Correspondence Factorial Analysis in order to find hidden information or knowledge. The results are presented under the form of multi-dimensional and interactive views. In this paper, the overall process will be presented through a case study. Keywords: information mining, data analysis, domain knowledge Part II: Information Analysis Comparing a Knowledge-Based Economy: In the Case of South Korea and the Netherlands Han Woo Park1, Loet Leydesdorff2, Heung Deug Hong3 and Sung Jo Hong4 1 YeungNam University, South Korea,[email protected] 2 University of Amsterdam, The Netherlands, [email protected] 3 Miryang National University, South Korea, [email protected] 4 Dongguk University, South Korea, [email protected] The competitive advantages of a knowledge-based economy can no longer be attributed to single nodes in the network. The ‘knowledge infrastructure’ of a nation-state can be operationalized in terms of networks. For example, Etzkowitz & Leydesdorff (1997) proposed to consider the networked knowledge infrastructure in terms of a triple helix of university-industry-government relations. For the study of the structure of a knowledge-based economy, this paper proposes a triangulation strategy. More specifically, ‘knowledge infrastructure’ in the three institutional domains can be measured using different indicators: scientometric, patent, and webometric indicator. This methodological strategy is particularly useful in order to examine the networks among universities, industries, and governments in terms of their effects on the development of science, technology, and innovation. The paper employs this triangulation strategy to examine the current state and problems of South Korea and the Netherlands through the comparison of the respective knowledge bases of the two countries. The Netherlands can be used as a benchmark for the comparison with South Korea because of its rapid rise as a global hub of the (scientific and innovative) knowledge and (digital) information community. A comparative study of the knowledge bases of a national economy in the Asian and Western regions, respectively, has hitherto not yet been covered by existing studies, since these have been focusing on developments in North-American and/or Western-European nations. Furthermore, past research has usually dealt with conditions that encourage the spread of knowledge-based dynamics (e.g., ICT) in the economy across nations. Little attention has been paid to explaining the differences in the endogenous development of knowledge-based economies among countries. This study tries to fill this gap in the existing literature. Keywords: scientometrics, webometrics, international comparison, science, technology, innovation, knowledge, triple-helix Graph-Based Representations of Motion Pictures: An Attempt to Discover Statistical Features of Aesthetic Merit Gregory H. Leazer, Jonathan Furner and Rachel Napper, Department of Information Studies; University of California, Los Angeles, USA, [email protected]; [email protected]; [email protected] Traditional information retrieval (IR) approaches such statistical representations of term frequency and distribution have limited application in the retrieval of motion pictures. The traditional approaches are most successful with the retrieval of textual documents with relatively stable and homogeneous linguistic features where the terms used in a document are an accurate representation of the document's topic. While this paper does not argue for a broad and rigorous distinction between works of fiction and non-fiction, topicality is a less significant feature with general literary or artistic documents, where narrative and other aesthetic features become the principle components for 289 characterizing a document. This is especially true for cinema, where textual components (in the form of caption data) will have even less use as it is supplemented by non-textual features such as imagery and music that cannot be represented by typical term-based models. We describe the use of three alternative methods for clustering films in an IR system. A large filmperson incidence matrix is generated by listing the principle cast, directors, producers and screenwriters associated with each film. These social features provide a partial representation of the artistic merit of a film because the people associated with a film are an accurate (though partial) representation of a film's aesthetic characteristics. These attributes are used to measure distances for all pairs of films by creating a distance matrix: any two films are considered to be adjacent if there is any overlap in the people associated with both films. The distance between each pair of films is modelled as a graph and measured by the shortest path used to connect them through their adjacent members, using a technique that is similar to those used in the analysis of small-world social phenomena. The second and third distance measures are our innovation and involve the creation of a similarity matrix and the creation of a graph that expresses the amount of overlap in the people associated with any two films using Dice's coefficient. A "product distance" matrix is derived that expresses the distances between any two films based on the product of the similarity weights on a path that connects those films. We also derive an "accumulative difference distance" matrix for which the similarity weights are subtracted from one. Distances are based on the sum of these difference weights on a path that connects two films. The distance, product distance and accumulative difference distance matrices are applied to the Internet Movie Database (IMDb) and described. Empirical findings include the observation of smallworld networks consistent with findings from other studies of collaborative activity. We also report how the three distance measures are used to generate rankings from a user-defined set of relevant films. Keywords: social networks, information retrieval, motion picture retrieval, document clustering Using Search Engines and Web Crawlers in Social Science Research Mike Thelwall, School of Computing and Information Technology, University of Wolverhampton, Wulfruna Street, Wolverhampton WV1 1SB, UK, [email protected] The importance of the Web for publishing and seeking information is a powerful argument for the need for social scientists to develop sound methods to analyse it. In this paper I will introduce Webometrics, quantitative web research that has grown out of infometrics. Webometric research is concerned with measuring web phenomena and using the results to make inferences about characteristics of the pages or their creators. Hyperlinks have been the most common object of study, but there has been little agreement about what they should represent. I will give a brief summary of one strand of hyperlink research and then draw out some methodological lessons for quantitative web research. Experts in citation analysis have conjectured that hyperlinks may be usable in an analogous way to citations: counting links to web pages/web sites/countries could be used to measure their online impact or to track the flow of scholarly communication. After many studies of this issue, hyperlinking between university web sites has been shown to be very different to citations in patterns of use. The findings, summarised below, point to the care needed to interpret any hyperlink data. Why are inter-university links created? Relatively few, probably less than 1%, are directly equivalent to citations in the sense of connecting two scholarly pieces of work equivalent to journal articles or conference papers. Many links are purely symbolic, not referring to information but acknowledging a relationship between the source and target university, such as joint membership of a research project. About 90% of inter-university hyperlinks indicate some kind of scholarly connection, however, with the rest relating to recreational, administrative or support activities. Are inter-university link counts useful? Counts of links to a university can be used to estimate its research productivity, or, if this is known, can be used to assess whether its web presence is at the level to be expected for its research productivity. 290 For those analysing hyperlinks outside of academic sites, the most important lesson that can be learned from this research is the need to assess the meaning and value of link counts through (a) investigations of the purpose of random samples of individual links, and (b) correlating link counts with related measures that are hypothesised to be similar. Both approaches are needed: researchers should not assume that obvious reasons for link creation are the ones that are used in practice. Keywords: web, informetrics, webometrics, hyperlink, scholarly communication Part III: Digital Libraries Digital Libraries and Heterogeneity Maximilian Stempfhuber, Social Science Information Centre, Lennestr. 30, 53113 Bonn, Germany, [email protected] Digital – or virtual – libraries try to combine information from various sources and in different formats. Metadata play an important role in searching for this information, especially in situations where the original data (e. g. full-text documents or survey data) are not directly accessible for searching. In addition to finding and harvesting metadata, the integration of this differently structured and encoded metadata in a single retrieval model is not an easy task as well. Standardization is one way of dealing with heterogeneity and complexity, but models have to be provided for the remaining part that can not be standardized. The ways of dealing with this heterogeneity are methods of intelligent mappings between metadata schemes or knowledge organization structures. How, when and to what extent such mappings can be constructed automatically and for what type of data or domain they can be applied successfully, has to be evaluated. Keywords: digital libraries, meta data, information retrieval, heterogeneity Searching and Browsing Multiple Subject Gateways in the Renardus Service Michael Day, UKOLN, University of Bath, Bath BA2 7AY, United Kingdom, [email protected] This paper describes the main features of the Renardus cross-search and cross-browse service. Renardus is made up of a number of participating subject gateway services. Cross searching is based on the Z39.50 protocol. A review of data models in use by partner services helped define a minimum set of Dublin Core-based metadata elements that could be utilised as a common model for the Renardus service. This provides the basic infrastructure for interoperability between all participating gateways. The Renardus Service also uses classification mapping to enable subject browsing across all gateways with the Dewey Decimal Classification (DDC) across. The paper outlines the use of classification systems by Renardus partner gateways, the general mapping approaches taken by the project, the definition of mapping relationships, and provides some information on technical solutions. There follows a demonstration of how the mapping information is used within Renardus to support browsing and advanced searching and of several features that have been implemented to aid end-user navigation in deep subject-browsing structures. Keywords: information retrieval, metadata, subject gateways, cross-search, classification mapping Ontology Switching for the Social Sciences. Methods for the Underlying Corpus Analysis Robert Strötgen, University of Hildesheim, Institut für Angewandte Sprachwissenschaft [email protected] Semantic heterogeneity can be defined as the use of different documentation languages or ontologies for different document sets in digital libraries or other integrated document collections. This is also an important issue for the “Semantic Web” that aims to integrate documents from the Internet indexed with different metadata sets. A user searching these collections may miss documents because he does 291 not use the right search terms. Automatic translation between terms of different ontologies (“ontology switching”) is one approach to solve this problem. The underlying statistical methods based on corpus analysis (co-occurrence of terms) allow the automatic identification of semantic relations (or “crossconcordances”) between terms from different ontologies. These methods require parallel corpora with documents that are indexed with terms from different ontologies. For the social sciences suitable corpora are very rare. Most documents are even completely missing any metadata or subject description with controlled vocabulary in particular. Therefore, special modifications of the ontology switching approach help to deal with this lack. One alteration is simulating a parallel corpus for un-indexed document collections using machine learning (i.e. Bayesian networks) and linguistic methods like stemming, morphology, part-of-speech tagging and named entity recognition. Another change is translating not between two ontologies but between an ontology and suitable free text terms. This demands much more effort because usually much more free text terms than keywords exist in a document and they are uncontrolled. In order to diminish the computation cost and to find useful relations some preparations on the corpus are necessary. This talk will introduce into the concept of ontology switching and describe the special scenario and experiments for the social sciences in more detail. As an outlook to further work, this domain will be compared with the domain of patent documentation. Keywords: information retrieval; semantic heterogeneity; semantic web; ontologies; machine learning Bilateral Transfer Modules for the Conceptual Integration of Heterogeneous Document Bases in Content Analysis Xueying Zhang, Nanjing [email protected] University of Science and Technology, Hong Kong, Nowadays users are confronted with a more highly decentralized and heterogeneous information space than before. Although integrated thesauri and standardization efforts have been widely proposed as the best methods for establishing compatibility and convertibility of indexing languages for over 40 years, they are no longer sufficient to achieve the desired interoperability and efficiency. Transfer modules are presented to interlink document sets flexibly before the vagueness treatment between documents and query terms. In contrast to costly cross-concordances, the proposed transfer modules in this paper are developed by the combination of the dependency coefficient in rough set theory and the indexing specificity in information languages. The experiment indicates that these proposed modules could not only designate explicitly semantic relationships but also find more correlated terms, especially with low frequency of occurrence in the corpora. Further, sets of rules are introduced to filter sensible results and to integrate results generated at different conceptualization levels. In addition, some principles are presented to determine the maximum number of combination terms in creating “one to multiple” relations and “multiple to multiple” relations. Finally, the experimental results are evaluated by quantitative measures and compared with implemented cross-concordances and other statistical measure. Keywords: rough set theory; transfer module; indexing language; compatibility; content analysis, heterogeneity 292 SESSION: TIME SERIES ANALYSIS Organizor: Rainer Metz Contact: Central Archive for Empirical Social Research, University of Cologne, Liliencronstr. 6, D50931 Cologne, Germany, Tel. (49) 221 4769436; -34 (Secretary), Fax (49) 221 4769455, [email protected] Keywords: time series analysis; trends, cycles, structural breaks, outliers; stochastic processes; causal and intervention analysis. Abstract For the statistical analysis of social and economic processes methods of time series analysis are indispensable. This methods allow the identification of patterns of development (trends and cycles), of irregularities of common features and causal relationships. The most important basis for the statistical models is the theory of stochastic processes. The session tries to bring together scholars whose starting point is a concrete problem in socio-economic analysis for which they use specific times series models. The contributions therefore should be not merely theoretical but oriented towards socioeconomic applications. Comparison of Seasonal Variation of Births and Deaths in Urban and Rural Areas of Silesia (Poland) Zofia Mielecka-Kubien, Department of Statistics University of Economics, 40-226 Katowice, Poland, ul. Bogucicka 14, [email protected] Seasonal variation can be frequently observed in socio-economic time series. Because of different size of populations and some formal properties of time series (additive or multiplica-tive relation among their components) the range of seasonal changes is in many cases difficult to compare. On the other hand there is often need of comparison of seasonal component in different populations or in time series of different kind. In the paper a simple method of standardization and estimation of seasonal indices is proposed. In many cases the method enables comparisons of seasonal variation, and the results can be clearly interpreted. Theoretical considerations are illustrated by comparisons of seasonal cycles of birth and death processes in the Silesian part of Poland between urban and rural subpopulations in the recent years. Significant differences in the seasonal changes of the considered time series were found. Keywords: seasonal component, births and deaths, comparisons Population at Risk (PAR) Rates and other Denominators used to Measure Unemployment Rates Ray Thomas and Hasan Al-Madfai, 35 Passmore, Tinkers Bridge, Milton Keynes MK6 3DY, UK, Tel 44 1908 679081, Fax 44 1908 550401, [email protected] Traditional measures using population denominators treat unemployment as a stock and indicate the importance of unemployment as a social and economic condition, but fail to take account of unemployment duration. Population at risk (PAR) unemployment rates measure flows between unemployment duration groups and can be used to summarise the scale of inheritance of unemployment between successive duration groups. PAR rates indicate labour market response to unemployment through the use of time as a denominator. Times series for Claimant (registered/insured) unemployment for Britain for the period 1985 to 2003 show declining PAR exit rates with increased duration of unemployment consistently through the 293 trade cycle. The stability of this relationship supports the treatment of duration as an explanatory variable of the composition of unemployment duration groups and casts doubt on appropriateness of the use of duration as an explanatory variable for investigation of exit probabilities in unadjusted hazard functions. A variability decomposition approach taking into account the inheritance between successive duration groups is used to deal with strong seasonal variation. PAR rates for short duration unemployment groups show earlier but less sensitivity to changes in labour market conditions than longer duration groups. This difference supports the distinction between 'frictional' and 'structural' unemployment in Beveridge's classic typology and denies the validity of hysteresis theories associated with the growth of long term unemployment in the 1980s. The paper examines the pros, cons and interpretations of unemployment rates using population and time denominators. The topics covered include the use of different rates for policy making and forecasting, and the availability and timeliness of supporting data on ILO unemployment from labour force surveys and on registered and insured unemployment from administrative systems. The paper discusses possibilities of application of PAR approach to other datasets relating to population stocks. Keywords: unemployment, population at risk, denominators, seasonal variation, stocks, flows Auto-regressive Time Series Cross Section Analysis (TSCS) for Proportions Data with Random Coefficients. A Flexible Model and its Application in Regional Science Michael Windzio, EMPAS, University of Bremen, Germany, [email protected] In many studies of sociological, economical, political and geographical science, units of observation are spatial units such as countries, regions and constituencies. Moreover, dependent variables measured as proportions data are very common in these fields. Focusing on change over time and using data consisting of time series of each spatial unit, the structure of data and the family of statistical models are called “Time Series Cross Section” (TSCS) Analysis. In most cases, the error structure of these models must account for first or higher order autoregressive residuals. This paper deals with the question of how to estimate such models in a flexible way, whereby two issues are being elaborated: The first issue concerns the problem of how to correctly model proportions data. This kind of data easily leads to similar pitfalls as binary state space data does, if modelled as a dependent variable. Following W. H. Greene, the solution is similar as well: the dependent variable can be transformed via the logistic link function and the model linearly estimates effects on the log odds of the dependent variable. Secondly, the dependent variable’s process of change over time is of particular interest. But assuming, however, that every spatial unit undergoes the same curve of change over time may be unrealistic. In order to relax this assumption, the TSCS Analysis can be generalized to a random coefficient model. Linear and non-linear effects of time can be modelled in a flexible way which allows unique curves of change over time for each spatial unit. The advantage of the random coefficient model is twofold: it offers a solution to the problem of the heterogeneity of units being a nuisance, since the flexible treatment of temporal change profiles avoids misspecification (e.g. if a linear process of change does not fit to all units). Furthermore, it allows the heterogeneity of units to become a matter of interest because each profile of change can be plotted over time and compared with other profiles. This comparison could result in a typology of change profiles. The paper offers a short outline of the Random Coefficient TSCS model. Moreover, some empirical results are presented in order to underline the flexibility of this model. Keywords: random coefficient model, TSCS Analysis, longitudinal data, panel analysis 294 SESSION: THE STATISTICAL ESTIMATION OF LINEAR AND NONLINEAR STOCHASTIC DIFFERENTIAL EQUATIONS Organizor: Hermann Singer Contact: Lehrstuhl für Angewandte Statistik und Methoden der empirischen Sozialforschung, FernUniversität Hagen, Tel.: 02331 987 2615, [email protected] Keywords: Stochastic differential equations; Parameter estimation; Discrete time measurements; Sampling; State space models; Filtering Abstract Stochastic differential equations (SDE) describe the time evolution of quantitative state vectors in continuous time. In the social sciences and economics, however, only discrete time measurements are available (e.g. daily, weekly, quarterly etc.). Then, the drift and diffusion parameters of the continuous time system (SDE) must be estimated on time series or panel data. The transition probability between measurements, as required for the likelihood function, can be computed exactly only in the linear case (LSDE), whereas for nonlinear systems, (in general) only approximations can be obtained. More generally, a measurement equation can be joined (continuous-discrete state space model) in order to model measurement error and latent variables. The session will invite papers on the parameter estimation of SDE and state space models including - Application of linear systems (LSDE) - Linearisation (Extended Kalman filter EKF etc.) - Analytical approximations - Monte Carlo approximations - other numerical approaches A Survey of Estimation Methods for Stochastic Differential Equations Hermann Singer, FernUniversität in Hagen, Germany, [email protected] This article gives a short review of key issues and of existing estimation methods in differential equation modeling. It covers both linear and nonlinear stochastic differential equations (SDE) and continuous-discrete statespace models including panel data. Keywords: Ito calculus; sampling; exact discrete model; transition densities; extended Kalman filter; nonlinear filtering A Comparison of Different Procedures to Estimate the Damped Linear Occilator for Panel Data J.H.L. Oud, University of Nijmegen, The Netherlands, [email protected] There is a growing interest in the application of the damped linear oscillator in panel research (Boker, 2002; Oud, 2002; Singer, 1993, 2003) of partially observed systems. Examples in psychological research are discussed by Boker. The damped linear oscillator is a special case of differential equation modeling, which since Newton is the standard approach of dynamic phenomena in natural science. Different procedures, exact and approximate, have been used to estimate the damped linear oscillator. In his software package LSDE (Linear Stochastic Differential Equations), Singer used the exact discrete model (EDM) and the expectation-maximization (EM) algorithm in combination with a quasiNewton algorithm. Oud used also the exact discrete model but in combination with structural equation 295 modeling (EDM/SEM procedure). In addition to this procedure, which requires highly nonlinearly oriented SEM software, Oud also proposed the less demanding approximate ADM/SEM procedure. A final approximate procedure was proposed by Boker: the multivariate linear differential equation (MLDE) procedure. The four procedures will be compared in a simulation study and their advantages and drawbacks expounded. Keywords: damped linear oscillator, stochastic linear differential equations, panel data. Multivariate Multilevel Models of Dynamical Systems using Latent Differential Equations Steven M. Boker and Stacey S. Poponak, University of Notre Dame, USA, [email protected] Multivariate longitudinal data frequently measures systems of variables that both change over time and influence one another over time. One method for dealing with systems that have an intrinsic dynamic, i.e. which influence their own level over time, is to model the variables as a differential equation. Such models are particularly well suited for systems that have a "set-point" or equilibrium around which the variable fluctuates. A simple form of such a model is a second order linear differential equation that under some parameter constraints becomes a damped linear oscillator. Suppose two variables are measured longitudinally and mutually influence one another's behavior. Such a system of variables can be modeled as simultaneous differential equations with coupling between them. Each variable regulates its own behavior as well as the behavior of the other variable. simple metaphor for this type of system is two pendula with a spring between them. Consider the example of feelings of intimacy between a husband and wife. For each person, these feelings are likely to fluctuate from day to day around some equilibrium value. The husband's feelings of intimacy are likely to have an effect on the wife, and her feelings of intimacy are in turn likely to influence her husband. These individuals are unlikely to have the same equilibrium values or the same model parameters as one another. Each husband and wife might have their own equilibrium level and each married couple might have their own frequency and strength of coupling of their feelings to one another. We propose a model and method that allows estimation of a mixed effects latent differential equation. This method simultaneously (a) estimates latent variables from multiple indicators, (b) estimates the derivatives of these variables, (c) estimates effects between these variables, and (d) allows these effects to vary across groups (e.g. married couples). We use the structural equation modeling program Mx to fit this mixed effects latent differential equation model to a sample set of daily diary data of self-reported feelings of intimacy from a study of marital social relations. The advantages and disadvantages of this approach over standard multigroup latent growth curve modeling are discussed. The primary advantages accrue when (a) one does not have a true zero for time against which to align all groups or (b) one expects that exogenous influences will disrupt stationarity. Keywords: dynamical systems, random coefficients, latent variables, differential equations. Continuous-time Modeling of Bi-directional Influences between Family Relationships and Adolescent Problem Behavior Marc J. M. H. Delsing In the behavioral sciences, the cross-lagged panel design has become an increasingly popular tool to disentangle the over-time influences two variables have on one another. Traditionally, cross-lagged panel analyses are performed in discrete time. This approach, however, has several important disadvantages. A major problem, for example, is that results from discrete-time analyses are highly dependent on the length of the time interval between measurements. Consequently, results of researchers using different time-intervals are difficult to compare and often seem to be contradictory. The problems associated with discrete-time analysis can be solved by means of continuous-time 296 analysis, using stochastic differential equations. The present contribution presents an application of this continuous-time modeling procedure to the assessment of the bi-directional influences between family relationships and adolescent problem behavior. Estimating Dynamical Systems Disturbed by Noise Lutz-M. Alisch and Alexander Robitzsch, [email protected] Technische Universität Dresden, Germany, In order to ascertain whether a sample of curves under fluctuations is a realization of a dynamical system disturbed by noise, it is convenient to look at two different approaches. Given a sample of curves such that they are n realizations of a stochastic process, our concern is to identify an SDE (first or second order) with solutions which would converge to the observed trajectories sufficiently strong. The main idea here is to estimate an SDE from one trajectory and to show that the other curves from the sample correspond to the stochastic flow generated by the SDE. From a second point of view it is possible to estimate the (second order) differential operator via a method developed out of the statistical toolbox of functional data analysis (FDA) called random principal differential analysis (RPDA) which allows one to study the space of solutions. In detail we will discuss the estimation of an SDE of the form dX_t= b(t,X_t) dt + sigma(t, X_t) dW_t with b and sigma unknown coefficient functions. We illustrate the estimation procedures with respect to empirical data. Then we compare the results with those obtained by using parametric SDE-techniques, hierarchical modelling and PDA (to get an estimation of an ODE with time varying coefficient functions). Moment Equations and Hermite Expansion for Nonlinear Stochastic Differential Equations with Application to Stock Price Models Hermann Singer, FernUniversität, Hagen, Germany, [email protected] Exact moment equations for nonlinear Itô processes are derived. Taylor expansion of the drift and diffusion coefficients around the first conditional moment gives a hierarchy of coupled moment equations which can be closed by truncation or a Gaussian assumption. The state transition density is expanded into a Hermite orthogonal series with leading Gaussian term and the Fourier coefficients are expressed in terms of the moments. The resulting approximate likelihood is maximized by using a quasi Newton algorithm with BFGS secant updates. A simulation study for the CEV stock price model compares the several approximate likelihood estimators with the Euler approximation and the exact ML estimator (Feller, 1951). Keywords:Stochastic differential equations; Nonlinear systems;Discrete measurements;Maximum likelihood estimation; Moment equations;Extended Kalman Filter;Hermite Expansion STREAM: GENERAL STATISTICS SESSION: ASSESSMENT OF MODEL FIT Organizor: Tamas Rudas Contact: Department of Statistics, Faculty of Social Sciences, Eotvos Lorand University and TARKI, Budapest, fax: + 36 1 372 2912, [email protected] Abstract This session is a continuation of a similar one held at the conference in 2000 in Cologne. Statistical models are routinely used in data analysis. In some situations, a formal test of fit, and in some others the measurement of the goodness-of-fit is of interest. Assessment of model fit refers to both activities and it is carried out under a number of varying circumstances, and technical and philosophical 297 frameworks. These include the effects of different sampling schemes and sample sizes, assumptions concerning data quality, specifications regarding subject matter significance, the use of computer intensive methods, Bayesian or frequentist frameworks, etc. Abstracts of proposed contributions to the session – theoretical, methodological or applied – are welcome in all aspects of assessment of model fit. Papers emphasizing particular problems – and hopefully proposing solutions – relevant in social science applications are particularly encouraged. Each speaker in the session is expected to have about 20 minutes for presentation and if the talks (or, at least, part of them) will address related issues, a short discussion may also be organized. Authors are asked to submit their abstracts to the session organizer by e-mail. Analysis of Deviance with Penalized Likelihoods David Firth, University of Warwick, UK, [email protected] In social science as in other areas of application, maximum penalized likelihood is often a valuable alternative to maximum likelihood, for fitting statistical models to data. Examples include: ridge regression and its generalization to non-linear models; models (such as generalized additive models) which include smoothing-spline components; reduced-bias estimation of generalized linear models; and pseudo-Bayesian incorporation of prior information on model parameters. Important applications include model-based estimation of small-area statistics from national surveys, and small-sample logistic regression analysis. The standard analysis of deviance for model assessment and model choice, based on a sequence of likelihood-ratio tests, does not apply directly when the likelihood is penalized. Similar considerations apply also to information criteria such as AIC. The problem will be reviewed and possible solutions explored, with particular emphasis on the most difficult case where the penalty is not a function of parameter values alone. Keywords: ridge regression, shrinkage, bias reduction, logistic regression, orthogonality Examining the Scalability of Intimacy Permissiveness in Taiwan Pei-shan Liao and Su-hao Tu, Center for Survey Research, TYRC, Academia Sinica, Taiwan, [email protected], [email protected] The issue of permissiveness toward premarital sexual or intimate behaviors for different age groups, gendered subjects, at different levels of intimate relationships, has been a focus among researchers. Considering the progressive nature of intimate behaviors, the items designed to measure attitudes toward premarital intimacy permissiveness have been found to form a reliable Guttman-type scale. However, few studies have thoroughly examined the appropriateness of applying such Western-driven measures to different social contexts and cultures, particularly in societies that are generally conservative toward sexual attitudes. This paper aims to examine the scalability of intimacy permissiveness measures in Taiwan by using island-wide survey data. The premarital sexual or intimacy permissiveness scale (PSP or IPS), including items used in previous studies, has been widely used and with little attention to its reliability and validity. Revisions of these scales have, nonetheless, lacked methodological examination. A reliable premarital intimacy permissiveness scale is needed to ascertain whether the attitudes of people in Taiwan toward various forms of intimate behavior remain conservative. Among models for scaling analysis, Guttman scaling model and two latent class models —the latent-distance model and Goodman’s scale model— were employed to examine the scalability of the scale measures. The data were drawn from the 2002 Taiwan Social Change Survey (Questionnaire No. 2), conducted by the Institute of Sociology, Academia Sinica. The respondents for this survey were randomly selected by using stratified sampling and face-to-face interviews. A total of 1,983 completed interviews were obtained. This study found that the premarital intimacy permissiveness measures were able to form a reasonable Guttman-type scale based on three scaling models. In order to further 298 validate such a scale, group comparisons were conducted between gender and age cohorts. Gender and age differences were found consistent with previous studies and mostly reconfirmed the appropriateness of the PSP measures designed for the Taiwan Social Change Survey. Keywords: Intimacy permissiveness, Guttman scale, latent class models, scalability Additive and Multiplicative Effects in a Fixed 2 x 2 Design using ANOVA can be difficult to Differentiate: Demonstration and Mathematical Reasons Johannes A. Landsheer1, Godfried van den Wittenboer2 and Gerard H. Maassen3 1 Department of Methodology and Statistics, Faculty of Social Sciences, Utrecht University, The Netherlands, [email protected] 2 Department of Education, University of Amsterdam, The Netherlands 3 Department of Methodology and Statistics, Faculty of Social Sciences, Utrecht University, The Netherlands Several researchers have expressed doubts as to whether ANOVA is fit to detect interaction in a fixed 2 x 2 design. To study this, values of a dependent variable are created with linear construction formulae, with and without a product term. Our simulation demonstrates that when the dependent variable results from a product term, ANOVA often fails to detect a significant interaction effect, while at the same time two strong additive main effects are indicated. As a consequence, multiplicative models are rejected in favor of an additive model. Mathematically it is shown that the ANOVA results can indicate strong main effects as the result of a solely multiplicative contribution of both factors. It is recommended to abandon the fixed ANOVA model for the study of interaction effects and use a regression model with assessed level values of each independent variable instead. Keywords: ANOVA, data modelling, simulation, interaction, power, interpretation How to Evaluate the Fit of Linear and Generalized Multilevel Models Wolfgang Langer, Martin-Luther-Universität Halle-Wittenberg, Institut für Soziologie, Germany In the early nineties Bryk & Raudenbush (1992,2002) and Boskers & Snijders(1994,1999) have proposed two methods of R2s to evaluate the fit of the between- and within-context-regressions, using the principle of proportional reduction of error. Nobody uses the pseudo-r2s proposed in literature (for instance Langer 2000) to evaluate the fit of hierarchical generalized regression models like the binary or ordinal logit/probit model. I suggest the maximum-likelihood-r2 developed by Maddala (1986) and Long (2000) to assess the overall fit of hierarchical linear models. To evaluate the fit of binary and ordinal multilevel logit regression models I propose the Aldrich&Nelson pseudo-r2 (1984) with the Veall&Zimmermann correction (1992,1993,1994). I present two empirical examples using data from educational and victimization research to show how to apply these pseudo-r2s to real data. Collinearity involving Ordered and Unordered Categorical Variables John Hendrickx1, Ben Belzer2, Manfred te Grotenhuis2 and Jan Lammers2 1 Freelance, [email protected] 2 Department of Methodology, Nijmegen University, The Netherlands, [email protected], [email protected], [email protected] Belsley (1991) proposed evaluating collinearity in regression models by adding random disturbances to suspected variables and assessing the consequences for parameter stability. This paper extends this basic principle to models with categorical variables. Categorical variables are randomly misclassified such that the destination category is random for unordered variables and near the original category for ordered categorical variables. This technique can be applied to nonlinear models such as logit and probit as well as linear regression. Macro programs are presented for implementing the technique. Keywords: collinearity, regression, categorical, assumptions, nonlinear 299 Testing the Fit of Conditional Independence with a Continuous Control Variable Wicher Bergsma, Eurandom, University of Eindhoven, The Netherlands, [email protected] Many tests of independence between two random variables are widely available, for example, the chisquare test, tests based on the correlation coefficient and Kendall's tau, or Hoeffding's test of independence. In this talk a method is presented using which these tests can be also applied to the testing of conditional independence with a (possibly continuous) control variable. Briefly, the question is addressed regarding the generalization of the method: how to test whether a given unconditional model fits for all the values of a control variable. Assessment of Model Fit using Incomplete Data Tamas Rudas, Department of Statistics, Faculty of Social Sciences, Eotvos Lorand University Budapest, Hungary; and TARKI Social Research Centre, Budapest, Hungary, [email protected] The talk presents a general framework for the analysis of survey data with the fact that some observations are missing taken into account. The approach presented aims at taking the unobserved part of the data into account when assessing model fit. The true distribution is modeled as a mixture of an observable and of an unobservable component. To assess model fit in this context, the mixture index of fit is used. The mixture index of fit does not postulate that the model of interest may account for the entire population, rather it considers the true distribution as a mixture of a component where the model fits and of another one where the model does not fit. The fit of the model with missing data taken into account is assessed by equating these two mixtures, one describing the observational process and the other one representing model fit, and asking, for different rates of missing observations, what is the largest fraction of the population where the model may hold true. In this framework a diagnostic procedure will be discussed and applied to survey data. Keywords: missing data, mixture index of fit model fit SESSION: NEW METHODS METHODOLOGY Organizor: Nathaniel Beck Contact: Department of [email protected] Politics, NYU, IN 726 POLITICAL Broadway, 7th (AND Fl., NY SOCIAL) 10003, USA, Keywords: Bayesian analysis, spatial models, transition models, measurement Abstract This sessions has a variety of papers which explore new methods used by political scientists and related social scientists. Some papers consider a variety of methods for data sets that have not previously been seen as amenable to analysis; others discuss state of the art methods with great potential. Time-Series--Cross-Section Data Nathaniel Beck1 and Jonathan Katz2 1 New York University, USA 2 Caltech, USA The analysis of time-series--cross-section (TSCS) data continues to burgeon. Discussions of the methodology to analyze such data are also burgeoning, with at least 4 separate papers appearing this 300 year (in addition to several related time series papers). We would like to take this opportunity to consider various methodological suggestions and update our own views on TSCS analysis. Amongst the issues we consider or reconsider are: a. Differences between TSCS and "panel" data b. Fixed effects c. Fixed effects with a lagged dependent variable d. Modeling dynamics (the issue of lagged dependent variables) d'. The consequences of minor misspecifications in modeling dynamics e. Improved modeling of spatial effects f. Modeling heterogeneity The approach will consist of analytics, Monte Carlos and comarisons in "real" data. References Nathaniel Beck (NYU) and Jonathan Katz (Caltech) Time-Series--Cross-Section Data: 2004 New Directions in Spatial Analysis: Space is more than Geography Nathaniel Beck1 and Kristian Gleditsch2 1 New York University, USA 2 UCSD, USA, and PRIO, Oslo, Norway Most applications of spatial statistics and spatial econometrics have considered only dependence over geographic space. In the social sciences, however, the relevant metric may not always be geographic space. In this paper we extend techniques for dealing with spatial dependence to alternative metrics. We consider two applications from international relations and cross-national research. Many applications examine dyadic data, where each observation reflects the interaction of two units A and B. In such studies, two dyads composed of the same state are unlikely to be dependent on one another. We also consider estimating a model where we examine the contribution of different metrics to dependence among observations. More specifically, we estimate a spatial autoregressive model of democratic institutions, treating a country’s level of democracy as conditional on other states that are geographically close, connected through economic linkages, and cultural links. American Public Opinion in the 1930s and 1940s: the Analysis of Quota-Controlled Sample Survey Data Adam Berinsky, Department of Political Science, Massachusetts, Institute of Technology, USA The late 1930s saw the birth the survey research enterprise. Given the importance of public opinion during these critical years in American politics, it is perhaps surprising opinion poll data from this time have largely been ignored by political scientists. One reason why these data may remain unexamined is that early polls were conducted using methods now viewed as shoddy. Modern opinion polls are conducted using probability sampling methods. However, opinion polls in the U.S. before the 1950s were conducted using less expensive quotacontrolled sampling methods. The particulars of this method varied across survey organizations, but all polling firms sought to interview certain predetermined proportions of people from particular segments of the population. The goal of this method was presumably to obtain a descriptively representative group of citizens. But, in practice, these techniques failed to gather representative samples for two reasons. Not all pollsters were interested in obtaining descriptively representative samples. Gallup was most interested in predicting elections, so he drew samples to represent each population segment in proportion to votes they usually cast in elections, rather than in proportion to number in the population. Thus Gallup' s "representative" sample was intended to represent the voting public, not the full population. Because 301 Southerners, Blacks, and women turned out at low rates in the 1930s and 1940s, these groups were deliberately underrepresented. For instance, Gallup designed his samples to be 65 to 70 percent male through almost the entire war. As troubling as these deliberately induced sample imbalances may be, the practice of quota sampling also introduced a number of unintended biases. Apart from having to fulfill certain demographic quotas, interviewers were given great discretion to select particular citizens to poll. Since interviewers preferred to work in safer areas and tended to survey approachable respondents, the "public" they interviewed often differed markedly from the public writ large in a number of important ways. For example, comparisons between census data and AIPO data show that Gallup's respondents were better educated than the mass public at the time. The 1940 census indicated that about 10 percent of the population had at least some college education. But almost 30 percent of a 1940 Gallup poll sample had a college education. Political scientists who are aware of the limitations with polls conducted before 1950 have arrived at a simple solution: They reject the polls out of hand. But rather than discard this valuable data, I have developed statistical methods to account for the non-random manner in which the respondents were selected. Specifically, I have developed specific analytic strategies to correct for the biases of the data at both the individual and aggregate levels. At the individual level, to correct for the sample selection and interviewer-induced biases introduced by the quota-control sampling method, researchers should introduce statistical controls for: (1) the variables that made up the quota -- such as gender, region, and occupation (2) the education level of the respondent, when available, and (3) a series of fixed effects for the interviewer number, where available. At the aggregate level to correct for the sample selection and interviewer-induced biases researchers should use post-stratification weights to bring the distribution of the variables in the sample in line with those in the census Estimating Spatial Autoregressive Coefficients in Spatial Lag Models Robert J. Franzese Jr. and Jude C. Hays, University of Michigan, USA We explore some of the methodological problems that complicate estimation of spatial autoregressive coefficients in standard spatial-lag models. For analysts directly interested in the spatial relationships in their data, accurate estimation of this coefficient is essential, but at least two kinds of problems arise in this context. One stems from the endogeneity of the spatial lag in the standard model; the other is due to imperfect measurement of the spatial lag. One way to estimate the autoregressive coefficient is to include the spatial lag in the model despite its endogeneity, a procedure sometimes called the Generalized Population Potentials (GPP) estimator (Land and Deane 1992). A couple of instrumentalvariable IV estimators have been proposed to address the endogeneity problem, but the relative performance of these estimators, especially in non-infinite samples, is unknown (Anselin 1988). We provide a set of experimental results that compares the GPP and IV estimators under a wide range of conditions. Another kind of problem arises because the true spatial lag is not directly observed. Prior to estimation, analysts make choices about the weights in the "spatial-diffusion" matrix, which choices, being imperfect, leave measurement error in the spatial lag. Oftentimes analysts choose arbitrary weighting structures that may reduce the variance in the measured spatial lag relative to the true spatial lag-for example, when the average value of the dependent variable in the other n-1 units is used as a proxy for the true spatial lag (Franzese 2002). We explore the possibility that this induces a "reverseattenuation" bias in estimates of spatial autoregressive coefficients. Recently, scholars have begun to include multiple spatial lags in their models as a way to test competing hypothesis about the sources of spatial interdependence (e.g., Simmons and Elkins 2004). These hypotheses generate different weights in the spatial-diffusion matrix and therefore different (imperfectly measured) spatial lags. We argue these spatial lags contain measurement errors that-given missing data and boundary problems-are likely to correlate. In other words, when spatially important units are excluded from the sample, some measurement error enters in the same direction across the different lags. We examine the extent to which such correlated measurement error in multiple spatial lags complicates hypothesis testing (Achen 1985). 302 Biases in the Effects of Parental Background on Voting Jannes de Vries1, Nan Dirk de Graaf1, Rob Eisinga2 1 Department of Sociology, University of Nijmegen, The Netherlands, [email protected] 2 Department of Methodology, University of Nijmegen, The Netherlands Parental occupational status, denomination and voting behavior affect the voting behavior of respondents, even while holding constant respondents' own occupational status and denomination. Information on parental occupation, denomination and voting behavior is usually supplied by their adult sons and daughters, who are the respondents in survey research. This information is likely to contain measurement error since the information adult children supply is about a situation in the (sometimes far) past, and since this information is not about themselves (proxy information). For example, the influence of parental voting behavior could be overestimated because respondents bias answers about their parental voting behavior towards their own voting behavior. This paper investigates whether the conventional retrospective measurement of family background leads to biases in the estimates of effects of family background on voting behavior. The data we analyze stem from three cross-sectional surveys from the Netherlands (Family Surveys Dutch Population 1992, 1998, and 2000). In these surveys three informants (primary respondents, one of their parents, and one of their siblings) were interviewed, and all three of them supplied information about parental occupation, denomination and voting behavior. Three structural equation models are estimated to assess whether conventional research yields reliable estimates of family background. The first model uses only retrospective information given by respondents. The second model includes the information supplied by the parents themselves and by one sibling. Parental occupational status, denomination and voting behavior are now latent variables, each having three indicators: each informant forms one indicator. This is a model with random measurement error. The third model is estimated using the three informants again, but now errors in the answers that respondents give, are allowed to correlate with characteristics of the respondents. The error terms in respondents' information on parental occupational status, parental denomination and parental voting behavior are correlated with respondents' own occupational status, denomination, and voting behavior. This is the model with systematic measurement error, and we argue that this model optimally yields the effects of parental background on voting behavior. Event Data Based Network Analysis (EDNA) Thomas Widmer1 and Vera Troeger2 1 Department of Political Science, University of Zurich, Switzerland, [email protected] 2 Swisspeace, Berne, Switzerland, [email protected] This paper explores the independently developed contributions offered by two streams of research: event data analysis and social network analysis. The first approach, event data analysis, deals with data collecting enterprises in the field of international relations used to circumscribe and explain interstate interactions. This approach was first introduced in the seventies and saw a second wave of strong advances since the mid nineties, especially based on new artificial intelligence supported data collecting procedures. The second approach, embedded in the theoretical developments in the field of domestic and comparative politics, is the so-called social network analysis approach. The last two decades have witnessed various innovations in the methods used to analyze data gathered through standardized questionnaires or based on written evidence in a comparative setting. In this context, one of the most recent innovations is the analysis of dynamic networks. This contribution demonstrates that the combination of these two approaches within the event data based network analysis (EDNA) framework has three central advantages. On a theoretical level, our approach to EDNA allows a more stringent, reasonable application of the theoretical hypothesis at hand in various fields of political science. In a methodological perspective we show, that EDNA has decent advantages - especially its explanatory power - compared to the approaches so far used in both schools. On a practical level, we show how EDNA can be used to support a sound early warning system that allows for efficient analysis that is better suited to understanding the complexity of the world system. Network analysis so 303 far either analyses statically cross-sections of valued graphs, as the quadratic assignment procedure for example does or dynamic binary networks by p* or markov chain models that base on logistic models. The challenge is to apply dynamic models to valued graphs where the dyads are either formed by counts, e.g. the frequency of interactions or scaled values, e.g. the mean degree of conflict of interactions. To analyze such valued networks dynamically one has to account for different underlying distributions, e.g. negative binomial or poisson distributions for event counts or normal distributions for scaled data. Another problem is not only to consider the serial dependence of successive networks but also account for contemporary dependence. That is: the interaction within a dyad not only depends on past values, however, it as well depends on the interactions of the dyadic actors with other actors during the same period. Our paper accounts for all mentioned problems at a time and proposes an integrated dynamic network model for the degree of conflict within an actors network. Matching Methods for the Causal Analysis of Observational Data Markus Gangl (corresponding author) and Thomas A. DiPrete, Social Science Research Center Berlin (WZB), Reichpietschufer 50, 10785 Berlin, Germany, Tel. (030) 25491-141; Fax (030) 25491-222, [email protected] Having close linkages with the counterfactual concept of causality, nonparametric matching estimators have recently gained in popularity in the statistical and econometric literature on causal analysis. In this paper, we first alert social scientists to key features of the Rubin causal mode (RCM, also known as the counterfactual concept of causality) that has become the standard tool for statisticians and econometricians to formalize the problem of causal inference in (social science) applications based on observational rather than experimental data. We then discuss the empirical implementation of a counterfactual analysis by propensity score matching methods that provide a stringent nonparametric framework for thinking about and drawing causal inferences in the social sciences. We describe the assumptions underlying matching methods relative to standard regression analysis. We emphasize the utility of the counterfactual framework for both micro and macrolevel empirical work and illustrate the application of matching estimators in an analysis of the causal effect of unemployment on workers' subsequent careers. SESSION: RECENT DEVELOPMENTS IN NONPARAMETRIC ITEM RESPONSE THEORY Organizers: Klaas Sijtsma and L. Andries van der Ark Contact: Tilburg University, The Netherlands, [email protected], [email protected] Keywords: automated item selection procedures, Mokken scale analysis, non-parametric item response theory, ordinal measurement, test and questionnaire data, Abstract Test or questionaires are often designed to measure the respondents' score on a hypothetical construct. Examples are aggression, neuroticism, political attitude, religiosity, arithmetic ability and verbal intelligence. Nonparametric item response theory models are based on sets of assumptions that pose few restrictions on the data, but still are restrictive enough to obtain ordinal measurement of persons and items on the basis of the total score on a test or questionnaire. The essential part of nonparametric item response theory is the ordinal relationship between the expected scores on the items in the test and the latent trait that represents the hypothetical construct. In this session new developments in nonparametric item response theory are presented. Examples of appropriate topics for this session are: Mokken scale analysis, procedures for selecting items into distinct subsets (examples of such procedures are MSP and 304 DETECT), estimation procedures, invariant item ordering, differential item functioning, person-fit analysis, methods for missing data handling in test and questionnaire data using few assumptions, and applications of nonparametric item response theory to the construction of new tests and questionnaires or the secundary analysis of available data. Detecting Discontinuity On a Cognitive Developmental Ability: A Comparison of the Binomial Mixture Model and the Latent Class Model. Samantha Bouwmeester, Department of Methodology and Statistics, Tilburg University, The Netherlands, [email protected] The nature of cognitive change is an important topic in cognitive developmental theory. In particular the question whether cognitive development is characterized by broad developmental stages (Piaget’s Theory) or by continuous change (Information Processing Theory) was important for decades (Flavell, 1970; Brainerd, 1978). Nowadays the emphasis has changed from broad developmental stages into the discontinuity or continuity in the development of specific abilities. For example, Thomas (1989) studied discontinuity in classification performance; Formann (2003) studied discontinuity in the waterlevel-task; Hosenfield, Van der Maas, and van den Boom (1997) studied discontinuity in analogical reasoning; and Van der Maas (1998), and Jansen and Van der Maas (2002) studied discontinuity in balance scale tasks. An advantage of this move is that discontinuity / continuity can be investigated empirically and modeled statistically. In particular the binomial mixture model (BMM) is used to study discontinuity. When using the BMM the peaks and gaps of a frequency distribution of a numbercorrect score on a particular cognitive ability are described by a number of binomial distributions. A disadvantage of this BMM, is that the existence of multiple strategies might not be observable by peaks or gaps in the frequency distribution and therefore the binomial distributions might not be the appropriate distributions to detect different strategies. Moreover, by analyzing the frequency distribution of number-correct scores it is assumed that a particular number-correct score describes maximal one strategy that seems only realistic when a Guttman scale or Rasch model fits the items of the test. Finally, the BMM assumes an equal difficulty level for all items in the test, which makes the model restrictive. An alternative model is the latent class model (LCM). In fact the BMM is a restricted LCM. In a latent class model, the item responses are analyzed instead of the number-correct scores. Therefore, number-correct scores are allowed to describe different kind of strategies. Moreover, no binomial distributions are assumed and the difficulty level of the items can differ. In our study we compared the BMM and the LCM on a number of aspects using empirical cross-sectional data collected by a test measuring transitive reasoning. Because the two kinds of models are nested, their fit could be compared directly. It was concluded that the LCM is a more appropriate and realistic model to determine the existence of multiple stages than the BMM. Keywords: binomial mixture model, cognitive modelling, discontinuity, developmental change, latent class model, transitive reasoning References Brainerd, C. J. (1978). The stage question in cognitive-developmental theory. The behavioral and brain sciences, 2, 173–213. Flavell, J. H. (1970). Stage-related properties of cognitive development. Cognitive Psychology, 2, 421–453. Formann, A. K. (2003). Modeling data from water-level tasks: a test theoretical analysis. Perceptual motor skills, 96, 1153–1172. Hosenfield, B., Van der Maas, H. L. J., & van den Boom, D. C. (1997). Detecting bimodality in the analogical reasoning performance of elementary schoolchildren. International Journal of Behavioral Development, 20, 529–547. Jansen, B. R. J., & Van der Maas, H. L. J. (2002). The development of children’s rule use on the balance scale task. Journal of Experimental Child Psychology, 81, 383–416. 305 Thomas, H. (1989). A binomial mixture model for classification performance: A commentary on Waxman, Chambers, Yntema and Gelman (1989). Journal of Experimental Psychology, 48, 423–430. Van der Maas, H. L. J. (1998). The dynamical and statistical properties of cognitive strategies: Relations between strategies, attractors, and latent classes. In K. M. Newell & P. C. M. Molenaar (Eds.), Applications of nonlinear dynamics to developmental process modeling (pp. 161–176). Hillsdale, NJ: Lawrence Erlbaum Associates. A Markov Chain Approach for Ranking Individuals from Item Responses Matthew S. Johnson, Baruch College, CUNY New York, USA, [email protected] The raw score has long been used to rank individuals from their item response data. When the design is not balanced, for example, if different respondents receive different sets of items then the use of the raw score may not be justified. This paper examines the efficacy of a ranking system developed as an extension of one suggested by Callaghan, Porter, and Mucha (2003) for paired comparisions data. The ranking system posits a latent voter whose job is to select the “best” individual in the sample. The voter moves from one individual to another according to a Markov chain. The transition probabilities of the Markov chain are formed by assuming that the latent voter changes his opinion of who is “best” according to the following rules: First the voter randomly selects another individual in the sample to compare with the current “best” examinee. If the two individuals have no items in common, then the voter chooses to move to the new selected individual with probability one-half. If the two individuals do have items in common, then the voter randomly selects one of the common items to compare the two individuals. If the two individuals receive the same score on the item then the voter randomly chooses one of the individuals with probability one-half. If the responses are not identical, then the latent voter moves chooses the individual with the higher score on that item with probability 0.5+∂, where ∂, some number between 0 and 0.5, is a sort of discrimination parameter that may vary by item. The individuals are ranked according to the limiting distribution of the Markov chain. This limiting distribution represents the likelihood that a given individual is the “best” in the sample. In addition to describing the ranking system, the paper compares the method with suggested alternatives from the literature, including raw scores, normits, and parametric methods. The paper shows that the ranking system is equivallent to ranking by the raw scores when the design is ballanced and discusses properties of the limiting distribution of the Markov chain. The ranking system is demonstrated utilizing data from both attitudinal surveys and educationg testing. Keywords: Markov chains; nonparametric IRT; scaling; attitude measurement A Bayesian Framework for the Monotone Homogeneity Model of Non-Parametric Item Response Theory George Karabatsos, Department of Educational Psychology, Program of Measurement, Evaluation, Assessment, and Statistics (MESA), College of Education, University of Illinois, Chicago, USA, [email protected] This presentation introduces a general Bayes inference framework useful for analyzing dichotomousscored item responses, under the assumptions of the Monotone Homogeneity (MH) model of nonparametric item response theory, which nests the unidimensional one-, two-, three-, and four parameter logistic models as special cases. Given a matrix of binomial parameters, where the rows are the different rest-score groups, and the columns are the test items, the monotone homogeneity model is wholly represented by an order-constraining prior distribution on these parameters, which assigns zero 306 probability to item response functions that are non-monotonic over the latent trait (i.e., over the restscore). A Gibbs sampler is presented, that allows for fast estimation of the posterior distribution of the binomial parameters under the monotonic constraints implied by the MH model. Furthermore, any set of test data can be statistically compared with the posterior predictive distribution of the model, in order to rigorously test the global, item, and person fit to the MH model, as well as to test the assumption of local independence. Moreover, the posterior predictive distribution can be easily extended to test invariant item ordering, and for testing item bias. A demonstration of the Bayes inference framework is provided through the analysis of real test data from the National Assessments of Educational Progress. Extensions of the framework are discussed, which include the introduction of the Bayesian-hierarchical version of the MH model, which nests the ordinary MH model as a special case. Multiple Imputation of Item Scores in Test and Questionnaire Data, and the Influence on Psychometric Results Joost van Ginkel, Department of Methodology and Statistics, Tilburg University, The Netherlands, [email protected] Missing data can be a serious problem in psychological research. Listwise deletion is the most commonly used method for dealing with missing data, although more sophisticated methods are available, based on multiple imputation. In general these methods work well, but are often too complicated to understand for social scientists that have no statistical background. In the context of test and questionnaire data several easier imputation methods have been developed as well. The aim of this study was to compare the performance of the easier imputation methods to a more complicated method under several conditions. In a simulation study we compared the results of Mokken scale analysis and reliability analysis for complete test data with the results for imputed test data. Seven multiple imputation methods were studied: random imputation, two-way, two-way with normally distributed errors, response function with completely observed cases, response function with incompletely observed cases, corrected item mean substitution with normally distributed errors and multivariate normal imputation. The simple methods two-way with normally distributed errors and corrected item mean substitution with normally distributed errors performed better than the more complicated method multivariate normal imputation. Furthermore, the bias of the results due to imputation was generally very small. Keywords: multiple imputation, missing data, multivariate normal imputation, two-way imputation, response function imputation, Mokken scale analysis, item nonresponse A Nonparametric IRT Model for the Circumplex: Generalization to Polytomous Data Wijbrandt H. van Schuur, [email protected] Department of Sociology, University of Groningen, The circumplex model is an Item Response Theory model for the measurement of subjects and items on a circular latent trait. Examples in psychology can be found in the study of personality (the Big Five) and emotions, and in sociology in the study of human values and vocational interests. The circumplex model can be regarded as an additional tool within Facet Theory. In the circumplex model subjects and items are located on the circumference of a circle. In the model for dichotomous data, each item is represented with two item step parameters, indicating the beginning and end of the arc along the circle that represents its latitude of acceptance. Subjects are represented with one location parameter. Subjects who are represented within the latitude of acceptance of an item will respond positively to the item in the deterministic model, or have a high probability of responding positively to the item in the probabilistic model. A coefficient of homogeneity is used, analogous to Loevinger's 307 coefficient of homogeneity in the Mokken scale, that is based on all quadruples of items. The response patterns 1010 and 01010 in an ordered quadruple violate the model. In the model for polytomous data, the highest item category is represented as a single arc, whereas lower item categories are represented as two arcs, located to the left and right of the arc for the highest category. Response patterns that are not either monotonous or single peaked violate the model. A procedure for calculation scale values for the polytomous model is proposed, as well as the generalization of the coefficient of homogeneity, and several other model diagnostics. A substantive application is given. References Mokken, R.J., Van Schuur, W.H., & Leeferink, A.J. (2001). The circles of our minds: a nonparametric IRT model for the circumplex, in: Boomsma, A, Van Duijn, M.A.J. & Snijders, T.A.B. (Eds.), Essays in item response theory. Lecture notes in statistics, Vol. 157, (pp. 339-356). New York: Springer Tracey, T.J.G., & Rounds, J.B. (1997). Circular structure of vocational interests. In: R. Plutchik & H.R. Conte (Eds.) Circumplex models of personality and emotions. Washington D.C.: American Psychological Association. On Ordering Properties of Classical Optimal Scaling Matthijs J. Warrens, Department [email protected] of Psychology, Leiden University, The Netherlands, Guttman (1941) proposed a generalization of principal component analysis to multivariate categorical data. The method is known under many names but will be referred to, following McDonald (1983), as classical optimal scaling (abbreviated as COS). The method has some interesting properties when applied to categorical scores with an assumed latent dominance structure. Such data are usually analyzed by unidimensional cumulative item response models, either parametric or nonparametric (Mokken, 1971). The work of Schriever (1985) on the dichotomous case is extended to the polytomous case. Conditions are specified under which the category scores of the unidimensional COS are ordered. The results are demonstrated on generated data, conforming to the polytomous perfect scale and the graded response model. Keywords: Classical optimal scaling, ordered category scores References Guttman, L. (1941). The quantification of a class of attributes: a theory and method of scale construction. In The Committee on Social Adjustment (Ed.). The prediction of personal adjustment (pp. 319 – 348). New York: Social Science Research Council,. McDonald, R.P. (1983). Alternative weights and invariant parameters in optimal scaling. Psychometrika, 48, 377 – 391. Mokken, R.J. (1971). A theory and procedure of scale analysis. The Hague/Berlin: De Gruyter/Mouton. Schriever, B. F. (1985). Order dependence. Dissertation, CWI-Tact 20. Amsterdam: Center for Mathematics and Computer Science. 308 SESSION: GRAPHICAL MODELLING Organizor: Nanny Wermuth Contact: Chalmers/ Gothenborg University, Mathematical Statistics, 41296 Gothenborg, Sweden, [email protected] and Johannes Gutenberg Universitaet Mainz, Germany, [email protected] Some Aspects of the Analysis of Graphical Markov Models D.R. Cox, Nuffield College, Oxford, United Kingdom, [email protected] Graphical representations play a number of constructive roles in the analysis and interpretation of relatively complex systems. The establishments and checking of conditional independencies establishes and confirms the structure of the graph. The role of triangular representations of such systems is discussed and some other developments of statistical analysis are reviewed. An empirical example is given. Computational Aspects of Chain Graph Models with Mixed Response Variables Vanessa Didelez, University College London, UK, [email protected] Chain graph models are used to represent the dependence structure in multivariate data, where some variables are regarded as reponses, some as explanatory and some variables take the role of intermediate variables. They can help investigating the data generating process and may reveal interesting direct and indirect relations. However, when the multivariate responses are of mixed scale, these models are difficult to fit. The task can be simplified by exploiting collapsibility allowing the decomposition of the multivariate model into lower dimensional or even univariate regressions. In this talk, I will motivate the use of chain graph models with a study on self-care in later life. Simple graphical conditions for collapsibility will be given. Implications for model selection as well as for latent variable models will be discussed. Checking for a Confounder in Models Generated over Directed Acyclic Graphs Elena Stanghellini1 and Giovanni M. Marchetti2 1 University of Perugia, Italy, [email protected] 2 University of Florence, Italy, [email protected] We discuss a method to assess whether, in a non-parametric context, a marginal association between two observed variables is confounded by associations to a common third unobserved variable. We partition the random vector Y into two subsets, the observed variables and one single unobserved variable. We assume that the data generating process can be described by a parent graph, i.e. by a directed acyclic graph complemented by a full ordering of the variables. We assume that the ordering is known from a priori information. The object of our inference involves the generating edge matrix, a matrix form of the parent graph. This is an upper triangular binary matrix having ones along the main diagonal and an ij-element, for i<j, is 1 if and only if node j is a parent of node i. Given the parent graph, the structure of associations implied for the observed variables can be derived by combining directly probability statements (see e.g. Dawid, 1979), by using the a separation criterion for directed acyclic graphs (Pearl, 1988) or by transforming the generating edge matrix (Cox and Wermuth, 2004). But knowledge of the structure of associations among the observed variables only, does not in general permit to uniquely reconstruct the edge matrix of the parent graph. We explore situations in which this reconstruction is possible and discuss the underlying assumptions. 309 An Application of Marginal Log-linear Models to Examine Changes in Social Mobility in Hungary During the Transition Period Renáta Németh, Eötvös Loránd University, Budapest Hungary,[email protected] The talk will analyze social mobility data for Hungary from the years 1987, 1992 and 1999. The main concern is put on testing Donald J. Treiman’s industrialization hypothesis that was posed in 1970 and is still widely cited today in the context of transition. The objective variables involved in Treiman’s hypothesis are completed with subjective social position and subjective social mobility, aiming to take into account perception as a pathway through which objective factors have an impact on individual behavior. The fitted models are graphical models based on directed acyclic graphs and the values of marginal log-linear parameters (as proposed in Rudas, Bergsma, 2004) are used to gain insight into the strengths of associations among the variables and the details of the computations will be illustrated. From a substantive point of view, the main findings include that mobility trend between 1987 and 1992 had an opposite direction to the trend between 1992 and 1999, showing a quick and sensitive reaction to the economic changes. That is, when investigating transition process we should distinguish these two periods. Marginal Models for Multivariate Dependencies Tamas Rudas1 and Wicher Bergsma2 1 Tarki Budapest, Hungary, [email protected] 2 Eurandom, Eindhoven, Netherlands, [email protected] Cox and Wermuth (1996, 2004) discuss a great variety of graphical models to analyze multivariate dependencies. These include models based on directed acyclic graphs, and on partially directed graphs, graphs with solid and dashed edges (both arrows and lines). In general, models with arrows correspond to complex regression-type models with possibly several response variables. Bergsma and Rudas (2002) developed a theory of marginal models for categorical data, based on marginal log-linear parameterizations. These parameterizations generalize the multivariate logistic transformation of Glonek and McCullagh (1995) and the standard log-linear parameterization. Statistical models based on directed acyclic graphs are marginal models (see Rudas and Bergsma, 2004). The talk will investigate some of the more general regression-type models based on partially ordered graphs with solid and/or dashed lines from the point of view of their natural parameterizations. In these parameterizations the conditional independencies implied by the graph are represented by zeroes among the marginal log-linear parameters and the remaining, non-restricted parameters can be interpreted as measuring the strength of certain, possibly quite complex, effects allowed by the model. The application of this theory to the continuous case will also be considered. The Use of Bayesian Networks for Imputation Paola Vicard1, Marco Di Zio2 and Mauro Scanu2 1 Università Roma Tre, Italy, [email protected] 2 ISTAT, [email protected], [email protected] Dealing with data sets affected by partial non-responses is a, pervasive problem in statistical surveys. A common approach to, deal with partial non-response is to replace (i.e. impute) missing, items with artificial plausible values, under the assumption that, the missing data mechanism is Missing At Random (MAR, see Rubin, 1976). The main reason why the approach based on imputation is so widely used is that it produces a complete data set that can be easily analysed with the usual statistical techniques. A wide variety of imputation techniques has been developed (e.g. see Little and Rubin, 2002). 310 While most of them have desirable properties when studying univariate characteristics, i.e. tend to reproduce the univariate distributions, the preservation of relationships among variables is still a problem to be furtherly studied. Bayesian Networks (Cowell et al., 1999) are a well-known tool to study the relationships among variables in complex surveys. Consequently they can be usefully applied for imputation. In this work we present an overview of some approaches based on Bayesian networks to impute missing items. Some comparisons with standard imputation techniques will be provided. SESSION: THE NETS BETWEEN NEURAL NETWORKS AND NETWORK ANALYSIS. FUZZY LOGIC, COMPLEXITY AND/ OR NETWORK ANALYSIS SOCIAL THEORY Organizers: M. Ampola and Dott. A. Givigliano Contact: University of Pisa, Italy. [email protected], [email protected] Keywords: Network Analysis, Fuzzy Logic, Neural Networks, Languages, Complexity Social Theory Abstract The study and the analysis of the social reality as a relational network could bring to a double point of view. The first one concern a reading of the social world as a complex multidimensional network of relationships, made of one dimension for each relation. The dimensions aren't strictly distinguished, but they could be seen as a complex unity (following the idea of complexity expressed by Edgar Morin). The second point of view concern the classical and recent network analysis approaches, which identify the relations as a set of interactions among differents points in the social space. Is it possible to unify this two approaches? The complex approach and the network one have of course a logic behind them: which is this logic? And is it one logic or different kinds of logic? The aim of this session is to analyze this two points of view and the possible convergence and differences between them today. Is it a problem of data-analysis or there are other kinds of reasons? Social Network Analysis, Social Theory and Research Design: Experiences from an Italian Contexts Andrea Salvini, Department of Social Sciences, University of Pisa, Italy, [email protected] Over the last few years Social Network Analysis has experienced a further process of advance at the technical-methodological acquisitions level (just think for example to the use of sophisticated statistical tools applied to networks dynamics analysis), but also at the level of the spread in not only Anglo-Saxons scientific contexts (f.e. in France and Italy). However, the increase of awareness about methodological soundness of this approach hasn't still found - in Italy at least - a correspondent richness of surveys and studies in which such awareness finds application, save due exceptions. In literature many technical contributions are found, but there are insufficient references to empirical experiences in which that approach turned out to be actually useful, f. e. in the field of social politics. The present work has a twofold purpose: on one hand, to give the picture of the problems affecting the relationship between Social Network Analysis theoretical reference and methodological approach, showing how these problems influence the construction of a consistent research design in empirical research. Concerning this, the relations between some theoretical approaches - such as Structuration Theory and Exchange Theory - and Social Network Analysis will be investigated, and the potentialities in terms of constructing integrated and consistent processes of research will be pointed out. On the other hand, the purpose is to represent some concrete experiences of social research carried out in Italy, in particular in Tuscany, about the topic of social service and volunteering, in which it has 311 been tried to link up some of the already mentioned theoretical references with Social Network Analysis practice. These research experiences have utilised the ego-centred networks approach in order to elaborate relational indicators able to interact - through the construction of suitable models with other indicators obtained through the application of "traditional" methodologies. To conclude, we intend to support that only the integration between the "new" relational approach and the others "conventional" forms of research may allow to valorize Social Network Analisys techniques, and to acquire more suitable tools and processes to the comprehension of social complexity. Keywords: Methodology, Social Network Analysis, Structuration Theory, Exchange Theory. The Complex Dispositional Space of the Social Relation and the Nets between Social Networks and Neural Networks. The Problem of the Language. Alfredo Givigliano, Department of Social Sciences, University of Pisa, Italy. [email protected] A modality of reading social reality inside an approach of research founded, structured and projected on a perspective which belongs to the family of the network analysis, concerning its interpretative function, integrated with a complex reading of the same reality, it can be built on the base of a space of the possibilities of the subject that becomes a space of the dispositions inside, by and in function of language. Such modality doesn't concern only the possibility of analysis of the single subject, but above all the description and reconstruction of the space of the possibilities, than the space of dispositions, of the social relationships in their complexity. In this case the language, the every day life language, is more clearly identifiable from a side as the object that allows to study the relationships, from the other, it is the tool inside the sociology for this study. A strategy and a project of research, from a logical and methodological point of view, and of techniques one, that inserts this reading inside a matrix of network analysis, as it regards the phase of the analysis of the data, it can be constructed on the base and in function of the use of specific neural networks. The nets in matter, would work not only as a tool of comparison, but also of determination of the same topology of the space of the dispositions. According to this idea the social relationships don't individualize something static but something dynamic that exploiting, in phase of description, the vague structure of the relationships themselves, allows to build a continuum of the same vagueness, therefore, not to assume this characteristic in definitive and substantive way, but to report it to the specific dynamics and situations of the phenomenon that the researcher intends to study and to analyze. A particular attention, inside this perspective, it must be sets to the study and the analysis of the language, of the natural languages and of the sociology own, which structure the relationships themselves, as well as the social relationship that results to be the one of the researcher towards his own object of study. The hypothesis to be analyzed and to develop is the tension among the two languages that it resolves inside the same natural language, that determines that sociology one according to logical and epistemological modality. Keywords: Space of possibilities, Space of dispositions, Complexity, Vagueness, Social Network Analysis, Neural Networks. The Aim of a Neural Social Researcher Simone Gabbriellini, Laboratory [email protected] of Social Research, University of Pisa, Italy, The analysis of the relational data seems to be so more difficult according to the researcher’s range of action. A methodology, founded upon the use of neural networks, allows to leave on background, in the analysis phase, the theoretical formalization of the distribution of the relationships that forms the social data in examination, according to the statistic tradition, leaving this involvement bypassed by the software. 312 With the opportune construction of indexes, it is possible to analyze the matrix of weights produced by the net, nevertheless it doesn't give us back any function able to approximate the course of the analyzed phenomenon, cause we didn’t give the function first, at the beginning of the process! A neural network is an object able to reproduce, with a minimum amount of error, the intensity of the relationships that tie the values of the variable that we build from the segment of reality that we want to examine. This ability cause the net to be comparable to a liquid which we can dose the stringiness: too diluted, it is not able to generalize from the training set to real set of analysis, becoming too tied to the peculiarities of the first one (a problem called overfitting); on the contrary, if we make the fluid too much viscous, preventing it to slip in every ravine of our object, it will become able to make generalizations, but not able to gather the peculiarities of our object. The aim of the neural social researcher is, in this case, not the choice of a theoretical distribution to analyze data, but the choice of the correct values of the parameters of the net that manage its abilities of analysis. It is in a some way a methodology that works case by case: it can produce scenaries and simulations, but doesn't make generalizations in the traditional statistic sense of the term. Is this a consequence to which we have to undergo, a sort of implicit renouncement, or is this a precise operational choice, in consequence of assuming social reality as something not forceable by generalizations that can result, at the end, only misleading, only partial representations able just to produce uncorrect images of the relational world? Is the network approach always correct? Which are the limits and the risks in the use of a neural approach tout-court? Is there any problem for which would be more correct a less complex approach? Keywords: neural, network, social, research, relational, data. Social Capital: Theoretical Considerations Stefania Milella, Laboratory of social research, University of Pisa, Italy, [email protected] The concept of social capital has met with success in sociological analysis, which uses it in order to study a lot of social phenomena. It's a complex and multidimensional concept, marked out by the interaction between a subjective dimension and a structural dimension. This paper proposes to identify some significant dimensions of the concept of social capital, making an overview of the most important theoretical contributions about the topic. In particular, we refer to relational approach about social capital (Bourdieu, Coleman), that supports that social capital consists of the resources deriving from social relations tissue in which a person is embedded. Bourdieu uses the concept of social capital in order to explain the processes of reproduction of social classes, as third kind of capital coming up by the side of human capital and economic capital. According to Bourdieu, social capital consists of opportunity of access to material and symbolic resources that the individual uses in his/her relational life. The first source of social capital is family, subsequently the actor increases and continually modifies his/her endowment of social capital through the access to different relational ambits (Bourdieu, P., 1980). In Coleman's relational vision, social capital is constituted by resources deriving from social relation tissue in which social actors are embedded. While human capital is an individual attribute, social capital exists in the social relations, and provides resources that can be used by the individuals in order to realise interests and goals that would not be attainable in its absence (Coleman, J. S., 1990). Within this theoretical frame, it comes out that social capital and social networks are tightly linked, but, at the same time, they demand a distinction at a conceptual level. Social capital is incorporated into social relations, but it can't be simply identified with them. While social networks may be both opportunities and constraints for action, social capital is always productive, because it refers to resources used by social actors in order to realise their interests. Social capital productivity is influenced by social networks features, but it also depends on particular resources activated in social ties. Sociological analysis deals with social capital both as individual resource and as collective resource; from this theoretical perspective it comes out that social capital is a situational concept, escaping a rigid definition, demanding of being analysed and interpreted referring to social actors and their social networks, and to the particular context in which social action is carried out. Keywords: social capital; social networks 313 SESSION: MISCELLANEOUS Organizor: Karl van Meter Contact: National Center of Scientific Research, 59 Rue Pouchet, 75017 Paris, France, tel/fax 33 (0)1 40 51 85 19, [email protected] A Strategy for the Selection of Weighting Variables Barry Schouten, Statistics Netherlands, Methods and Informatics Department, Room 241, PO Box 4000, 2270 JM Voorburg, The Netherlands, tel. (+31) 70 3374905, fax. (+31) 70 3375990, [email protected] Estimates for population statistics can be seriously biased in case response rates are low and the response to a survey is selective. Methods like poststratification or propensity score weighting are often employed in order to adjust for bias due to nonresponse. One problem that many adjustment methods have in common is the choice which of the available auxiliary variables to use. In the case of poststratification it must be decided what strata are defined. In the case of propensity score weighting adjustment cells must be formed that have comparable response probabilities in order to keep a low variance. In this paper we use the generalised regression estimator to adjust for nonresponse and we propose a selection strategy of weighting variables that simultaneously accounts for the relation with response behaviour and the relation with the important survey questions. The selection strategy is applied to the Integrated Survey on Household Living Conditions (POLS) 1998. The selection strategy aims at the minimisation of the absolute bias that is maximally possible under an assumption that is weaker than the usual Missing-at-Random (MAR) assumption. We assume that stratum means may be biased by nonresponse, but that the relative distances between the stratum means are approximately preserved. Keywords: Nonresponse bias, Nonresponse adjustment, Weighting model, Missing-at-Random, NotMissing-at-Random, Poststratification. The Measurement of Duration of Studies in Higher Education Stephan Boes, Department [email protected] of Statistics, University of Dortmund, Germany, The duration of studies is an important topic of discussion in educational policy at the step into a knowledge based economy. It is measured on international standards (OECD) by a rough transition model called Chain Method, which calculates the transitions from one year of studies to the next with data on two consecutive study years. The Chain Method is based on several assumptions, so it gives a result which only applies to these assumptions, and it is based on censored data. In Germany, about ten per cent of university students are censored. The OECD uses the duration of study to calculate cost of study over the whole study course. A method similar to the Chain Method is used by the Standing Conference of the Ministers of Educational Affairs of the German Laender (Kultusministerkonferenz, KMK). They use it to calculate forecast of student numbers in the German system of Higher Education. The methods used can be given a formal background. We will show that the methods are special cases of the demographic life table technique. Our paper will give results for the calculation of the duration of studies using different assumptions within life table surroundings and some approaches beyond censoring. We use data of the German Federal Statistical Office of several consecutive years to calculate and compare “survival probabilies” for students in their course of study. Different assumptions on the Lx, based on demographic standards and real life experiences, lead to different results. The analysis of the 314 data before the censoring point gives rise to assumptions about their development beyond this point. The outcome is a range for the true duration of studies. Results for Germany from different agencies are compared with our range of results. We will further give results about differences in study behaviour between male and female students and about the way the reducing of drop-out rates and restrictions to long-term enrolment would influence the duration of studies. It will also be shown what the different results mean to the student forecast of the KMK. Keywords: Education Statistics, Transition Model, Life Table techniques STREAM: DATA COLLECTION, SAMPLING AND NONRESPONSE SESSION: IMPROVING THE VALIDITY OF FRAUD RESEARCH USING INSTRUMENTS FROM THE SOCIAL SCIENCES Organizer: Gerty Lensvelt-Muldersand Peter van der Heijden Contact: Utrecht University, Department of Methodology and Statistics, Postbox 80140, 3508 TC Utrecht, The Netherlands, Tel: +31302535857, Fax: +31302535797, [email protected] Keywords: Fraud, research methods, white collar crime, underestimations Abstract Fraud is a penetrative phenomenon with the power to disrupt the economy and to attack the foundations of the welfare state. Obtaining accurate information on the types and extent of fraud is therefore extremely relevant, but reliable and valid information on the prevalence of fraud is hard to obtain. Official statistics on fraud tend to underestimate the size of it because of large non-response rates and low validity of the data due to social desirable answering and down right lying. In this session we want to give fraud researchers a chance to present a paper that can add to the quest for methods that enhance the validity of fraud estimates and decrease dark figures. The aim of this session is to give a broad overview of the methods that are used nowadays and to enhance our insight in different forms of fraud. The session should be interesting for a broad range of fraud researchers. For this session we especially invite researchers that study social security fraud, organisational fraud and white-collar crimes using instruments form the social sciences to present a paper and add to the discussion in this important field. Evaluating the Validity of (Vicarious) Self-Reports of Deviant Behavior Using Vignette Analyses Stefanie Eifler, University of Bielefeld, Postbox 100131, D-33501 Bielefeld, Germany, Tel. 0049(0)521-1064643, fax 0049-(0)521-1066479, [email protected] Focus of this paper are vignette analyses of delinquent and criminal behavior. Vignette analyses (or scenario techniques) use descriptions of hypothetical situations which are presented to subjects within the framework of a survey. Vignette analyses have been used to measure attitudes towards social situations as well as behavioral intentions evoked by the hypothetical situation presented. Vignette analyses are established as a research instrument in the study of social norms employing a factorial survey approach (e.g. Opp & Jasso, 1997, Rossi & Anderson, 1982). Scenario techniques have been also used to simulate everyday life behavioral decisions of actors, especially in the field of deviant and criminal behavior (cf. Eifler, 2001). However, vignette analyses measure behavioral intentions only, they are not suitable to reveal any information about behavioral decisions of actors in the presence of real opportunities. In addition, the similarity of behavioral intentions measured using scenario 315 techniques – in vitro - and behavior measured in vivo remains unknown since no studies evaluating the validity of vignette analyses have been carried out so far (cf. Finch, 1987). Against this background the study presented here consists of an empirical examination of the validity of vignette analyses in the field of criminal and deviant behavior. Data collected using scenario techniques on the one hand and using non-reactive approaches or non-invasive observations on the other hand are compared systematically. It is shown that for some forms of criminal behavior vignette analyses and noninvasive observations reveal similar results while for other forms of deviant behavior vignette analyses lead to underestimations of the behavior in question. The results are discussed with regard to their methodological implications. Keywords: validity, vignette analysis, factorial survey approach, self-report, criminal behavior, deviant behavior Statistical and Technical Methods for the Detection and Prevention on Interviewer Fraud in Data Collection Joe Eyerman, Paul Biemer, Craig Hill, Joe Murphy and Chris Ellis, RTI International, 3040 Cornwallis Road, P.O. Box 12194, Research Triangle Park, NC 27709-2194, Tel: (919) 541-7139, Fax: (919) 316-3867, [email protected] Data falsification and interviewer fraud can significantly increase the costs of survey research and introduce bias in study estimates. This threat has been magnified in recent years as cash incentives for respondents have become more common and greater in amount. This has increased the value of fraud for the individual interviewer by making the cash payments a significant part of the return on falsification. As a result, a new class of white collar criminal may be attracted to the data collection field, further threating the validity of our data. Furthermore, decreasing response rates and increasing pressures on interviewers to achieve high response rates may encourage falsification. RTI International has developed a suite of statistical and technical methods for the early detection, attribution, and remidiation of fraud in the data collection process. This paper will report on the overall fraud detection process employed by RTI and will highlight three techniques currently being used or developed for fraud detection in large, federally sponsored surveys. Included will be a discussion of Computer Assisted Recorded Interviewing (CARI) methods currently employed at RTI to detect fraud. This paper will also discuss data mining techniques currently being applied to survey metadata and response data to detect patterns of fraudulent behavior. Finally, this paper will review some of the emerging research at RTI that uses text mining and voice recognition software to evaluate fraudulent patterns in text fields and recorded interviews. We believe these techniques may be broadly applicable to fraud research in all areas of the social sciences. In addition, we believe these techniques could be improved by drawing on the advances made in other fields of fraud detection. Keywords: Fraud, CARI, Text Mining, Data Mining Research Designs for Fraud Studies: How to Chose the Optimal Research Design? Gerty Lensvelt-Mulders and Peter van der Heijden, Utrecht University, Department of Methodology and Statistics, POBox 80140, 3508 TC Utrecht, The Netherlands, Tel: +31302535857, Fax: +31302535797, [email protected] We will start this paper with a short expose of different methods that can be used to study the prevalence of fraud in society. Secondly we will discuss a scheme that we developed to guide the choice of a method dependent on the question that is asked, the level of information that is needed and the accessibility of data or respondents. As a rule estimating the extent of fraud is rendered difficult because of the so-called dark numbers problem. It is therefore very important to know which research method will result in the most valid 316 results, dependent on features of the research question. A literature study revealed that research methods could be classified into three broad clusters, i.e. surveys methods, methods for secondary data analysis and random audits. Seldom it is the case that there is just one single method that can perfectly estimate the prevalence of a specific form of fraud, or the amount of financial damage that it causes. It is the researcher’s task to choose a (combination of) method(s) that guarantee(s) an optimal match between research question and research method. For this purpose we developed a protocol that takes into account the features of the information that has to be collected, the features of different forms of fraud, and the assumptions of different research methods. Keywords: fraud, CAI, randomized response, network scale-up, capture-recapture, multiplier methods, random audits SESSION: PREVENTING, DIAGNOSING AND ANALYZING MISSING DATA Organizors: Joop. J. Hox and Edith D. de Leeuw Contact: Utrecht University, The Netherlands, [email protected] Abstract Social research often produces data sets with missing data, either because respondents refuse to participate in the research altogether, or because they do not provide data on some subset of variables. In some cases, the missingness is created by design. In most cases, the missingness is created unintertionally. The first important issue is therefore what measures can be taken in the data collection phase to reduce the expected amount of missing data. After the data are obtained, we must deal with two important tasks. First, it is important to diagnose the pattern of missingness. This can provide very practical information. For instance, we may find that most of the missing values concern only one specific variable. If this variable is not central to our analysis problem, we may delete it from our analysis, rather than keeping it, and therefore having to delete many cases. We would also like to know if the missingness forms a pattern, or if it is related to some of our observed variables. For if we discover a system in the missingness pattern, we may try to include that in our statistical analyses. The second task is to produce sound estimates of the parameters of interest, despite the incompleteness of the data. There are two major approaches to this problem. One is to make the data complete by imputing the missing values, and then do the analysis on the completed data. The other is to use a method, typically a likelihood-based procedure, that allows us to model incomplete data directly. This session aims to address all important methodological and statistical aspects of the missing data problem in social research. Mixed Missing: Mode Elemental Structures Antonio Alaminos, Mathematical Sociology. University of Alicante. Spain, [email protected] One of the main concerns is the nature of the missing values. Lets consider extremes for simplicity. If missing at random we have not to care about. But if missing shows structures that covariate with substantive variables we have to make decisions. There are, in fact, severals options to take. We are speaking about one country, one mode. But if you go cross-cultural (or more precisely, crossstatenations) and mixed modes (like WVS, ISSP, and others) many questions raise. For example, the simple one. What are we comparing? Reports and books usually go straight into variables distributions and coeficient comparisons. This is possible because the analist presume "tabula rasa" effect from data collections procedures. But this is not, frecuently, the real situation. This paper will expose the mixed missing mode imprint in international surveys. This will help to evaluate how deal with this problem. Also, to consider the real meaning of observed cross-national differences. 317 Social Distance, Interviewer’s Perception of Rapport and the Types of Invalid Response Su-hao Tu1 and Pei-shan Liao2 1 Center for Survey Research, Academia Sinica, 128 Sec. 2 Academia Road, Nankang, Taipei 11529, Taiwan, Tel.: 886-2-27884188 ext. 515, Fax: 886-2-27881740, [email protected] 2 Center for Survey Research, Academia Sinica, Taiwan Inspired by social distant theory, the extent to which the respondent volunteers to give an answer is associated with the similarity of characteristics between the respondent and the interviewer (Groves et al. 1992). While social distant theory emphasizes role-independent variables, i.e. the characteristics of the interviewer and the respondent, little attention in the previous literature has been put on rolerestricted variables such as the interviewer’s perception of his/her rapport with the respondent. This paper thus examines different types of invalid response associated with social distant and interviewer’s perception of interview rapport separately and jointly. The invalid response implying the respondent’s willingness to provide a substantive answer include “refusal’, “no opinion” and “don’t know”. In addition, we try to examine whether social-distant and/or rapport-perception effects would change for questions of a different nature. Four datasets analyzed are from Taiwan Social Change Survey (TSCS) conducted in 2000 and 2001. 2-level Hierarchical Linear Model (HLM2) is used under the recognition of the possible importance of interviewer effect and the assumption that the interviewers’ work is not randomly assigned. Keywords: social distant, interview rapport, invalid response, interviewer effect Socioeconomic Status of Survey Respondents with Missing Household Income Data Michael Wood, Department of Sociology, Hunter College, City University of New York, USA, [email protected] Survey data commonly contain relatively higher nonresponse on income and related items. Modern methods of data imputation offer solutions to item nonresponse, but such methods hinge upon the assumption that nonresponse is not related to the variable being estimated. This paper investigates sources of income item nonresponse using a U.S. national sample dataset (n=2000), conducted by face-to-face interviewers. The survey included questions about financial attitudes and knowledge as well as an assessment of respondent socioeconomic status made by the interviewers. Three main questions are investigated: 1) Is income item nonresponse related to income and socioeconomic status (especially, higher socioeconomic status)? 2) What, if any, differences in financial attitudes characterize income nonresponders from responders? 3) Does nonresponse on the income item correlate with generalized item nonresponse? The data analysis shows that: 1) Income item nonresponse is not significantly related to socioeconomic measures here, 2) Lack of knowledge about money and financial matters is related to income nonresponse, but income nonesponders are not generally different from responders on financial attitudes and beliefs, 3) a significant portion of income nonresponders reflect generalized nonresponse to survey questions. Keywords: missing data, income item nonresponse, data imputation, financial attitudes Statistial Editing and Imputation for Social Data Frank van den Eijkhof, Department of Methods and Statistics, Faculty of Social Sciences, Utrecht University, The Netherlands, [email protected] Also working for Statistics Netherlands (CBS) in The Hague, Voorburg, The Netherlands Improving the quality of datasets by cleaning raw data is mostly aimed at identifying missing values (item nonresponse) and finding univariate outliers, after which some imputation technique is performed. Official statistical bureaus do have more information; they investigate the possible and impossible combinations of values in records. The resulting constraints are called edits. An example of 318 an edit for censuses is, that a person cannot be married and be of an age lower than 16. Statistical editing consists of changing values in records such that no more univariate outliers exist and that all records are made consistent with the edits. Combining statistical editing and imputation is difficult, and Statistics Canada currently uses an approach (the Nearest-neighbour Imputation Method or NIM) based on nearest-neighbour imputation to solve this problem. For each record that is inconsistent with the edits (a failed record), a set of donors is taken from consistent records, selected on their distance to the inconsistent record. For each donor all variables are considered for which changing the value of the failed record into the analogue value of the donor positively influences the consistency of the edits. An exhaustive search is done for all possibilities, and sets of changes that render the failed record consistent with the edits are kept. A selection on these solutions is done to ensure a minimal change of the failed record, and the resulting solution is applied to the inconsistent record. We will show our investigation regarding the robustness and results of the NIM. For comparison part of the census of the United Kingdom (from 1991) is used. Keywords: Missing data, Statistical editing, Data Analysis, Imputation SESSION: NONRESPONSE Organizer(s): Claire Durand1 and John Goyder2 Contact: 1Department de Sociologie, Universite de Montreal, Canada, [email protected] 2 Department of Sociology, University of Waterloo, Canada, [email protected] Keywords: non-response, unit non-response Abstract This session has papers on any aspect of survey nonresponse. Topics would include (but are not restricted to): nonresponse on specific types of survey such as electoral surveys; response maximization (e.g. call management in telephone surveys, training and selection of personnel, incentives,...); theoretical approaches (e.g. leverage-salience exchange theory, social psychology of interaction and persuasion,...); research on the consequences of nonresponse; studies of panel attrition in longitudinal surveys; cross-cultural and cross-national studies in nonresponse; trends in non response in different settings and countries. Part I: Increase Response Rates The Influence of Advance Contacts on Response in Telephone Surveys: A Meta Analysis Edith De Leeuw, Joop Hox, Elly Korendijk and Gerty Lensvelt-Mulders, Department of Methodology and Statistics, Utrecht University, [email protected] Recently, the leading position of telephone surveys as the major mode of data collection has been challenged. Telephone surveys suffer from a growing nonresponse, partly due to the general nonresponse trend for all surveys, partly due to changes in technology specifically influencing contactibility by phone. One way to counteract the increasing nonresponse is to increase the number of contacts. In mail and face-to-face surveys advance letters have been shown to be an effective extra contact and raise response. This inspired researchers to perform experiments with advance letters in telephone surveys. Like an advance letter, leaving a message on an answering machine can be seen as a precontact and 319 opportunity to influence survey participation. Therefore, in this meta-analysis we present a quantitative summary of empirical studies on the effectiveness of two types of precontact on the response in telephone surveys. In the first part we summarize the effect of advance letters on response in telephone surveys. In the second part we look into the effectiveness of a second type of precontact: leaving a message on answering machines. Keywords: telephone interview, advance letter, prenotification, (telephone) answering machine, response, refusal Efficacy of Incentives in Increasing Response Rates Mansour Fahimi1, Roy Whitmore, James Chromy, Margaret Cahalan2 and Linda Zimbler3 1 Research Triangle Institute, 6110 Executive Blvd., Suite 420, Rockville, MD 20852-3903, Tel: (301) 230-4675, Fax: (301) 230-4647, [email protected] 2 RTI International, USA 3National Center for Education Statistics, USA Nonresponse has become one of the major challenges to conducting high quality survey research because the quality of survey estimates is contingent upon a high response rate. Now, more than ever before, securing a respectable rate of response is an objective that is hard to achieve in most surveys. This paper provides a summary of the results obtained from an experiment conducted to assess the effectiveness of incentives for increasing the response rate, overall and in particular for hard-to-reach individuals. Data for this research come the from the 2003 field test of the National Study of Postsecondary Faculty (NSOPF), conducted for the National Center for Education Statistics, US Department of Education. This nation-wide study involved collection of data from faculty at postsecondary institutions. The experiment consisted of a design where two levels of incentives for nonresponse follow-up were nested within three levels of incentives for early response. Keywords: Nonresponse, Incentive, and Experimental Design New Evidence on Amount of Payment on Prepaid Incentives on a Mailed Questionnaire John Goyder1, Guil Martinelli2 and Steve Brown3 1 [email protected] 2 Department of Sociology, University of Waterloo, Waterloo, Ontario, Canada 3 Department of Political Science, Wilfrid Laurier University, Waterloo, Ontario, Canada Research on cash incentives administered in prepaid mode, on mailed questionnaires, has consistently suggested that rates of response do not increase dramatically as the amount moves from a token to a substantive amount. James and Bolstein (POQ, Winter 1992 ), for example, found that a $10 payment in cheque form produced no increment over a $5 cheque. Based on two premises, the present research re-examines this question. First, inflation continually erodes the value of a $5 payment, seemingly the most plausible optimal amount for mailed questionnaires conducted in Canada. It is time to re-visit the issue of amount of payment. Secondly, our design, unlike many others, was mixed mode. After three mailings on a survey administered in Kitchener-Waterloo (Ontario, Canada), in which random halves received either (Cdn) $5 or $10 in cash, the final follow-up was by telephone. We can thus track the effect of the different amounts both over the three postal waves and as the mode of contact switched to telephone. Our finding was that $10 over $5 makes no measurable difference within the mailed portion of the fieldwork., but did quite dramatically improve the co-operativeness of sample listings once they were telephoned. The analysis also examines item nonresponse by payment condition. 320 A Plea for the Tailored Use of Respondent Incentives Gerry Nicolaas and Nina [email protected] Stratford, National Centre for Social Research, UK, There is overwhelming evidence that respondent incentives can improve survey response rates. Furthermore, meta-analyses of various experiments have shown that pre-paid incentives are more effective than promised incentives, monetary incentives are more effective than non-monetary incentives, and response rates tend to increase with the value of the incentive. It is generally assumed that an increase in response rates will improve the representativeness of the sample. However, this is not necessarily the case. First of all, incentives must improve response rates substantially to reduce non-response bias. Second, it is possible that incentives could have highly selective effects that increase the differences between respondents and non-respondents and consequently increase non-response bias. In this paper, we will make the case for tailoring the use and type of incentive to the design of the survey and to the characteristics of under-represented groups in the survey. We will use information from an incentive experiment, which was carried out to assess the effect of conditional differential incentives on response rates and non-response bias on the National Travel Survey for Great Britain. The results of this experiment showed that conditional incentives can increase full household cooperation. Furthermore, linking the incentive amount to household size was an effective method for reducing the under-representation of larger households. However, the results also showed that incentives can have selective effects on certain sub-groups. Whether these selective effects will reduce non-response bias depends on their strength and direction. Keywords: non-response, unit non-response, incentives, response rates, sample representativeness The Drop-off Pick-up Method: An Alternative Approach to Reducing Nonrespone in Mail Surveys Amy Ross-Davis and Shorna R. Broussard, Purdue University, Department of Forestry and Natural Resources, 195 Marseteller Street, West Lafayette, IN 47907, USA, [email protected] In this research, we explore the efficacy of using a personal delivery method for surveys as a means to decrease nonresponse bias. We selected two 9-mi2 sampling sites in north-central Indiana for the dropoff pick-up method and 12 9-mi2 sampling sites in the same region for the Tailored Design Method. Surveys were delivered to a census of forest landowners who owned at least 1 acre of forestland in the study area. Landowners assigned to the Tailored Design Method sites received a series of five mailings: pre-notification letter, survey with cover letter and packet of Eastern redbud seeds (Cercis canadensis) as an incentive, reminder/thank you postcard, 2nd copy of survey and cover letter, and 3rd copy of survey and cover letter via priority mail. Landowners in the drop-off pick-up sites received a hand delivered copy of the survey with a cover letter and a packet of Eastern redbud seeds (Cercis canadensis) as an incentive. Articles were published in local newspapers of each community prior to delivery to alert local residents of the study. This paper will report on the cost effectiveness of this method comapred to the Tailored Design Method, the efficay of this delivery method in terms of increasing respone rate, and will also include a detailed methodology of the Drop-Off Pick-Up techniuque as an alternative for mail surveys. Keywords: non-response; survey methodology; drop-off pick up method Recruitment Procedures and Response Rate: A Swiss Experiment Nicole Schoebi and Dominique Joye, SIDOS: Swiss Information and Data Archive Service for the Social Sciences, [email protected] In Switzerland, experiences have shown that response rates are low, sometimes even less than 50% in face to face survey. This could be the case even after a high number of contacts or attempt of refusal 321 conversion. The realisation of the Swiss part of the ESS 2002 has documented some important results in this sense. Among others, one hypothesis could explain partially the nonresponse rate in Switzerland: we have a sample of households and select, when in contact with the household, a person which will (eventually) answer to the questionnaire: This is normally done according the Kish method. Such a procedure can have two difficulties: in some cases, we have to convince two persons in order to have an answer: one giving the household composition, the second one answering to the survey. The refusal of one is of course enough to stop the process; such a way to contact the people is sometimes seen as contradictory by the respondents: from one side, we say that it is very important that a person answers, when, in the same time, we are using a random way to “choose” this person. A lot of potential respondents do not understand that their opinions are be very important for us but, at the same time, are randomly chosen inside the household or the fact that we are perhaps looking for another person that the one actually contacted and ready to answer! According to this, an experiment is actually realised in Switzerland, trying to break the survey process in two phases: one where the household composition is done, by saying that we are doing a very brief survey on life conditions (formation, work and residential conditions), asking also the permission to recall later for another survey; a second one realised two month later with questions on social integration, political orientation and attitude about surveys. In any case, we will have some information of the people refusing the second survey but having given the household composition in the first step. A second experience is linked to this first one by giving to some respondent (randomly chosen) a relatively high incentive (15 €) in case of refusal in order to see the impact of this for refusal conversion. Keywords: Nonresponse, selection of respondent, refusal conversion Part II: Trends and Longitudinal Data The Long-Term Effectiveness of Procedures for Minimising Attrition on Longitudinal Surveys Jon Burton, Heather Laurie and Peter Lynn, UK Longitudinal Studies Centre, Insitute for Social and Economic Research, University of Essex, UK, [email protected], [email protected], [email protected] A commonly accepted view is that longitudinal surveys are both expensive to run and present data quality problems if levels of attrition from the survey are high, especially where there is evidence of differential patterns of attrition for particular groups within the sample which may introduce some bias into estimates from the data. There are also concerns about maintaining overall sample sizes for those with continuous longitudinal records as well as the size of sub-groups of interest within the longitudinal sample. Longitudinal surveys therefore go to great lengths to try and keep respondents participating in the survey using a variety of means to do so. During fieldwork, these typically involve a range of tracing procedures, refusal conversion programmes or the re-issuing of non-contacts for further attempts. These procedures are costly to implement even though they affect a relatively small proportion of the sample at any given year of a survey wave. Nonetheless they are considered central elements for ensuring the long term viability of a longitudinal survey. Despite a large literature on attrition there is little evidence on the actual value of these types of procedures for the quality of longitudinal data and it is this question which primarily concerns us in this paper. The paper uses data from the British Household Panel Survey (BHPS) from 1994 to 2002 to assess the long term effectiveness of procedures for minimising attrition in maintaining sample sizes and for data quality. We focus here on two procedures: refusal conversion, and attempts to ‘track’ mover households to a new address. Since 1994, the BHPS data has collected information on these processes 322 of tracking and refusal conversion during fieldwork. These data allow us to examine how successful these procedures have been in maintaining the viability of the sample over time. We first examine the longevity of successful refusal conversion. Are we simply postponing drop out for a wave or two, or is the procedure effective at retaining units in the sample for many waves? We then compare sample sizes and sample characteristics, both overall and for subgroups, across multiple waves, with and without refusal conversion. With regard to tracking movers to new addresses, we examine success rates and their impacts on sample sizes and sample characteristics. We also examine the combined effects of both procedures. Finally, we make a preliminary assessment of the effects of the procedures on data quality through comparing key survey estimates. Mood and Socio-Economic Status Bias in Survey Nonresponse: Results from an 11Wave Panel Femke De Keulenaer and Katia Levecque, [email protected], [email protected] University of Antwerp, Belgium, The literature on socio-economic status (SES) and psychiatric disorder bias in survey nonresponse reports inconsistent findings. The literature has dealt at length with the likelihood of SES bias, however a number of studies report curvilinear effects, other studies find that participation decreases with SES, while still others conclude that lower SES groups are more likely to participate in surveys than higher SES groups. The scarce literature on psychiatric morbidity bias in survey nonresponse reports findings going form no significant effect of psychiatric disorder on nonresponse to a moderate or strong tendency of sick persons not to participate. This paper considers the representativeness of the Panel Study of Belgian Households (PSBH) over its 11-year history form 1992 to 2002 given the dynamics of entry and exit from the panel. The PSBH informs us about a wide range of topics, such as health, working conditions, housing and social relations. The first wave of the data collection was conducted in 1992 by face-to-face interviewing in a stratified multistage probability sample. Since then, data collection was continued on a yearly basis. Both household and individual data were gathered using a standardised questionnaire, which in the former case was administered to a reference person of the household. In the latter case, all individuals aged 16 or more were asked for personal information, including depressed mood (as measured by the HDLF-scale). Earlier analyses of attrition from the PSBH revealed that attrition probabilities were not random and decline with duration. In this paper, we want to assess whether SES and a depressed mood are factors that increase the risk of noncontact and noncooperation in the PSBH by analysing survey nonresponse in two phases. We hereby account for the fact that attrition of individuals in the PSBH can be the result of a decision-making process at household level, and/or a decision of the adult individual himself. We begin our analysis by looking at the response behaviour of households. Attrition in this analysis phase reflects survey nonparticipation caused by failed contact or lack of cooperation at household level. In the second phase of analysis we concentrate on participating households and assess whether the nonresponse behaviour of individual household members within these households can also be explained by SES and a depressed mood. In order to explain why certain households or individuals are at higher risk to attrite from the panel, we will estimate year-by-year attrition probabilities using discrete-time logit models. The discrete-time logit approach is particularly effective at handling time-dependent covariates and covariate-time interactions. This makes it possible to incorporate changing respondent characteristics and to test if attrition behaviour changes with the duration of the panel. Keywords: attrition, socio-economic status, depressive mood, Panel Study of Belgian Households 323 Survey Attrition and Estimation Strategies in the Medical Expenditure Panel Survey Trena M. Ezzati-Rice and David Kashihara, Center for Financing, Access, and Cost Trends, Agency for Healthcare Quality and Research, [email protected] The Medical Expenditure Panel Survey (MEPS), sponsored by the Agency for Healthcare Research and Quality, is an ongoing longitudinal panel survey designed to produce estimates of health care utilization, expenditures, sources of payment, and insurance coverage of the U.S. civilian noninstitutionalized population. The MEPS-Household Component (HC) consists of an overlapping panel design in which each new panel of households is interviewed a total of five times over a period of 30 months to yield use and expenditure data for two calendar years. The sample of households selected for the MEPS-HC is a sub-sample of households who completed the prior year’s National Health Interview Survey (NHIS), another large ongoing national health survey. A key design feature of the MEPS is the capacity to conduct both cross-sectional as well as longitudinal analyses to address current and emerging health care policy issues. The adjustment strategy used in MEPS to compensate for survey nonresponse includes: (1) an adjustment for dwelling unit nonresponse to account of household nonresponse after round 1 among those households sub-sampled from the NHIS and (2) a nonresponse adjustment to account for survey attrition at the person level. To inform the specification of nonresponse adjustment strategies to compensate for survey attrition, this paper examines and compares the characteristics of participants in the initial round of data collection for a recent MEPS panel with those who discontinue their participation in subsequent rounds. The identified auxiliary variables are compared with those currently used in MEPS to compensate for survey attrition. Also, to assess the effectiveness of the nonresponse adjustment strategies used in MEPS to minimize the impact of nonresponse bias in the resulting survey estimates due to sample attrition, national healthcare expenditure estimates are compared using annual, panel-specific, and longitudinal data. This study is part of ongoing efforts to ensure the accuracy of the MEPS survey estimates. Keywords: nonresponse, survey attrition, longitudinal panel survey Determinants of Mail Survey Response Rates over a 30 Year Period, 1970-2000. An Analysis of Community Studies at the German Institute of Urban Affairs Volker Hüfken, Department of Social Science, Heinrich-Heine-Universität Düsseldorf, Universitätsstraße 1, D-40225 Düsseldorf, Germany, [email protected] One of the most difficult questions to answer about response rates to mail surveys concerns whether response rates have been declining in recent years. Another question about which little is known is the extent to which characteristics of particular survey populations influence the level of response. 340 studies were analysed. The final response rate averages 56 %. The type of population, the length of the questionnaire affected final response independent of reminders. Results indicate too, that year of study did have a significant effect on the mail survey response rates. Keywords: mail survey, response rate over time, Understanding why People take part in Surveys: The Case of the Families and Children (panel) Survey Miranda Philips, Shaun Scholes and Debbie Collins, National Centre for Social Research, UK, [email protected] An important factor in ensuring the robustness of panel survey data is that it continues to be representative of the population group of interest. This can be achieved in a number of different ways including: the use of a refreshed panel sample design, whereby new cases are added to the study at specified intervals so allowing changes at the aggregate level to be examined; and 324 the maintenance of the existing panel, through activities such as tracing movers, keeping in touch exercises, use of advance letters and incentives. The latter point is commonly adopted on panel surveys yet there is little evidence to demonstrate how effective such strategies are in reducing panel attrition, whether different strategies work better with different types of respondents and why these strategies work. This paper explores these issues, drawing on work carried out on the Families and Children Study (FACS). FACS is a government-funded longitudinal panel survey, which follows families with children to monitor how their circumstances change over time. The survey has been running since 1999 and families are interviewed every year. The survey includes a refreshed panel sample design to allow changes at the aggregate level to be examined. From the outset the survey also adopted a number of panel maintenance strategies aimed at minimising panel attrition, such as tracing movers, keeping in touch exercises and incentives. After the third wave of data collection in 2001 it was decided to carry out a piece of qualitative work with FACS respondents to explore: the factors affecting initial and on-going participation in the study; the factors affecting the withdrawal of co-operation; and the effectiveness of existing panel maintenance strategies. A number of recommendations for improvements to existing panel maintenance strategies were made, specifically in areas concerned with informing respondents about the nature and purpose of the study, securing respondents’ (ongoing) participation, in providing reassurance about participation in the survey, and with dealing with panel refusals. A number of these recommendations were implemented in the subsequent round of the study (2002). Analysis of response data allowed us to evaluate the impact of the implementation of these strategies as a whole on reducing panel attrition for the survey, both overall and for different types of respondents. This paper presents the findings of this evaluation and makes recommendations for further research. Keywords: panel attrition, panel maintenance strategies, respondent’ motivation Nonresponse trends in the United States National Health Interview Survey, 1997-2003 Howard Riddick, Survey Planning and Development, Division of Health Interview Statistics, National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, Maryland, USA. The National Health Interview Survey (NHIS) is one of the major data collection systems of the National Center for Health Statistics of the Centers for Disease Control and Prevention. The NHIS uses a multistage sample designed to represent the civilian noninstitutionalized population of the United States. A new sample is drawn for each annual survey. The questionnaire is administered in the household using computer-assisted personal interviewing (CAPI) and has three major collection components: the family, the sample child and the sample adult. Basic information is collected on all family members from an adult living in the household. Additional information is collected for one randomly selected sample adult and one randomly selected child. A knowledgeable adult reports for the sample child, while the sample adult self-reports. Trends in nonresponse will be presented for the household, sample adult, and sample child. Efforts undertaken to improve or maintain response rates will be presented and discussed. Such efforts include an interviewer incentive program, modifying the interview close out period in one region to allow for more weekend days for interviewing, conducting a split sample test of refusal aversion training emphasizing active listening and role playing, and the sponsorship of a response rate summit in February 2002. Sponsored by the United States Census Bureau and the Interagency Household Survey Nonresponse Group, the purpose of the summit was to provide a forum for discussion among experts in the field about how to address concerns related to the decreasing response rate trend in household surveys. Expert panel members included representatives from university survey centers, private survey companies, and from Statistics Canada. Panel recommendations will be discussed. Keywords: nonresponse 325 Part III: Cinficence in Surveys and Motivation to Answer Understanding Nonresponse of the Twelfth Graders in the National Assessment of Education Progress: Social Isolation Theoretical Approach Young Chun, American Institutes for research, 1990 K Street NW Suite 500, Washington DC 20006, [email protected] The purpose of this paper is to explain ‘nonresponse of twelfth graders’ in national assessments such as the National Assessment of Educational Progress (NAEP) according to “social isolation theory” and using models of nonresponse measurement developed by Groves and Couper (1998). We have used data on 12,000 twelfth graders in the 2000 High School Transcript Study (HSTS) and linking 2000 NAEP surveys of students, teachers and principals all collected by the National Center for Education Statistics. By identifying factors central to reducing nonresponse at three levels of analysis (i.e., student, school, and social-psychological perception), we have the findings that address both theoretical implications for the external validity of a social isolation theory of nonresponse and practical implications about measures of interventions to be recommended to reduce nonresponse in the national assessments of students. Unit Nonresponse among Foreigners Caused by Non Willingness to Participate in Survey Research. An International Comparison of Strategies Dealing with the Problem Remco Feskens, Joop Hox, Hans Schmeets, Gerty Lensvelt-Mulders, Department of Methodology and Statistics, Utrecht University, Heidelberglaan 1, 3584 CS Utrecht, tel. ++31-30-253 14 90, [email protected] High non-response rates among ethnic minorities are a major cause of concern. There are in particular problems with the willingness to cooperate in a survey- earlier research suggests that refusals to cooperate are the main cause of non-response. The major consequences of non-response, increasing costs for reaching a desired response rate and more seriously, the reduction of validity are especially a problem among ethnic minorities because they are likely to differ systematically from other groups. In this paper we will make an international comparison of strategies used to reduce non-cooperation among ethnic minorities. This because little is known about eventual differences in strategies trying to minimize non-response among ethnic minorities caused by refusals. We discuss an international comparison of the non-willingness of foreigners to participate in survey research, their profile and the tailor made strategies used by the national statistical agencies of the United Kingdom, France, Germany, Finland, Sweden and the Netherlands. The effectiveness of these tailor made strategies will be evaluated. It will not only be of theoretical interest to see what strategies other countries use in trying to reduce nonresponse among foreigners. This paper will also present some useful recommendations regarding tailored strategies in persuading ethnic minorities to participate in survey research. This may help increase the validity of survey research and ideally make it also more cost-effective than the present used strategies. Keywords: unit nonresponse, refusals, hard to contact groups, tailor made strategies. 326 Task Dedication, Task Alienation and What Data Collection Instruments Mean to Respondents Patricia A. Gwartney1, Vikas Kumar Gumbhir1 and Anthony Leiserowitz2 1 Department of Sociology, 1291 University of Oregon, Eugene OR 97403-1291, USA, 541-346-5007 [email protected], [email protected] 2 Decision Research, 1201 Oak Street, Eugene OR 97401, USA, 541-485-2400 [email protected] This paper continues the authors’ explorations of the meanings survey respondents ascribe to surveys, censuses and political polls and how those meanings affect latent aspects of survey participation. We find that survey respondents’ meanings about data collection instruments, as well as their ability to visualize those meanings, independently affects item response and nonresponse. We frame the research with symbolic interactionist concepts we call “task dedication” and “task alienation” and affective imagery analysis from word association research in cognitive psychology. Survey instruments include keywords, e.g., “community” and “election,” that evoke different subjective meanings in respondents and vary by their knowledge, beliefs, values and demographic characteristics. Yet methodologists have paid little attention to keywords in survey introductions, questions, and answer categories – and even the word “survey”. For the past few years, we have sought to uncover certain keywords’ differential meanings for respondents and to map their distribution across populations. In one subproject, we experimented with using “survey,” “census” and “political poll” as stimuli in word- and image-association questions. We randomly assigned 806 respondents to receive one of the following question pairs: What is the first thought What is the first thought that What is the first thought that that comes to mind when comes to mind when you comes to mind when you the term “U.S. hear the term “political you hear the word hear poll”? census”? “survey”? And what is the first picture And what is the first picture And what is the first picture or image that comes to or image that comes to mind or image that comes to mind mind when you think of a when you think of the “U.S. when you think of a “political poll”? census”? “survey”? To explore conceptual ambiguity in respondents’ understandings of these terms, we employ affective image analysis to classify their subjective meanings and multivariate analysis to examine their correlates. Respondents who answered the word association questions with a concrete visual image about a survey, census, or political poll we label “visualizers.” One key finding is that survey respondents’ ability to visualize significantly correlates with taskdedicated behaviour as they completed the interview. Specifically, visualizers exhibit significantly reduced item nonresponse (“don’t know” and “refuse” answers), more words answering open-ended questions, and shorter surveys, net of standard control variables, including cognitive sophistication. We find that item response and nonresponse also varies consistently by certain meanings that survey respondents assign to data collection instruments. Keywords: Item nonresponse, telephone surveys, symbolic interaction, affective imagery Measuring the Respondent’s Attitude towards Surveys. Geert Loosveldt and Vicky Storms, K.U. Leuven, Department of Sociology, Van Evenstraat 2B, Leuven, Belgium Measuring the respondent’s attitude towards surveys is not obvious. In the paper some of the measurement problems will be examined. The key problem is that a survey must be used to measure the respondent’s attitude towards surveys. One of the most problematic aspects of collecting information about attitude towards surveys by means of a survey is the fact that it is not possible or a least extremely difficult to measure this attitude for a relevant group of people who refuse to participate in the interview. One can assume that the respondent’s attitude has an effect on the 327 respondent’s decision to participate in the interview and that refusers have a more negative attitude than respondents. As a consequence the measurement of the respondent’s attitude towards surveys is extremely sensitive for non response error and will be biased in a positive direction. The analysis of the relationship between the respondent’s attitude and indirect indicators of the respondent’s willingness to participate in an interview (doorstep reactions, evaluation of doubts to participate, evaluation of future participation in surveys, …) will be used to investigate this assumption. One can also assume that the respondent’s attitude towards surveys is related to other substantive attitudes which are relevant to ‘the survey taking climate’ (feelings of insecurity, trust in institutions, political alienation, social involvement …). Results of this analysis in combination with information about the bias in the measurement of the respondent’s attitude towards surveys will be used to evaluate measurement bias in these substantive attitudes. In the analysis data will be used from the survey ‘Cultural shifts in Flanders: 2002’. In this survey a ‘General attitude towards surveys scale’ was used. The results of the analysis support the assumption that the measurement of the respondent’s attitude towards surveys is biased in a positive direction and that this information can be used to evaluate bias in the measurement of other substantive attitudes. Keywords: non response error, measurement error, the survey taking climate Does the Performance Matter? A Multilevel Analysis of the Effects of Individual and Performance Characteristics on Response in Audience Research Henk Roose, John Lievens and Hans Waege, Ghent University, Ghent-Social Science Research and Methodology & Cultural Policy Research Centre ’Re-Creatief Vlaanderen’, Belgium, [email protected] This paper focuses on the effects of characteristics of the performance on the decision to cooperate in an audience survey. Respondent related correlates of survey response, such as topic salience, gender, age, educational attainment, etc. were already analysed in the specific survey setting of audience research. Little is known, however, if and to what extent features of a performance might influence survey response in an on-site audience survey. Non-contacts in on-site audience research tend to be related to the on-site physical impediments faced by the contacting agencies and with the characteristics of the playhouse in which the performance is staged. Characteristics of the audience and the performance itself are less salient. With refusals this is different. Both individual as well as performance features influence the inclination to cooperate. Using a multilevel logistic regression model we map response behaviour in audience research, simultaneously taking into account the characteristics of the members of the audience and some features of the performance, such as length, genre, complexity, size of the playhouse, etc. Results are interpreted within an explicit theoretical framework based on rational choice and social exchange perspectives. We collected the data by means of a two-stage sampling procedure and analyse the survey participation to a postal survey in the second stage, ignoring the nonresponse in the first. The data are part of an extensive audience survey in cultural institutions in Flanders, in which the audience of more than 50 cultural institutions was enrolled in a study on arts participation. Keywords: audience research, response, multilevel analysis Survey Experience, Generalized Attitudes Towards Surveys and Respondents’ Willingness to Answer Questions in Subsequent Surveys Volker Stocké, Project Program SFB 504, "Rationality Concepts, Decision Making and Economic Modeling", L13,15, University of Mannheim, D-68131 Mannheim, Germany, Tel: +49(0)621-1813432, Fax: +49(0)621-181-3451, [email protected] This paper analyzes the effects of the evaluation by respondents’ of their last survey interview on their attitudes towards surveys in general and whether these attitudes have an effect on their cooperation in 328 future surveys. Data from a locally defined random probability sample is used to answer these questions. Firstly, subjects are found to differentiate between three dimensions when evaluating their survey experience. These dimensions are the burden respondents experienced during their last survey interview, how they judged its ‘entertainment value’ and how irritated they felt themselves to be by confusing questions. According to our second finding, only the burden experienced affects how subjects evaluate surveys in general. This association is found to be substantially stronger when the evaluation of the burden experienced in past interviews is cognitively more accessible as measured by the time needed to generate these judgments. This accessibility is found to be affected by the mode of administration of subjects last survey interview. According to our third result, respondents’ with a critical attitude towards surveys in general are more likely in the complete interview to answer ‘don’t know’, to refuse to answer questions and to be judged by the interviewer to be in general less willing to answer questions. However, the predictive power of attitudes towards surveys for respondents’ cooperation depends on the cognitive accessibility of these evaluations: fast attitude judgments are the better predictor for the subjects’ willingness to answer questions. The cognitive accessibility of attitudes towards surveys is found to increase with the number of subjects’ survey interviews in the past. In summary, the burden experienced by respondents in past survey interviews impairs their evaluation of surveys in general and thereby increases the prevalence of non-response in following surveys. In this way, survey researchers themselves partly determine the conditions for their future work. Keywords: Attitudes towards surveys, item non-response, respondents’ burden, response latencies, survey experience SESSION: AIDED RECALL TECHNIQUES IN SURVEY INTERVIEWS Organizors: Wander van der Vaart Contact: Vrije Universiteit, Faculteit der Sociale Wetenschappen , Afdeling Methoden en Technieken, De Boelelaan 1081-c , 1081 HV Amsterdam, The Netherlands, [email protected] Keywords: Questioning procedures, recall accuracy, autobiographical memory, empirical tests Abstract In many disciplines information on past behavior and life events is central to scientific theories and empirical research. The most frequently used method to obtain this information, is to reconstruct the past by means of retrospective questions. Since retrospective data often suffer from recall error, social researchers at times employ aided recall techniques to enhance data quality. Respondents are provided with checklists, cues, landmarks, time frames and the like. Timeline procedures and event history calendars integrate a number of such techniques and appear to be promising instruments in survey research. Still, the application of aided recall techniques often concerns ad hoc solutions that lack a proper theoretical framework. The session concerns research on theoretically based principles of aided recall techniques in surveys and their empirical tests. Center of attention are the (social) cognitive processes underlying aided recall techniques and their implications for questioning procedures and questionnaire design. While the primary focus of the matter is on retrieval processes, connections may exist with each of the major components of the response process - comprehension, retrieval, judgment and estimation, reporting. In addition, issues may be raised as to what extent aided recall techniques can go together with standardization of survey research and whether they can be tailored to different types of respondents, recall tasks and modes of data collection. 329 Improving the Quality of Retrospective Reports: Calendar Interviewing Methodologies Robert F. Belli, University of Nebraska-Lincoln Gallup Research Center, USA, [email protected] Event History Calendars (EHCs) are predicted to encourage respondents to use idiosyncratic cues available in the structure of autobiographical memory, improving the quality of retrospective reports. Although structured, EHCs encourage a more flexible interviewing style in comparison to traditional standardized Question-List (Q-list) methods. Such flexibility allows the introduction of effective retrieval cues that include already-reported events from respondents’ pasts that cannot be easily scripted in advance. Three methodological studies have directly compared the quality of retrospective reports from EHC and Q-list interviews. 1) The Panel Study of Income Dynamics (PSID) 2-year Calendar Methods Study used paper and pencil instruments and focused on events that occurred 1 to 2 years previously. Interviewers and respondents were randomly assigned to EHC (n = 309; 84% cooperation rate) and Q-list (n = 307; 84% cooperation rate) conditions. Prior year panel reports from the PSID were used as a standard of comparison. In data quality comparisons that examined retrospective reports of the occurrence and timing of residence changes, employment and unemployment histories, spells of missing work, and receipt of social entitlements, the EHC significantly outperformed the Q-list in 12 of 23 comparisons, and the Q-list provided better retrospective reports in only one of them. 2) The Lifecourse Intimate Partner Violence Study used paper and pencil EHC (n = 40) and Q-list (n = 359) instruments in face-to-face interviews. Although there were no validation data, results supported the EHC providing better quality retrospective reports of intimate partner violence among a cohort of African American women aged between 18 and 54 years. Specifically, the EHC provided significantly more reports of intimate partner violence, and reports of first exposures at younger ages for the older women. 3) The PSID Lifecourse Health and Economic Measures Study used computerized instruments. Lifecourse PSID panel reports were used as a standard of comparison. Respondents and interviewers were randomly assigned to conditions, with data collected on 313 EHC (94% cooperation rate) and 318 Q-list (97% cooperation rate) respondents. For most outcome measures, the EHC outperformed the Q-list in data quality, including residential changes, cohabitation, amount of annual employment, and health status history. There was no advantage in EHC interviews for unemployment and marriage histories. EHCs provide better quality retrospective reports in both telephone and face-to-face modes, in both paper and pencil and computer-assisted instruments, and in shorter (2-year) and lifecourse reference periods. Advantages of EHC interviews occur without a substantive increase in interviewing time. Keywords: Retrospective reports, calendar methods, data quality Collecting Event History Data about Work Episodes Retrospectively: What Recall Errors Occur and how can we Prevent them? Maike Reimer, Max-Planck-Institute for Human Development, Center for Sociology and the Study of the Life Course, Berlin, Germany, [email protected] Event histories describe time periods as temporally ordered sequences of events or episodes and transitions that capture the individual’s movement through a state space reflecting relevant institutional features of a given life domain. In this paper, I examine how and how reliably episodes and transitions are constructed when collecting work histories in a retrospective survey, and describe the improvements in the data collecting procedure that were developed to prevent recall error. I compare data from standardized interviews in the “East German Life History Study" (EGLHS) that were collected once in 1992 and once in 1996. The recall task in the EGLHS is threefold: respondents have to 1. place their work-related activities and roles at a given point in the past within the state space; 2. reconstruct a transition into the next state, and 3. date start and end of the episode down to a month. Inconsistencies between reports are produced at all three stages of the reconstruction process. In the later interview, about 44 % of the respondents omit or insert episodes, fuse or de-fuse episodes, 330 characterize episodes and transitions differently, move transitions into and out of the reference period or produce completely different sequences. Overall, the later sequences show less change and are more conventional, but changes in any direction occur. Reconstructions are also influenced by the changes of labor market institutions and experiences from 1992 to 1996. Inconsistent reports are especially likely where the organizational structure of memory does not correspond well to the state space. Based on these findings, a computer-assisted data collection instrument for telephone interviews called “L-DEX” is currently being developed, which helps interviewers and respondents to cooperatively reconstruct the sequences as intended. In a calendar-like module they can go back and forth in time and relate episodes within and across domains to access individual memory structure and to check and resolve temporal inconsistencies. Special attention is paid to the crucial role of the interviewer as “reconstruction guide” and interface between social scientific concepts, data/standardization requirements, and memory processes. Keywords: event history data, aided recall techniques, recall accuracy, Influence Dating Formats on the Telescoping Effect and the Distribution of Autobiographical Memory Steve M. J. Janssen1, Antonio G. Chessa1 and Jaap M. J. Murre2 1 University of Amsterdam, The Netherlands, [email protected] 2 University of Amsterdam, The Netherlands and University of Maastricht, The Netherlands The effect of different formats for dating news and personal events was examined in two experiments. In the first, subjects could choose between ago (e.g. “three weeks ago”) and calendar formats (e.g. “June 15, 2003”), while in the second, they were assigned to either the ago or the calendar format condition. We measured temporal displacement of news events and distribution of autobiographical memory ages. We found in both experiments backward telescoping for recent news events and forward telescoping for less recent news events. The ago format led to smaller errors in dating recent events, but larger errors in older events. Furthermore, in the second experiment participants who had to date personal events in the ago format reported more events that occurred recent, while subjects answering in the calendar format reported more events that occurred less recent. Keywords: Dating formats, telescoping, autobiographical memory. Timeline Methods and Related Aided Recall Applications in Social Sciences and Health Studies: a Review. Wander van der Vaart, Department of Social Research Methods, Vrije Universiteit Amsterdam, The Netherlands, [email protected] In the research practice of sociologists and health scientists concerns about the quality of retrospective reports has given rise to the use of so called timeline methods. These aided recall procedures aim to help respondents to gain better access to long-term memory by employing a graphical time frame in which life history information can be represented. Most considerations on timeline methods are to be found in life course research in sociology and especially in health behaviour and treatment studies in health sciences. Both disciplines focus on the retrospective reconstruction of (series or sequences of) behaviour and events during an extended time period and share similar concerns about recall bias. In life course studies, respondents are typically asked to recall their history on domains like employment, education, relationships, residences, financial behaviour, et cetera. Health studies usually focus on more specific topics like risk behaviour (e.g., alcohol and drug use, sexual behaviour) and medical events (e.g., cancer diagnosis, clinical treatments, hospitalisation). Timeline methods are widely employed in the health science, but they often lack sufficient theoretical foundation. In quantitative life course studies relatively little attention has been devoted to research on 331 timeline methods, though some research indicates that they may have beneficial effects on data quality. This paper reviews applications of timeline methods in both disciplines in order to obtain more insight into the potential benefits of timeline methodology and its effective mechanisms. Keywords: Timeline methods, aided recall, recall accuracy, autobiographical memory SESSION: DATA COLLECTION Organizers: Katja Lozar Manfreda and Marek Fuchs Contact: Faculty of Social Sciences, University of Ljubljana, Kardeljeva ploscad 5, 100 Ljubljana, Slovenia, [email protected] Abstract In recent years data collection techniques have diversified: web-enabled surveys, mobile phone surveys, electronic diaries, mixed mode surveys are just few of the recent developments. Many of these data collection techniques have undergone serious assessments in terms of their impact on coverage, non-response and measurement error. Others have received little attention from a methodological point of view so far. The session aims to bring together scholars from different backgrounds and organizational settings in order to discuss the methodological impact of the new data collection techniques on the survey process. Hence it is the goal of the session to improve the awareness that even though these data collection techniques are new in terms of the technology or the communicational channel involved, they are still to be judged according to the traditional standards of data collection. Thus, making use of the advantages of these techniques should not compromise well established criteria of data quality. Papers presented in this session should report results from evaluations using one data quality indicator or another. If possible, papers may report results from experimental comparisons with other data collection techniques. Patterns of Mobile Phone Usage and their Impact on Participation in Mobile Phone Surveys Vasja Vehovar, University of Ljubljana, Slovenia, [email protected] The popular use of mobile phones in developed countries created also an increasing trend of fixed telephone noncoverage. As the majority of EU countries crossed the 80% penetration rate of mobile phones, they also crossed or approached the 10% share of mobile-only households. Finland is on top of this trend with 35% of mobile-only household in 2003, what actually prevents the fixed telephone probability sample surveys of the general population. There exist sever problems with this surveys mode, particularly the issues of costs, frames and legislation. However, the experiences with mobile phone interviewer surveys in many countries showed that the cooperation problems remain on approximately same level as with fixed telephone surveys. This paper concentrate on another potential problem of this survey mode: the impact of the attitudes on the mobile phone usage (including the patterns of usage) on survey participation. For this purpose, the representative sample as well as the mobile phone sample were conducted: both included the questions that characterized the segments of users according to the usage pattern and corresponding attitudes. Based on this data, the users were structured into three segments: intensive pragmatic users, intensive emotional users and less intensive users. The basic hypothesis was that the first and the third group showed some specific contact rate patterns. However, with the second group, composed of emotional users (they care about their mobile device and they developed certain relation or attachment with it), we encountered specific measurement problems. 332 Comparing Anonymous and Non-Anonymous Research Methods Asking Sensitive and Socially Desirable Questions. When You See What Do You Get? Differences between Face-to-Face, Online and DesktopVideoconferencing Research Methods asking Sensitive and Socially Desirable Questions. Hans-Ullrich Mühlenfeld, University of Bremen, Institute of Empirical and Applied Sociology (EMPAS) Celsiusstrasse, 28359 Bremen, Germany, Tel: +49 421 218-2673, Fax: +49 421 218-7474, [email protected] Research findings show that sensitive questions and questions about socially desirable behaviour provoke a) the tendency to give a socially desirable answer and b) variance between more or less anonymous interview modes. This paper investigates the tendency to give socially desirable answers between three interview modes: a Computer Assisted Personal Interview (CAPI), a web-based questionnaire (ONLINE) and an innovative method called Internet Assited Personal Interview (IAPI), which uses a desktop video conference system to facilitate communication between the interviewer and the interviewee. All three methods differ in anonymity, with the CAPI as the least anonymous interview situation and the ONLINE as the most. Consequently, the interviewee's reaction to sensitive questions and questions concerning socially desirable behaviour should differ between modes. The results of a study using 90 undergraduate psychology students revealed interesting differences between the three interview modes, showing that the IAPI is almost as efficient at asking such questions as the CAPI. Keywords: social desirability, computer mediated communication, interview mode effects, video conference Improving Questionnaire Design through the Provision of Electronic Access to Large Scale Survey Questionnaires in the UK Martin Bulmer, Harshad Keval and Julie Lamb, ESRC Question Bank, Department of Sociology, University of Surrey, UK, [email protected] The construction of questionnaires is an essential element in improving data collection techniques in social survey research. The ESRC Social Survey Question Bank is an innovative attempt within the UK social research community to make available the questionnaires from major social surveys via a freely accessible web site: URL; http://qb.soc.surrey.ac.uk. The surveys selected for inclusion are in the main major national surveys using probability sampling, the majority conducted by the Office for National Statistics and the National Centre for Social Research. The Question Bank holds the majority of these survey questionnaires in PDF format, easily accessible and searchable via the WWW, amounting currently to approximately 30,000 pages, extending back to 1991. The whole of the United Kingdom, including Northern Ireland, is covered. Some fifty surveys are at present included. The Question Bank, established as part of the UK Centre for Applied Social Surveys, was set up in 1995 to improve the quality of UK survey research and in particular to try to improve survey measurement. In addition to providing access to questionnaires, many of which were not readily accessible previously, the Question Bank attempts to present commentary upon the measurement of different survey variables, written by experts, as well as to provide access to the Harmonisation project of the Office for National Statistics which seeks to standardise survey measurement across government surveys. This aim of trying to achieve a degree of concordance in survey measurement across different surveys is, as others have found, extremely problematical. The approach adopted by the Question Bank to this problem has been an inductive one, to make a very large amount of material available to easy access on the WWW, and to invite the user to search this material using a powerful search engine on the QB site. This requires ingenuity and insight on the user’ part to bridge the gap between concept and variable, or between the abstract idea and the actual question. Sometimes this is easier to achieve than in other cases. The commentary is intended to sensitise the user to some of the issues which can arise, but the amount of material on converting 333 concepts into variables is not extensive. The Question Bank relies upon the user drawing their own conclusions about the comparability of questions and the most appropriate questions to include in the type of survey which the user is designing. Keywords: Question Bank, questionnaire, question, survey design, WWW Effects of Survey Sponsorship and Mode of Administration on Respondents’ Answers about their Racial Attitudes Volker Stocké, Project Program SFB 504, "Rationality Concepts, Decision Making and Economic Modeling", L13,15, University of Mannheim, D-68131 Mannheim, Germany, Tel: +49(0)621-1813432, Fax: +49(0)621-181-3451, [email protected] This paper analyzes whether or not the type of survey sponsor affects respondents’ answers about their racial attitudes, and if so, which processes might be responsible for this effect. This is done with a split ballot experiment where the survey sponsor was either specified to be a university or a commercial company. It was assumed that respondents perceive positive racial attitude answers to be more socially desirable when they are interviewed by an academic sponsor and that they adapt their response behavior accordingly. One explanation for such sponsor-induced effects is that they are based on impression management strategies and are entirely interviewer mediated in the sense that subjects assume interviewers to share the desirability judgments of the respective sponsor organization. This is tested by randomly assigning subjects to either an interviewer- or a self-administrated mode of data collection, whereby the interviewers’ ability to perceive and sanction the responses is systematically varied. Respondents from a random probability sample (N=218) answered the items of the blatant- and subtle-prejudice scales after being randomly assigned to one of the four experimental conditions. As expected, university sponsorship provoked more positive racial attitude answers. However, these differences were only found for interviewer- but not for self-administered interviews. Starting from an identical average response behavior under the condition of high response privacy, the university sponsor provoked more positive and the commercial sponsor more negative attitude reports when interviewers recorded the answers. This interaction effect between the type of sponsor and the privacy conditions proved to be statistically significant for the responses on the blatant- but not for the subtleprejudice scale. In summary, different types of sponsors seem to influence the expectations respondents perceive on the part of the interviewers and thereby subjects’ beliefs about which answers are situational instrumental in order to obtain social approval. Keywords: Mode of administration, racial attitudes, social desirability, sponsor, SESSION: SAMPLING METHODS Organizor: Siegfried Gabler Contact: Centre for Survey Research and Methodology, Mannheim, Germany, [email protected] Abstract The session wil present new developments on the field of sampling and/or estimation methods. There are papers dealing with issues such as: Adaptive sampling, Coordinate sampling, Telephone sampling, Small area estimation , Variance estimation, Weighting, Design effects. 334 Part I Comparing Sampling Frames Renáta Németh and Ildikó Zakariás, Eötvös Loránd University of Sciences, Hungarian National Center for Epidemiology, Budapest, Hungary, [email protected], [email protected] In cases when the target population is the adult population of a certain area, it is reasonable to use an accessible list of adults for the sampling design that is usually a multi-stage, stratified design with equal sampling probabilities of individuals (Design A). However in certain cases (area sampling, telephone surveys) this type of survey design is impossible to apply; the same holds in cases when there is no such list of individuals available. This problem is often sorted out by using a different type of two-stage design: households are selected first with almost equal sampling probabilities, and then one adult member of each selected household is chosen applying a quasi-random procedure e.g. Leslie Kish grid, last birthday method etc (Design B). When evaluating the representativity of samples obtained by design B, researchers often refer to the undersampling of males and overrepresentation of elderly people, originating the phenomenon from the practical realisation of the interview, e.g. males being more difficult to find at home, and less willing to participate. In our paper some theoretical evidence will be given that explains the representation problems without considering these assumptions (Renata Nemeth, working paper, 2003, www.lisproject.org). According to our findings the major disadvantage of design B to design A lies in the resulting unequal selection probabilities of individuals that can have severe consequences: the derived sample will not be representative for even basic demographic characteristics of the population such as age or sex. It will be proved that this effect is inherently connected to the sampling method and not to practical problems. The literature does not prove this connection, nor does it normally mention it. According to the results the expected performance of design B will vary country by country; this effect will be examined for different countries where design B is in use. In our lecture we will also provide empirical evidence underlining the theoretical significance of the problem: the results of the Hungarian National Health Interview Survey conducted by the Hungarian National Center for Epidemiology (HNCE) in 2000 and 2003 applying design A will be directly compared with other Hungarian surveys applying design B; the differing levels of representativity will be clearly revealed. The different types of weighting methods (Horwitz Thompson estimator) and the differing design effects will also be examined. In cases when the implementation of both design A and design B is feasible, it may be worth considering the above mentioned aspects. We did the same in 2002 when the HNCE joined the WHO World Health Survey covering 70 countries. Although the WHO sampling plan suggested design B with the Leslie Kish grid as a sampling design, referring to the above findings and also to the existence of a proper quality list of the Hungarian adult population we suggested the implementation of design A instead; the proposal has been accepted by the WHO. Keywords: sampling frame, Leslie Kish grid Methods for Achieving Equivalence of Samples in Cross-National Surveys Peter Lynn1, Siegfried Gabler2, Sabine Häder2 and Seppo Laaksonen3 1 Institute for Social and Economic Research, University of Essex, UK, [email protected]; 2 Zentrum für Umfragen, Methoden und Analysen (ZUMA), Mannheim, Germany, [email protected], [email protected]; 3 University of Helsinki, Finland: [email protected] A central objective of the European Social Survey (ESS) is to improve the quality of cross-national survey design and implementation. This requires advances in the standards of survey methodology in 335 many countries and advances in the conceptualisation and implementation of ideas of standardisation and comparability. In this paper, we will describe the procedures that were used to ensure equivalence of the sample designs used in the 23 countries participating in round 1 of the ESS. This involved the careful specification of the permissible parameters of the sample design as well as ongoing interaction between the ESS sampling panel and the ESS national co-ordinators in each country. A number of aspects of these procedures were novel and we will focus particularly on these and will attempt to evaluate the success of each. These include: Specification of the national sample size in terms of the “effective sample size” of interviews; Provision of guidance on how to predict the design effect due to variable selection probabilities ( DEFFP ) and the design effect due to clustering ( DEFFC ), and how to use these predictions to determine the necessary sample size; Intensive procedures for interaction between the members of the sampling panel, between the sampling panel and the national co-ordinators, and between the sampling panel and the ESS central co-ordinating team. Keywords: Cross-national surveys, design effects, European Social Survey, sample design Design Effects in the Analysis of Longitudinal Survey Data Chris Skinner and Marcel de Toledo Vieira, University of Southampton, United Kingdom, [email protected] The design effect measures the inflation of the sampling variance of an estimator as a result of the use of a complex sampling scheme. It is usually measured relative to the variance of the estimator under simple random sampling. Many social survey designs employ multi-stage sampling, leading to some clustering of the sample and this tends to lead to design effects greater than unity. There is some empirical evidence that design effects from clustering tend to decrease the more complex the analysis. For example, design effects for regression coefficients are often found to be less than design effects for the mean of the dependent variable in the regression. Evidence of design effects close to unity for such analyses may be used by some analysts of survey data to justify ignoring the sampling design in complex analyses. In this paper we present some evidence of an opposite tendency, for design effects to be higher for complex longitudinal analyses than for corresponding cross-sectional analyses. Our empirical evidence is based upon data from the British Household Panel Study. This survey follows longitudinally a sample of individuals selected in 1991 by two-stage sampling, with clustering by area. Data are collected in annual waves. Our analyses are based upon a subsample of women aged 16-39. The dependent variable is a gender role attitude score, derived from responses to six five-point questions, and treated as a continuous variable. Covariates include age group, economic activity and educational qualifications. Longitudinal regression models include random effects for women. Data are analysed for five waves of the survey when the gender role attitude questions were asked. The design effects for the regression coefficients are found to increase the more waves are included in the analysis. A similar tendency is observed for estimates of the time-averaged mean of the dependent variable. A possible theoretical explanation is provided. The implication of these findings is that standard errors in analyses of longitudinal survey data may be very misleading if the inital sample was clustered and if this clustering is ignored in the analysis. Keywords: clustering; design effect; longitudinal analysis; random effects model; Impact of the Geographical Size of Clusters on the Precision of Survey Estimates Kevin Pickering and [email protected] Sarah Tipping, National Centre for Social Research, UK, The samples for most large-scale face to face surveys in the UK are selected to be clustered within geographical areas. This is done to make the fieldwork more manageable and, as a result, reduces the 336 costs for the survey (for a fixed sample size). There is a negative impact of clustering a sample and that is a reduction in the precision for survey estimates – the greater the level of clustering, the lower the precision of the estimates. This reduction in precision is observed because respondents within the same cluster tend to be ‘more similar‘ than people in the general population. In order to issue larger sample sizes within a set budget, there has been a tendency in recent years for the samples for some Government surveys to be clustered within smaller geographical areas. Therefore, rather than samples being clustered with Postcode Sectors (where the average number of addresses is 2700 per sector) or Wards (where the average number of addresses is 2360), there has been a trend towards clustering samples within half- or quarter-Postcode Sectors or Wards. This presentation will investigate the impact of different geographic sizes for clusters (full-, half- and quarter-postcode sector) on the precision of estimates for a number of Government surveys, including the Health Survey for England, the National Travel Survey and the Survey of English Housing. By comparing the loss in precision from using smaller clusters against the gains in cost efficiency, we will assess the net effect on efficiency (i.e. precision per cost) of different levels of clustering on survey estimates. Keywords: clustering, precision, cost efficiency, Postcode Sector Part II GenD: An Evolutionary Algorithm for Re-sampling in Survey Research Daria Croce1 and Cinzia Meraviglia2 1 University of Milan-Bicocca, Italy: [email protected] 2 University of Eastern Piedmont, Italy: [email protected] The paper presents a method for generating new records in databases of limited size (200-500 cases) by means of an evolutionary algorithm (close to, but different from, a genetic algorithm). This method can be of aid when budget constraints limit the number of interviews, or in case of a population that shows some sociologically interesting trait, but whose small size can seriously affect the reliability of estimates, or in case of secondary analysis on small samples. The applicative ground is given by causal research designs with one or more dependent and a set of independent variables. The estimation of new cases – i.e. cases inferred by the algorithm on the basis of the observed density function of the variables considered – takes place according to the maximization of a fitness function (see, among others, Eberhardt, Simpson and Dobbins 1996). The procedure outcomes a number as large as needed of new “virtual” cases, which (should) reproduce the statistical traits of the original population. The algorithm used is GenD, designed and implemented by Semeion Reserach Centre (Rome) (Buscema 2004); among its features there is an innovative crossover procedure, which tends to select individuals with average fitness values, rather than those who show best values at each “generation”. Our aim is to show how the estimation of “new” cases takes place, as well as to evaluate it from a methodological point of view. Some preliminary trials show that the “virtual” data generated by the algorithm are not only close to the observed sample in respect of many statistical indicators (mean and standard deviations, value range, univariate and multivariate distribution, etc.), but they also consent better estimation of the parameters of causal models. The quality of the new data will be tested by means of traditional statistical models as well as neural networks, according to a particularly thorough design (see Fig. 1): 1) the observed sample is half-splitted, only a half is analyzed via traditional techniques or neural networks; 2) cross-validation is performed to find out how good the parameter estimates are; 3) GenD calculates the pseudo-inverse of the estimated parameter matrix; 4) the best individuals at each generation form the “virtual” new population; 5) “virtual” data are tested against the second half of the observed sample (which has never been used since now, except than in crossvalidation); 6) if virtual data consent a better or equal performance (i.e. better estimate of the inverse of the parameter matrix), routine statistical analysis is performed to examine the functional distribution described by the new data. 337 As a new set of trials, GenD will be applied to a small data set from a survey on the role social capital in the reproduction of inequalities in a local area of North Western Italy. The questionnaire generated over 500 variables, mostly regarding the working life history of the interviewees. This new trial is part of a set of analyses to be performed in the context of a national research project for updating the Italian Scale of Social Desirability of Occupations (de Lillo and Schizzerotto, 1985). Fig. 1. Sample (N = 400) Training sample (N = 200) Causal analysis (standard techniques/ neural networks) Testing sample (N = 200) Cross-validation GenD → pseudo inverse parameter matrix (testing sample left aside) “new” records Causal analysis performed on the “new” sample; results tested against the testing sample Analysis of the References “new”sample Buscema M. (2004, forthcoming), Genetic doping algorithm (GenD): Theory and applications, Expert Systems, May, vol. 21, n. 2 de Lillo A. and Schizzerotto A. (1985), La valutazione sociale delle occupazioni (Social evaluation of occupations), Bologna, Il Mulino Eberhart R., Simpson P. and Dobbins R. (1996), Computational intelligence PC tools, London, Academic Press Sampling Process-Generated Data in order to Trace Social Change Nina Baur and Christian Lahusen, University of Bamberg, Germany, [email protected] When analyzing social change, researches frequently face the problem that new research questions arise. For these questions almost never survey data are available. Very often, the can researchers neither use retrospective interviews. In such cases, researchers have only two possibilities: They can start a new survey. However, this excludes the possibility of studying former change which might be the actual research question. The alternative is to construct a new data set using process-generated data. Social scientists rarely reflect sampling problems arising from and strategies for using this data type. Using the media discourse on unemployment as an example (focussing on newspapers for practical reasons), we suggest solutions for process-generated data. Researchers face three types of over layered sampling problems: Bias of Available Data I (Data Production Process): When using process-generated data, researchers cannot control the process of data production. Data are already biased during production. Bias of Available Data II (Data Selection Process):The more time passes, the more of the original data decay or are deliberately destroyed. This process, too, is biased, as humans have to actively want to preserve data available for later use. The researcher can influence neither data production nor data selection process. However, both process influence what data are available at all. Researchers cannot 338 change this situation. They only can estimate biases by qualitatively analyzing production and selection contexts. They also can triangulate media data with other data. Sampling Strategies for Process-Generated Data: Very often the amount of process generated data is so large that researchers has to sample them. For sampling media data, researchers need a multi-saged sampling procedure: First the question arises, which media to choose. The next question is which issues to choose. The third question is which articles to choose from one single issue. For the third question, we suggest simply sampling all relevant articles from one issue. However, researchers face a trade-off between this and the following sampling stage: They can either analyse many issues out of few newspapers (Strategy 1) or few issues from many different newspapers (Strategy 2). In Strategy 2, the chosen issues might not be typical for the discourse in this single newspaper. In Strategy 1, the chosen newspaper might not be typical for the overall discourse. In order to analyse discourses, researchers also need to focus change over time, thus facing a third problem: If the time span is large, researchers can only sample few issues per time unit – they might miss important fluctuations in discourse. If the time span is small, researchers might miss important longterm shifts in the discourse. We tried to solve this problem by generating two separate samples. Each sample can be used to estimate sampling errors the other sample. For Sample 1, one newspaper was chosen that we presumed to be “typical” for the discourse. The time span was very short (1997 to 2000). Thus minimising sampling problems on the second selection stage, as three issues per week could be analysed. For Sample 2, we chose ten different newspaper using the most different cases design. We also extended the time span (1964 to 2000), selecting only issues from every 4th year in order to eliminate effects of election and economic cycles. For selected years, we chose issues starting from a reference day. Starting from this date we used adaptive sampling strategies to choose as many relevant issues as possible. Keywords: Process-Generated Data, Sampling Bias, Multi-Staged Sampling, Adaptive Sampling, Most Different Cases Design, Longitudinal Research, Time Units, Time Span, Validity, Information Accessibility Estimating Domestic Tourism Expenditure in Developing Economies: Lessons from India Rajesh Shukla and Pradeep Srivastava, National Council of Applied Economic Research (NCAER), Parisila Bhawan, New Delhi, India, [email protected] The paper focuses on two central albeit inter-related questions. First, how should data on domestic tourism be generated? This is difficult question even in the context of OECD economies, but more so in developing countries with different institutional infrastructure and resource constraints. Second, it also addresses the issue of credibility; in a sense raison de etre’ of the Tourism Satellite Accounts (TSA) framework. More specifically, we seek to evaluate based on the Indian experience different possible modules of an inter-institutional platform that can contribute to enhancing credibility of the TSA in a developing country context. The Indian experience in developing a TSA is distinct for at least two reasons: first, the Government has commissioned a dedicated survey of domestic tourism and, second, the responsibility has been entrusted to a non-government organization, albeit one that is respected and has high credibility. These distinctions, virtually by definition would also constitute caveats in any attempt at generalizing to a broader context. Given these caveats, this paper has used the experience of domestic tourism data generation in India to highlight issues that may confront other countries undertaking similar processes. Household budget surveys, although active in India, are undertaken only once every five years on large samples, and on thin samples in other years. Since data already existing was relatively old, scarce and not readily amenable for use in a TSA, it was decided to undertake a dedicated household survey of domestic tourism. The survey would measure both number of tourists of various types and their expenditures. 339 The sampling design adopted recognized the heterogeneity of distribution of tourists in the population and weak priors on this underlying distribution. Consequently, the survey developed a sampling frame in the last stage of its multistage stratification, with the belief that the higher costs would be more than made up by the benefits in precision of estimates and other factors. In view of seasonal nature of tourism, the survey was undertaken in two rounds during the reference period of one year. The survey methodology put considerable emphasis on the role of training, which was found extremely useful in negating preconceived (and ill informed) notions about tourism in the general population, conveying accurately to investigators and respondents concepts and definitions underlying TSA, and enhancing enthusiasm and motivation in the investigators. Keywords: Tourism Satellite Accounts, Domestic Tourist, Usual environment, Multi-stage sample design. Sampling Undocumented Immigrants in Brussels via Snowball Sampling: Theoretical and Practical Considerations Mila Paspalanova, KU Leuven, Belgium, [email protected] The proposed paper concentrates on the problematic of sampling undocumented Eastern European immigrants residing in Brussels. By focusing explicitly on two groups of undocumented immigrants e.g. the Polish and the Bulgarian immigrants we would present theoretical and practical arguments for justifying the chain referral method or the snowball sampling procedure as the only available method for sampling understudied and unknown populations. The paper proceeds by firstly presenting a broad overview of the research traditions in the field of sampling the so called hidden or hard to reach human populations. The discussion of the advantages and the limitations of this non-random sampling procedure would be presented and exemplified in the light of classical and contemporary studies (such as the ones of Biernacki & Waldorf (1981), Eland Goossensen et al. (1997), Faugier & Sargeant (1997), Sudman (1972), Welch (2001)) adopting snowball sampling as a sampling tool. The second part of the paper would focus on the specific characteristics of the undocumented Eastern European immigrants in Brussels witch make the application of random sampling techniques practically impossible. In the same line of presentation we would also touch upon the situation of sampling (or better said of non-sampling) and studying undocumented populations in Brussels until now and we would illuminate the overall neglect of the need of applying any sampling procedure in the research on undocumented immigrants. In the final section of the paper we would present the practical ways of implementing snowball sampling for identifying undocumented immigrants in Brussels and we would discuss upon the encountered methodological problems. Similarities and differences in comparison with other related studies employing snowball sampling would be drawn. To conclude we would present the initial results of the sampling of undocumented Eastern European immigrants in Brussels. Keywords: snowball sampling, hard to reach populations; undocumented immigrants; Brussels References Biernacki, P., Waldorf, D. (1981) Snowball Sampling. Problems and Techniques of Chain Referral Sampling. Sociological Methods and Research, Vol. 10, 2, 141-163.; Eland-Goossensen, M., Van de Goor, L., Vollemans, E., Hendriks, V., Garretsen, H. (1997) Snwoball Sampling Applied to Opiate Addicts Outside the Treatement System. Addiction Research, Vol. 5, 4, 317-330.; Faugier, J., Sargeant, M. (1997) Sampling Hard to reach Populations. Journal of Advanced Nursing. Vol. 26, 790-797.; Sudman, S. (1972) On Sampling of Very Rare Human Populations. Journal of the American Statistical Association, Vol. 67, 338, 335-339.; Welch, S. (2001) Sampling by Referral in a Dispersed Population. Public Opinion Quarterly, Vol. 39, 2, 237-245. 340 Using Adaptive Sampling Methods in the Analysis of Social Phenomena on the WWW Robert Ackland, Centre for Social Research, Research School of Social Sciences, The Australian National University, [email protected] Over the past 10 years, the Internet has become an important tool for communication, with individuals and organisations increasingly using e-mail and the World Wide Web (WWW) for day-to-day personal and professional interactions. While the WWW is largely seen as a new information and communication media, it is also highly effective in facilitating the formation and maintenance of networks and this has particularly been evident in area of politcal and social movements. Social scientists are turning to the Internet as a rich source of data for research on interactions and the formation of networks, but currently do not have appropriate methods for collecting and analysing Web data. Given the potentially huge quantities of data that can be collected from the WWW, one particular problem facing social scientists is: how to construct appropriate samples for further analysis? In this paper, Web data collected as part of a study into the networking behaviour of Australian political parties wll be used to illustrate the applicability of different methods of adaptive sampling for the study of social and political phenomena on the WWW. Keywords: online communities, Web mining, graph theory STREAM: LATENT VARIABLES MODELLING SESSION: LATENT GROWTH CURVE STRUCTURUAL EQUATION MODELS, AUTOREGRESSIVE AND HYBRID MODELS Organizors: Jost Reinecke1 and Peter Schmidt2 Contact: 1University of Trier, Germany, [email protected] 2 University of Giessen, Germany, [email protected] Keywords: Growth Curve Modelling, longitudinal Data, Structural Equation Modelling, Autoregressive Modelling Abstract In the last years a fastly growing literature has been dealing with the use and new applications of latent growth curves within the structural equation modeling approach. Previously, growth curve modeling was mainly discussed in developmental psychology. In sociology and political science autoregressive models have been mostly applied to model longitudinal data. In several papers, Bollen and Curran have compared the strengths and weaknesses of latent growth models and autoregressive models for modeling longitudinal data and proposed hybrid models to combine them. In this section we want to bring in papers which take up this issue and compare these three types of models. Special attention is given to the problem which type of information in growth curves and autoregressive models can be modeled and gained. Besides more fundamental papers we look for applications in sociology, political science and social psychology. These applications should also include intervention and evaluation studies. 341 Combining Autoregressive and Semiparametric Mixture Models: Investigating the Selection-Facilitation Effect of Peers on Antisocial Behaviors Eric Lacourse1, Daniel Nagin2 and Richard E. Tremblay1 1 University of Montreal, Canada, [email protected] 2 Carnegie Mellon University, USA Being part of a delinquent group has been shown to facilitate antisocial behavior. Different analytical approaches have been used to show this socialization effect. Recent work, in the field of structural equation modeling (Curran & Bollen, 2001) demonstrated that it could be useful to combine an autoregressive and a growth curve model to describe complex processes that consider both population heterogeneity and state dependence. The present study expands on this work by combining both the autoregressive and semi-parametric mixture model (Nagin, 1999) using data from the Montreal Longitudinal Experimental Study. Although these statistical methods are often used separately there are many advantages to combine both. One main advantage of this combined analytic approach is to estimate the effects of time varying covariates (e.g. delinquent peers) on distinct developmental trajectories of antisocial behavior (e.g. early onset, late onset, desisters). Furthermore, this method enables the researcher to simultaneously model the association between trajectories of distinct but related behaviors, and to estimate the common heterogeneity using a set of risk factors. Potential extensions and limitations of this modeling strategy will be discussed. Keywords: semiparametric mixture model, antisocial behavior, autoregressive model Pay Off and Impact of Varying the Intercept in Latent Growth Curve Modeling Manuel C. Voelkle, University of Mannheim, Germany, [email protected] Latent growth models have become an increasingly popular approach to model change over time. Naturally, the focus lies primarily on the latent linear or nonlinear slope factor, whereas the latent intercept is just a by-product that is often merely considered due to reasons of completeness, rather than substantial interest. In many cases, however, researchers are not just interested in predicting individual growth curves, but also in predicting between subject variance at a specific point in time. By adjusting the factor loading vector of the latent intercept, the position of the intercept can be easily altered to any – observed and even unobserved – point of measurement. It will be illustrated, how changing the factor loading vector of the latent intercept, can be used to obtain more reliable estimates of between subject variance at a specific point of time. Paying closer attention to the position of the latent intercept, may also help to understand and resolve many alleged paradoxes and phenomena in the analysis of change, such as the Law of Initial Values (Wilder, 1957, Lacey & Lacey, 1962), the Overlap Hypothesis (Anderson, 1939, Bloom, 1964), or the Fanspread Effect (Campbell & Erlebacher, 1970). Theoretical and empirical evidence from the field of skill acquisition will be provided to demonstrate the superiority of this approach over conventional regression analyses, and implications for the concepts of reliability and validity will be discussed. Keywords: Latent growth curve modeling, latent intercept, factor loading matrix, reliability, validity, skill acquisition References Anderson, J. E. (1939). The limitations of infant and preschool tests in the measurement of intelligence. Journal of Psychology, 8, 351-379. Bloom, B. S. (1964). Stability and change in human characteristics. New York: Wiley. Campbell, D. T., & Erlebacher, A. (1970). How regression artifacts in quasi-experimental evaluations can mistakenly make compensatory education look harmful. In J. Hellmuth (Ed.), Compensatory education: A national debate (Vol. 3). New York: Brunner/Mazel. Lacey, J. I., & Lacey, B. C. (1962). The law of initial value in the longitudinal study of autonomic constitution: Reproducibility of autonomic responses and response patterns over a four year interval. In W. M. Wolf (Ed.), Rythmic functions in the living system: Annals of the New York Academy of Sciences, 98, 1257-1290. Wilder, J. (1957). The law of initial value in neurology and psychiatry. Journal of Nervous and Mental Disease, 125, 73-86. 342 Latent Growth Curve and Interaction Modeling: The Case of Travel Mode Choice Fan Yang Wallentin, Peter Schmidt and Eldad Davidov In recent years, a growing number of scholars has tested interaction and non-linear effects in their models. Many of these studies tested interaction effects between perceived behavioural control (PBC) and intention in the theory of planned behaviour. However, only a small number of them applied structural equation modeling. In these few studies, the utility of latent growth modeling has remained largely unexplored. In this study, we compare Maximum Likelihood and Robustified Maximum Likelihood to estimate the effect of the interaction of the slopes of PBC and intention in the theory of planned behaviour on the slope of the changing behaviour over three points in time. We apply real data collected in Frankfurt on travel mode choice of people moving to live in the town. Anomia, Left right Orientation and Group Related Enmity. Steffen Kühnel, Peter Schmidt and Elmar Schlüter In this paper we present results of a three wave representative panel of the german adult population dealing with group related enmity in the german ppulation. We present results and compare Autoregressive models, latent growth models and hybrid models dealing with the relationships between anomia, left right orientation and three facets of group related enmity that is discrimination of foreigners, anti-Semitism and Islamophobia. This includesthe modelling of observed and Latent Means. In a summary we compare the strengths and weaknesses of the three approaches for testing hypotheses concerning change processes. Keywords: Autoregressive, latent growth and hybrid models SESSION: MULTIDIMENSIONAL SCALING AND FACET THEORY Organizors: Patrick Groenen1 and Ingwer Borg2 Contact: 1Rotterdam, The Netherlands, [email protected] 2 Mannheim, Germany, [email protected] Abstract Using multidimensional scaling (MDS) in combination with facet theory (FT) has proved to be a successful approach in many fields of the social sciences. Some prominent examples are Schwartz’s universal circumplex structure of values; the radex of work motives by Elizur and others; and the multiplex of protest behaviour by Levy and Guttman. The approach consists of partitioning an MDS representation of the empirical items by projecting the different facets, facet by facet, onto the MDS space in an attempt to partition it into regions that contain only points which are classified in the same way by the facets. Using MDS and FT is not a main stream methodology which makes it difficult to keep track with developments. Applications are scattered in many fields, and data-analytic innovations are often technical, appearing in journals not read by many social scientists. The goals of the symposium is to offer a forum for state-of-the-art reviews, both technically and theoretically-empirically, and for presentations of series of empirical research that definitely is cumulative in nature. The intention is to show how MDS and FT can be used today in systematic theory construction. 343 What is not what in Facet Theory Wijbrandt H. van Schuur, Department of Sociology, University of Groningen, The Netherlands, [email protected] The idea behind Facet Theory is to understand complex concepts in a research domain by analytically dissecting their different aspects or facets, and find out how they are interrelated, such that proper classification leads to a proper description (i.e., measurement of aspects of concepts and measurement of individual persons) of the phenomena to which the concepts refer. In this paper I mention five areas in which facet theory can be improved: 1) the formulation of the mapping sentence, 2) the indirect or holistic approach to data analysis, 3) the direct or analytical approach, 4) substantive (social) scientific theories, and 5) the organizational structure of the Facet Theory Association. Ad 1) The classification of facets (e.g., in stem, modifying, response, axial, polar or modular facets) requires further investigation. Classifications within a single facet often do not follow elementary rules of logic. Ad 2) The indirect or holistic approach to data analysis hinges on a too specific concept of similarity, and too specific approaches to the analysis of such data. Within Similarity Structure Analysis (SSA) there is too much emphasis on the visualization of the first dimensions for all items. The zero-point on polar and modular facets are often not well understood. Alternatives to SSA (e.g., additive tree models, correspondence analysis) are not sufficiently pursued. Ad 3) The direct or analytical approach profits insufficiently from recent advances in probabilistic item response theory models. This does not only include probabilistic versions of the Guttman scale (e.g., the Rasch model or the Mokken model), but also IRT models with non-cumulative item response functions (e.g., latent class analysis, unfolding, or the circumplex model). Ad 4) An interpreted picture is not the end stage of research, but only a first step. There is too little cumulativity in understanding the theoretical meaning of the facets, and in their application to other types of study. Ad 5). The Facet Theory Association (FTA) should be an organization that adheres to specific scientific principles, and not to any specific human being, however brilliant. It must be willing, if necessary, to cut the umbilical cord with its founder. References Borg, I. & Shye, S. (1995). Facet theory. Form and content. Thousand Oaks: Sage. Advanced Quantitative Techniques in the Social Sciences, 5. Ley, S. & Elizur, D. (2003). Facet theory. Towards cumulative social science. Ljubljana: University of Ljubljana, Center for educational development Questionnaire Design and Data Analysis using Facet Approach: The Case of the International Census on Attitudes toward Languages Kazufumi Manabe, School [email protected] of Sociology, Kwansei Gakuin University, Japan, The purpose of this paper is to illustrate the utility of Facet Approach developed by L. Guttman and his group by using the example of the questionnaire design and data analysis of the International Census on Attitudes toward Languages. This census was conducted in 28 countries/areas of the world from 1997 to 1998 by our research team at the National Language Research Institute in Japan. Facet Approach is composed of three parts: Facet Design. This involved the construction (Structuples) and use of Mapping Sentences for questionnaire design. Facet Analysis. This includes Scalogram Analysis, Partial Order Scalogram Analysis, Smallest Space Analysis, Median Regression Analysis, and so on. Facet Theory. The First Law, The Second Law, and the Law of Polytone Regressions have been proposed and tested in diverse empirical contexts. 344 I attempt to summarize the basic ideas of Facet Approach by focusing on the above three areas, which I believe are useful for the systematic design of a questionnaire survey. Facet Design can be used to analyze questionnaires to determine the completeness of the information gathered. Some examples of such analyses are shown. Facet Theory guides the data analysis in that it identifies meaningful and interpretable relations which can be explored statistically. In this paper I demonstrate three different kinds of data analysis (Facet Analysis) guided by Facet Theory. Keywords: facet approach, facet design, facet analysis, facet theory, international census, languages New Type of Psychodiagnostical Test: Situational Operational Grid Anna S. Naumenko, [email protected] Nowadays there’s a strong need in creating precise psychodiagnostical instruments. In Moscow State University laboratory for computer psychodiagnostics we’re trying to invent and explore a new type of psychodiagnostical test, which combines the approaches of multidimensional scaling and repertory grid technique, proposed by G.A. Kelly. This is an attempt to find a compromise between ideographical and nomothetical techniques, between gaining qualitative non-standardized and statistical standardized information. The author of the new approach, professor Alexander G. Shmelyov has given it a name of situational operational grid. The task of the proband is to choose out of the proposed list one operation/behavioral type for each situation or to rank the degree of an acceptability of proposed operations/behavioral types for each situation. As a result the researcher gets estimations’ matrix (operations x situations). This test type is mostly used for evaluating the professional competence of different professions representatives. The situations are very specific for the concrete profession; that means: they vividly describe typical daily work occasions of this or that professional. Sometimes suitable pictures or portraits of the characters are used to enrich the situations descriptions and to make them more plausible. That demands careful operational analysis of the professional activity; that is usually made beforehand. To present time in our lab were constructed 3 situational operational grids suited for TV journalists’, realtors’ and top-managers’ competence assessment. All tests were validated using the results of expert professionals and novices diagnostics. In every professional group was shown significant difference between experts and novices. Keywords: multidimensional scaling, repertory grid technique, situational operational grid, professional competence estimation Explaining the Structure of Values of an Ideal Organization by Schwartz‘s Universal Value Theory Ingwer Borg1, Patrick J.F. Groenen2, Wolfgang Bilsky3 and Karen A. Jehn4 1 ZUMA, Germany, [email protected] 2 Erasmus University Rotterdam, The Netherlands, [email protected] 3 U Münster, Germany 4 Universiteit Leiden, Netherlands Shared values are arguably the most fundamental component of culture. To measure personorganization fit, O’Reilly et al. (1991) have developed an instrument, called the OCP or the Organizational Culture Profile, that is based on value judgments. It uses 54 value items such as being innovative, being people-oriented, risk taking and working long hours. The individual is asked to Qsort these items into nine ordered categories by (a) his or her personal view of the desirability of these values in an ideal organization, and (b) in terms of how well these values characterize a particular organization. The resulting two distributions are the basis for computing a person-organization fit 345 measure that is used to explain such dependent variables as commitment, satisfaction, performance, and conflicts. So far, there is little research on the structure of the OCP items. O’Reilly et al. (1991) report the results of an exploratory factor analysis. It shows eight factors such as innovation, risk taking or attention to detail. The data of Jehn et al. (1997) essentially replicate this factor structure. Yet, one wonders on how one should interpret differences in the number or the composition of factors, since there is no guiding theory. Bilsky & Jehn (2002) proposed to use Schwartz’s universal theory of values (Schwartz & Sagiv, 1995) for that purpose. They classified the OCP items into the typology of this theory, and then checked whether the item intercorrelations could be represented in an MDS plane such that the respective points would fall into regions reflecting this typology. The results roughly confirmed this conjecture. We here show that using a confirmatory MDS approach (Borg & Groenen, 1997), the OCP value items perfectly fit into the Schwartz theory with very little additional stress compared to the exploratory MDS solution used by Bilsky & Jehn (2002). We enforce a four-quadrant solution that can be interpreted to reflect two value dilemmas: the individual must decide to either favor an emphasis on results or on harmony on the one hand, and similarly decide on whether he or she prefers risk-taking or rather sticking to rules. Most of the factors of a standard exploratory factor analysis of the OCP items can, moreover, be explained by the two-dimensional MDS solution. Thus, there exists a good theoretical basis for the OCP items that, moreover, suggests much simpler organizational culture assessments in practice. References Borg, I. & Groenen, P. (1997). Modern Multidimensional Scaling. New York: Springer. Bilsky, W. & Jehn, K.A. (2002). Organisationskultur und individuelle Werte: Belege für eine gemeinsame Struktur. In M. Myrtek (ed.), Die Person im biologischen und sozialen Kontext (pp. 211-228). Göttingen: Hogrefe. Jehn, K.A., Chadwick, C., & Thatcher, S.M.B. (1997). To agree of not to agree: the effects of value congruence, individual demographic dissimilarity, and conflict in workgroup outcomes. The International Journal of Conflict Management, 8, 287305. O’Reilly, C.A. III, Chatman, J., & Caldwell, D.F. (1991). People and organizational culture: a profile comparison approach to assessing person-organization fit. Academy of Management Journal, 34, 487-516. Schwartz, S. (1992). Universals in the content and structure of values: theoretical advances and empirical tests in countries. In M. Zanna (ed.), Advances in Experimental Social Psychology, 25, 1-65. New York: Academic Press. Schwartz, S. & Sagiv, L. (1995). Identifying culture-specifics in the content and structure of values. Journal of CrossCultural Psychology, 26, 92-116. SESSION: CORRESPONDENCE ANALYSIS Organizer: Jörg Blasius Contact: University of Bonn, Seminar of Sociology, Adenauerallee 98a, 53113 Bonn, Germany, [email protected] Keywords: Simple correspondence analysis, multiple correspondence analysis Abstract In recent years a new technique has become more and more attractive in social science statistics: correspondence analysis. In the last decade, the number of theoretical publications as well as published empirical applications of correspondence analysis have increased enormously. However, there is still a large gap between the theory and the practice of the method. The aim of the session is to bring together theoretical and applied researchers in all the areas of social sciences where correspondence analysis is currently used, notably sociology, psychology, education, and political sciences. Contributions which combines theoretical aspects and social science applications are particularly welcome. Themes of the conference include all forms of correspondence analysis and related techniques: Simple, multiple, joint and multiway correspondence analysis, homogeneity analysis and biplots. 346 Nordic Business Researchers’ Use of Research Methodology: An Assessment of Researchers in Denmark, Finland, Norway and Sweden Tor Korneliussen, Bodø Graduate School of Business, 8049 Bodø, Norway, [email protected] This paper investigates the following research questions: Do business researchers in the Nordic countries use different research strategies and how are these research strategies related to each other? At the 15th Nordic Conference on Business Studies in 1999, 352 Nordic participants were provided with questionnaires, which 178 of them returned. The respondents were asked to explain to what extent they generally use twelve research methodologies. We used Likert-type items with five categories each, ranging from “strongly disagree” to “strongly agree”. Applying nonlinear principal component analysis, we choose a two-dimensional space the first of which is described as “qualitative vs. quantitative methodology”. This dimension seems to contrast use of “qualitative methodology” versus use of quantitative/experimental methodology. The second dimension seems to contrast development of formal theory versus empirical research. With respect to dimension 1, we can see that Norway reported the highest degree of use of quantitative methodology. Although both Sweden and Finland report high use of qualitative methodology, Denmark has the highest use of qualitative methodology. Business researchers in Norway seem to be quite quantitative-oriented, while researchers in Sweden, Finland and Denmark are more qualitativeoriented. This finding is statistical significant at the .10 level. Dimension two mirrors normative versus descriptive research methodologies. There is no statistically significant difference between the researchers of the Nordic countries with respect to this dimension. To visualize the structure of responses in the two-dimensional space, we use biplots of the indicators. The reported findings show that business researchers in the Scandinavian countries tend to use different research methodologies. Norwegian researchers reported the highest use of quantitative methodology. Although both Sweden and Finland report high use of qualitative methodology, Denmark has the highest use of qualitative methodology. Business researchers in Norway seem to be relatively quantitatively oriented, while researchers in Sweden, Finland and Denmark are more qualitatively oriented. Keywords: research methodology, non-linear principal component analysis, biplot Using Combined Methods to Develop Sociological Concepts. An Exploration of the Concept of Social Capital using Correspondence Analysis and Case Studies Research Francesca Odella, Dept. Sociology and Social Research, University of Trento, Italy, [email protected] The paper focuses on the methodological implications of combined methods, and specifically the application of correspondence analysis to qualitative research. Using material from an empirical research on social capital at local level, the paper illustrates how the joint use of qualitative and quantitative techniques facilitates exploration of specific dimensions of sociological concepts. In particular, the paper will describe the use of data collected via semi-structured interviews in correspondence analysis. The aim is to develop multidimensional perspectives from data collected thought case studies and to catch relational and cultural aspects of the concept of social capital. On the basis of such analysis, the paper will then discuss the advantages and the constraints of the transformation and coding of qualitative data. Keywords: correspondence analysis, case studies, social capital, combined techniques 347 The Analysis of Lifestyle Groups and Social Milieus in Germany Using Cluster and Correspondence Analysis Ingvill C. Mochmann1 and Yasmin El-Menouar2 1 Central Archive for Empirical Social Research, Cologne, Germany, [email protected] 2 Seminar of Sociology, University of Bonn, Germany, [email protected] In previous decades the analyses of lifestyle groups and social milieus have been quite popular in all social science disciplines, in particular in Germany. Cluster analysis and correspondence analysis are applied in order to find homogeneous groups in areas as leisure, consumption and voting behavior. Usually, no differentiation is made between lifestyle groups and social milieus, although the concepts differ with respect to the stability and tenacity of the groups found. In this paper the existence of lifestyle groups and social milieus in Germany by means of cluster and correspondence analysis. In the first step, lifestyle groups in Germany will be analyzed using cluster analysis. In the second step, bivariate analyses between lifestyle groups and several external variables in the areas of social structural characteristics, organizational membership, cleavages, interest representation and party preference will be carried out. Finally, a correspondence analysis including the lifestyle groups and external variables will be carried out with the aim to identify political social milieus in the electorate of eastern and western Germany. The aim is to show how correspondence analysis can be used to differentiate between lifestyle groups and social milieus as well as its use for the identification of social milieus in modern society. Keywords: cluster analysis, correspondence analysis, lifestyle groups, social milieus, electoral behaviour International Comparison with Correspondence Analysis. Attitudes towards Same Sex Partnership as an Empirical Example Maria Rohlinger, Central Archive for Empirical Social Research, University of Cologne, Germany, [email protected] In the mid nineties, various European countries discussed legislative initiatives towards liberalisation of same sex partnership. The topic tentatively evokes large contrasts of pro’s and con’s within the given societies. Primary investigators of the group of researchers, setting up the frame of International Social Science Programme (ISSP) choose to integrate the question concerning the acceptance of same sex partnerships into their questionnaires in 1994, thus allowing to portrait the background of the political discussion in the context of a wider scale of values and social predispositions. Comparisons of the nature of integrating simultaneously sets of multiple indicators parallel to social background, conceptualised within an international frame, are demanding for the data reduction techniques to be applied. Correspondence Analysis allows in the given example for a visualization of different attitudes in most of these aspects. Keywords: correspondence analysis, international comparison, same sex partnership The Use of Correspondence Analysis for Stratification Renato Salvatore, Department [email protected] Economia e Territorio, University of Cassino, Italy, In many social research surveys, large available sets (frames) of multivariate data, such as censuses, must be exploited to achieve the best cognition in the interactions between categorical variables. The reason is that some aspects of phenomena that we utilize to describe the underlying well-known connections between the main research variables, can lead us to a better specification of a subsequent 348 survey plan, e.g. a stratified sampling plan. One of the main objectives of a stratified sampling is to reduce the variance of the mean estimation for survey variables, splitting population in homogeneous strata. This causes a consequent reduction of the survey costs, due to the high level of homogeneity in the strata, where strata are represented by categories of descriptive variables that are related with those of the survey. The desired gain in homogeneity of the population strata can be achieved through complementary use of correspondence analysis and cluster analysis, driving to meaningful clusters. They can be considered the strata that solve, in some way, the homogeneity maximization problem. But population strata must be considered for another purpose: they could be domains of interest, in which we can evaluate the survey variables mean estimates. Clusters cannot be used for this objective, and we must inevitably exploit categories of descriptive variables. In this paper, the multiple correspondence analysis and cluster analysis techniques are used in order to detect the best set of categorical variables that satisfy the demand of an increased homogeneity of population strata. A ranking of categorical supplementary variables is listed, based on an aggregate measure in the factorial desired space of the distance between centres of gravity of categories and cluster centroids. An application is analysed in which the objective is to reduce the general costs of an agri-environmental impact assessment sample survey. The data are those from Italian 2000 Agricultural Census, the sampling technique is the optimum multivariate allocation based on the convex programming approach. Keywords: multiple correspondence analysis, stratified sampling, optimum allocation Measure versus Variable Duality in Correspondence Analysis Henry Rouanet and Brigitte Le Roux, Université René Descartes, Paris, France, [email protected], [email protected] The formal approach used by Benzécri to develop CA was not an accidental matter of notation, but an integral part of the construction (Benzécri & coll, 1973). The properties of CA are entirely founded on the underlying mathematical theory, essentially abstract linear algebra , as found in the classical mathematical books by MacLane, Halmos, etc.: finite-dimensional vector space, homomorphism, scalar product, etc. The cornerstone of the formal-geometric approach is the measure vs. variable duality, which formalizes the distinction between two sorts of quantities: those for which grouping units entails summing (adding up) values, such as weights, frequencies, etc., we call them measures (like in mathematical measure theory), versus those for which grouping units entail averaging values, such as scores, rates, etc., we call them variables. This duality is reflected in the duality notation (alias transition notation), putting lower indices for measures and upper indices for variables. See Rouanet & Le Roux (1993), Le Roux & Rouanet (2004). In the paper, we describe measure vs. variable duality in CA at the following two crucial stages of geometric modelling: i) Construction of clouds and the chi-square metric. The marginal frequencies of the table firstly provide reference measures over rows and columns. Secondly, they define Euclidean isomorphism from variable vector spaces to dual measure vector spaces, hence scalar products and Euclidean norms, therefore they determine without arbitrariness the chi-square metric over those spaces. ii) Principal directions of clouds and principal coordinates. The fundamental mathematical result is that the solution of spectral equations is the singular decomposition of two adjoint homomorphism and/or the associated bilinear form. Applying these results to CA immediately yields the transition equations and the reconstitution formulas. CA is a case in point to exemplify the superiority of the formal approach to multivariate statistics over the usual matrix approach. Keywords: correspondence analysis, geometric data analysis, formal approach, linear algebra References Benzécri J.P. & coll.(1973). Analyse des Données, Volume 2, Analyse des Correspondances. Paris, Dunod. Rouanet H. & Le Roux B. (1993). Analyse des Données Multidimensionnelles Paris, Dunod. Le Roux B., Rouanet H. Geometric Data Analysis (in press). Dordrecht, Kluwer. 349 The Space of Elite Opinions. The Case of Norway Johs. Hjellbrekke and Olav Korsnes, Department of sociology, University of Bergen, Norway, [email protected], [email protected] Elite studies typically focus on how to separate between the various types of elites. Do the elites constitute a homogenous group or not? The criteria for internal differentiation, for instance between political, administrative and economic elites, have usually been related to societal sectors, specific societal task, levels of power or to functional criteria, In this paper, we follow a different line of investigation, and identify elite configurations on the basis of their space of position taking, i.e. the agents’ attitudes and standpoints towards a wide range of political, cultural and social issues. Drawing inspiration from Bourdieu’s (1989) and Bourdieu & Saint-Martin’s (1978) analyses of the French field of power, and following the approach first outlined in Le Roux & Rouanet (1998, 2004) and Chiche et al. (2000), we will demonstrate the advantages of multiple correspondence analysis in an examination of position specific homogeneity and heterogeneity, and also of cross-cutting oppositions, in the Norwegian elites’ opinions. We will follow a two-step strategy, where we first identify the main polarizing opinion dimensions in the space of position taking. This will be done through an analysis of the respondents’ answers to 40 statements regarding ecology, economics, equal rights, feminism, neo-liberalism, nationalism, religious issues and welfare state issues. Secondly, through a detailed investigation of the cloud of individuals, and by using new tools in geometric data analysis, we will first examine how these dimensions affect the various groups internally. Thereafter, particular emphasis will be put on the degree of group separation and group overlapping. In this way, not only the elite configurations but also their intersections will be revealed. The data originate from a survey by the Norwegian Power and Democracy Project, distributed to 1711 people in top positions within the Norwegian society; positions within politics, academia, larger private and public companies, the central administration, the church, the judicial system, the military, cultural institutions and larger managerial and occupational organisations. Keywords: multiple correspondence analysis, geometric data analysis, elites, Bourdieu References Bourdieu, Pierre (1989) La noblesse d’état. Paris: Ed. de Minuit Bourdieu, Pierre and de Saint-Martin, Monique (1978). "Le patronat". In Actes de la recherche en sciences sociales, nr. 20/21, mars-avril 1978, pp. 3-83. Chiche, J., Le Roux, B., Perrineau, P. and Rouanet, H. (2000). "L'éspace politique des électeurs francais à la fin des années 1990". In Revue francaise de science politique, vol. 50,3, juin 2000, pp. 463-488. Le Roux, Brigitte and Rouanet, Henry (1998). "Interpreting Axes in MCA: Method of the Contributions of Points and Deviations". In Blasius, Jörg and Greenacre, Michael J. (eds.). Visualization of Categorical Data. London: Academic Press, pp.197-220. Le Roux, Brigitte and Rouanet, Henry (2004). Geometric Data Analysis Amsterdam: Kluwer Academic Publishers Processes of Social Differentiation in an Economic “Social Field” Lennart Rosenlund and Odd Einar Olsen, Høgskolen i Stavanger, Aalborg Universitet, Norway [email protected] This paper includes an effort to apply Pierre Bourdieu’s analytic framework to a social universe to which it, according to our knowledge, never has been systematically applied; that of organisational studies. It has been inspired by a remark by him in his late writing, when he, after a long period of apparent lack of interest turned his analytic eye towards the economy (Bourdieu 1997, Bourdieu 2000). He suggests that the enterprise as a social complex social entity is well suited to be approached analytically by the help of one of his key concepts; the enterprise as a social field. What will be examined are analyses based on data collected from what appears – at the outset - to be a highly homogenous social group; the data analysed consists of answers to a questionnaire distributed among managers and specialists employed by two major industrial actors and some of their subsidiaries within the Norwegian oil industry. 350 The social reality this concept aims to account for consists of a separable entity of the “global” social structure (the space of social positions as it is outlined in the Distinction (Bourdieu, 1984) that is presumed functioning according to a (semi-) autonomous logic of its own. In this, the social agents hold specific positions and are engaged in struggles over values that are of importance to all (various forms of capital). The concept has double analytical edges corresponding to the “two orders” of social reality. There is, on the one hand, an “objective” invisible system of social positions in which each and all of the social agents belonging to the field have a specific position, which tend to have bearings on the way they form and perform; their practices, via their habitus(es). Seen in this way, a social field is an objectivation or representation of the main forces of social differentiation and thereby an image of the prevailing power relations in the field. On the other hand, there are – precisely - social practices, or human symbolic activities - which are products of their habitus(es) - that unfold in perpetual struggles and competition that the social agents are engaged in, in the field. The concept of social field aims at giving the best possible representation of both aspects of the social reality, both the “structural” aspect and the symbolic one, the latter of which may be viewed as a system of position-takings, or stances. The paper will examine the structure of this particular social field seen as a system of social positions. The analysis has been undertaken by the help of GDA (Geometric Data Analysis) or multiple correspondence analysis on the basis of profiles of various forms of capital among the analysed respondents. We are going to consider a three-dimensional representation of the constructed social field as an optimal representation of the prevailing system of power relations within this social universe. Further, the analyses consider secondary characteristics of the field such as relations to division of labour, prevailing hierarchies of knowledge, demography, system of classification of jobs and departments, trajectories within the field (previous jobs) and affiliation to the local community. Lastly, we shall consider the constructed social field as a system of position-takings, i.e. how positions in the ”objective” social field are reflected in oppositions and divisions with regard to how the social agents perceive and value different aspects of their work. This last section is based on an excerpt of a battery of questions developed by one of the commercial opinion poolers in Norway (MMI) to measure “working climate”. Keywords: multiple correspondence analysis, geometric data analysis, social fields, organisational studies, Bourdieu References Bourdieu, Pierre (1984) Distinction – A Social Critique of the Judgement of Taste. London: Routledge & Kegan Paul Bourdieu, Pierre (1997) Le champ économique, Actes de la recherché en sciences socials, 199, Paris, Septembre 1997 Bourdieu, Pierre (2000) Les structures socials de l’economies. Paris: Seuil. Le Roux, Brigitte & Rouanet, Henry (2004). Geometric Data Analysis Amsterdam: Kluwer Academic Publishers Specific MCA with Special Reference to Questionnaires Brigitte Le Roux1 and Jean Chiche2 1 Université René Descartes, Paris, France, [email protected] 2 CEVIPOF, Paris, [email protected] The motivation for devising Specific Multiple Correspondence Analysis (Specific MCA) was the need to analyse questionnaires with non-responses, getting rid of the constraints of complete disjunctive coding while preserving the structural properties of MCA. In the paper, the method is presented within the general framework of Geometric Data Analysis, then we deal with the special case of questionnaires, following the line of the paper by Le Roux (1999) and the book by Le Roux & Rouanet (2004). We first recall the properties of principal directions and variables of a Euclidean cloud. Then we apply these properties to a protocol for which individual profiles are measures, and we derive the properties of bi-weighted Principal Component Analysis (PCA). Then, we recall the formulas of MCA viewed as a particular bi-weighted PCA. 351 We then compare Specific MCA to standard MCA, writing inequalities between eigenvalues and studying the rotation of principal subspaces when one goes from the global analysis to the specific one. Finally, we outline two applications of specific MCA: Bourdieu (1999), on the publisher space in France, and Chiche et al. (2002) on the political space of French electors. Keywords: multiple correspondence analysis, geometric data analysis, questionnaires. References Bourdieu P. (2000). Une révolution conservatrice dans l’édition, Actes de la recherche en Sciences Sociales, 127. Le Roux B. (1999). Analyse spécifique d’un nuage euclidien, Mathématiques, Informatique et Sciences Humaines, 37, 65-83. Chiche J., Le Roux B., Perrineau P., Rouanet H. (2002) . L’espace politique des électeurs français à la fion des années 1990. Revue française de Science politique, 50, 463-487. Le Roux B., Rouanet H. Geometric Data Analysis (in press). Dordrecht, Kluwer. The Space of Central Bankers in the World Frédéric Lebaron, University of Picardie, France, [email protected] The space of the educational, academic and professional trajectories of central bank governors, alternate governors, and monetary policy board members proves the coexistence and rivalry between different types of "symbolic capital". There are for example "insiders", who come from the central bank institution on the one hand, and "outsiders", whose legitimacy may be academic, economic or political, on the other hand. There are leaders, who come from the financial world and the private sector on the one hand, and then leaders who come from the political or university areas. Central banks are the places where these different forms of symbolic capital affront each other and combine. Multiple correspondence analysis helps to reveal the structure of this very particular social space inside the economic world, and to answer specific comparative questions related to the social characteristics of central bank leaders from different regions. This paper will present the utility of MCA to treat these particular sociological problems. Visualizing Complex Change Ken Reed, Deakin University, Melbourne, Australia, [email protected] Greenacre (1984) demonstrates how correspondence analysis can be used to visualize change over time. He illustrates this with a correspondence analysis of a table cross-classifying disciplines and time periods to tabulate change over time in the numbers of doctorates awarded in each discipline. This paper presents a simple method of displaying change when multiple variables are used to characterize objects at multiple points in time. The analysis itself it quite straightforward: correspondence analysis is applied to a matrix constructed by stacking the data table for each time point. In the example used in the paper, data are collected on the characteristics of staff in all Australian universities for three time periods (1993, 1996 and 2000). These tables are stacked so that each university appears three times in the matrix (once for each year). The drawback in this technique is that it increases the complexity of the resulting display because each object is represented by as many points as there are time periods (three, in this case). Part of the appeal of correspondence analysis is its ability to provide simple displays of complex data structures. The added complexity of increasing the number of data points can be managed by displaying objects as lines that represent their trajectory over time through a multidimensional space. Keywords: correspondence analysis; visualization; analyzing change. 352 The (Missing) Importance of Dimension One in Factor Analytic Approaches Jörg Blasius1 and Victor Thiessen2 1 University of Bonn, Germany, [email protected] 2 Dalhousie University, Halifax, Canada, [email protected] In the social sciences, factor analytic approaches are used to detect the underlying dimensions of the domain of interest. For example, one may wish to examine the dimensions of attitudes towards work. With ordinally-scaled items, one usually applies principal component analysis (PCA) followed by varimax rotation. A main problem in this approach is the importance attributed to each of the dimensions. Several well-known techniques, such as the scree test and the eigenvalue criteria can be applied to solve this problem. These methods have in common that the first k eigenvalues are used for interpretation. The eigenvalues are in descending order: the first eigenvalue is always treated as most important, the second as second-most important and so on, since they sequentially extract diminishing amounts of variation in the data. However, data from survey research always contains response errors, response styles, and other forms of non-substantive variation. Furthermore, the first dimension might reflect primarily the degree of agreement / disagreement towards the object of interest – and this might be caused to a considerable extent by the response behaviour of the respondents. In this case, dimension one (or dimension two) has a very simple interpretation; in the case of job satisfaction on a metric scale it differentiates only between those who are satisfied in general and those who are not. Applying PCA to this kind of data, the first dimension might reflect this general agreement; varimax rotation improves the visibility of the item structure whereby the interpretation of the first dimension usually changes, i.e. the interpretation as response style or as agreement/ disagreement disappears. Considering the ordinality of the items, non-linear principal components analysis (NLPCA) is the more appropriate method. Within an iterative process the category values (e.g. 1,2,3,4,…) are replaced by optimal scores and at the end of the process the PCA-algorithm can be used to perform NLPCA. The structure of this solution is already optimised and cannot be improved via varimax rotation. Using several examples, we will demonstrate and compare both methods for the interpretation of the first k dimensions. Keywords: principal components analysis, non-linear principal components analysis, response structure, response style A Measure of Asymmetries in Brand-Attribute Dominance Relationships: Correspondence Analysis of Matched Matrices Anna Torres-Lacomba, Universidad Carlos III de Madrid, Spain, [email protected] Brand image is measured to manage brand-equity with implications among the entire marketing program, which should be coordinated to create congruent and strong brand associations. Relevant literature related with to brand-attribute associations has been characterized to be “incomplete”, just measuring brand-by-brand tables, (with respect to a set of attributes) or, if complete, (measuring brand-by-brand tables and attribute-by-attribute tables for a set of brands), to be quite conceptual (Farquhar & Herr, 1990). For example, when measuring advertising effectiveness we need to assess effects on both, brand and attribute dominance (i.e., measuring new, positively evaluated associates being evoked by the brand after the introduction of a new advertisement (attribute dominance) and measuring new cues that are likely to evoke the brand (brand dominance)) (Holden & Lutz, 1992). We present correspondence analysis (Greenacre, 1984) as a visual tool that facilitates simultaneous consideration of both analyses due to a particular way of coding data named CA of matched matrices (Greenacre, 2000). Besides offering a map that enables us to visualize strength and uniqueness in brand associations, it captures asymmetries in strength between brand-by-brand and attribute-by- 353 attribute analyses, with the implication that both analyses can be determinant to take managerial decisions. As an illustration, we expose a study of brand associations for a set of brands of deodorants. Keywords: brand-attribute dominance, asymmetry in strength, correspondence analysis. References Farquhar, P. H, Herr, P.M., & Fazio, R. H. (1990). “A relational model for category extensions of brands”, Advances in Consumer Research 17, 856-860. Greenacre, M.J. (1984). Theory and Applications of Correspondence Analysis. Academic Press, London. Greenacre, M.J. (2000). “Analysis of matched matrices”, Working Paper at Universitat Pompeu Fabra 539. Holden, S.J.S.& Lutz, R.J. (1992). “Ask not what the brand can evoke; ask what can evoke the brand?”, Advances in Consumer Research, 19, 101-107. Applying Correspondence Analysis and Logistic Regression in a Survey on Self Healing Methods Zerrin A∏an, Levent Terlemez and Sevil Şentürk, Anadolu University,Turkey, [email protected], [email protected] and [email protected] One of the oldest instincts which the human being has is protection, especially against diseases. Protection is one of the basic parts of self healing process. Self healing is an extensive process which comprises many methods ranging from non medical mixtures or applications to home made medicines. Self healing process can be defined as acting to cure own medical problems without any professional support. The method – having a present medicine that is in the refrigerator or suggested buy a friend used in self healing process is not chosen randomly. There are many sociological variables which affect the self healing process and they are worth to be the subject of a survey since the results of such a survey can be used to guide people who face medical problems. In order to determine such methods used for public and personal health, a public survey was carried out on 750 individuals (age 15-56) who live in the province of Eskisehir in Turkey. The priority of the alternatives, for example, visiting a pharmacy, applying a particular method or visiting a doctor was inquired. Logistic regression is a method which is useful to explain the relationship among response variables and independent variables, where response variables are observed in two, three or more categories. In this study, we have used logistic regression and a complementary study, correspondence analysis, to show the relationship graphically. The relationship among the behaviours of individuals and their demographic characteristics were modelled by using logistic regression, and correspondence analysis is applied as an complementary study. In application, first of all, the effect of age, sex, marital status, education level, occupation and income level as well as present health condition, on appreciating self health, is explained by a model. As a result of that model, it can be said that the effect of age, occupation and present health condition is reasonable. After analyzing that model, the relationship between categorical variables (age, sex, occupation, preferred precautions, and worth of personal health ) is shown graphically by multicorrespondence analysis. Keywords: self healing, logistic regression, correspondence analysis Comparing Nonlinear Principal Component Analysis with Factor Analysis and Ordered Logistic Regression with an Application to Country Data Dicle Taspinar and Ozlem Deniz, Istanbul Commerce University, Turkey, [email protected], [email protected] Ordered categorical dependent variables can be related with both quantitative and qualitative variables by using an ordered logistic regression model. But this model estimates the probability of the observed units being in the declared categories, not the relation between them. The first known studies on this 354 subject were prepared by McCullagh (1980). In ordered logistic regression analysis there has to be no multicollinearity as in other regression analysis. The multicollinearity problem can be solved by applying factor analysis to the independent variables that converts them into independent factors. Nonlinear principle component analysis, which is an optimal scaling method, can be an alternative twofactor analysis where the independent variables are categorical. In this study the countries ranked by the UNDP’s Human Development Index have been used and the effects of factor analysis and nonlinear principle component analysis were compared in order to solve the multicollinearity problem in ordered logistic regression. In conclusion, a model stating the development statuses of the countries was defined. Keywords: non-linear principle component analysis, ordered logistic regression, optimal scaling Power Analysis and Sample Size Determination in the Context of Correspondence Analysis George Menexes and Iannis Papadimitriou, Department of Applied Informatics, University of Macedonia, Thessaloniki, Greece, [email protected], [email protected] The Correspondence Analysis is mainly considered as a descriptive technique designed to analyse simple two-way and multi-way tables containing some kind of correspondence or association between the rows and the columns. The nature of the resulting information is similar to the one produced by the Factor Analysis techniques. The method facilitates the exploration of the categorical variables’ structure included in the table. The most common type of that kind of tables is the two-way frequency cross-tabulation table. In the case of two variables where the data have been collected using the simple random sampling scheme, the statistical significance of the total inertia of the table can be tested by means of the X2 distribution. If the total inertia is considered an index of a global effect size, then the observed power of the X2 test of independence can be estimated. This test is recommended and applied in order to evaluate the rejection of the null hypothesis (no association between the two categorical variables). Generally, the power γ of a statistical test is defined as the probability of detecting a statistical significant difference or a relationship between the measured variables, if it actually exists within the population. The power is equal to 1-β (β is the Type II error rate) and it is related to α (Type I error rate), the effect size, the sample size n and the sample variance. The relationship form is such that power increases when the effect size, α and n increase. On the other hand, it decreases when the variability of data increases. In the context of the Statistical Power Analysis proposed by Cohen, the minimum required sample size can be estimated, a priori, in such a way that a predetermined effect size can be detected as statistically significant by the X2 test, at significance level α with power γ. Initially, this study examines the mathematical relation between the total inertia and the effect size proposed by Cohen for two cross-tabulated categorical variables. After that, the relation between total inertia and other association measures are presented. At this stage, a methodology for a priori and post hoc Power Analysis of the X2 test is proposed using the square root of total inertia as a measure of effect size. Using the proposed methodology, the minimum required sample size can be estimated in case of a survey or of an experimental study in which the Correspondence Analysis method will be applied. Finally, a generalization of the method is attempted in the multivariate case. The MS-EXCEL spreadsheet (enhanced with the π-face add-in) was used to carry out all the necessary computations. Keywords: correspondence analysis, chi-square test, sample size, power analysis 355 SESSION: CLASSIFICATION AND VISUALISATION Organizor: Simona Balbi Contact: Dipartimento di Matematica e Statistica, Università "Federico II" di Napoli, Italy, [email protected] Abstract Statistics has always pursued the goals of classification and visualisation as two of its main objectives. Nowadays, this assertion is more effective, because researchers usually deal with huge masses of data to be studied and to be interpreted. In the frame of Multidimensional Data Analysis, the aim of synthesising the structure of relations in the data has always been faced by giving special attention to the possibility of providing a visual perception of information, related to the identification of typical behaviours within the phenomenal variability. In the field of social sciences important results have been obtained by using this quantitative approach. However, a session with this title has not been thought only for paying a tribute to tradition, with the aim of presenting (innovative) solutions to classical problems, or interesting applications of standard techniques. The main idea consists in focusing attention on problems related with knowledge extraction from huge data sets, and on the possibility offered by new technologies and computational tools. Papers dealing with complex situations, e.g. unstructured data bases, like the one composed by documents, are also solicited Cluster Based Conjoint Analysis Carlo Natale Lauro1, Germana Scepi1, Giuseppe Giordano2 1 Dipartimento di Matematica e Statistica, Università degli Studi di Napoli “Federico II”, Italy, [email protected] 2 Dipartimento di Scienze Economiche e Statistiche , Università degli Studi di Salerno, Italy From a Customer Relation Management point of view, business strategies must not be oriented only to retain existing customers, but also to widen the customer base. Information about the customer is the key to success. Different marketing strategies can be followed for creating a good relationship with customers and for satisfying their needs. A specialisation strategy consists in designing different products/services according to consumer preferences. Therefore, homogenous segments of consumers are defined. Each segment is a class of consumers with similar preferences and the segments are as much as possible different each other. Specialised products are so offered in order to maximise Customer Satisfaction in each class. Clustering methods are generally applied for defining a suitable number of homogenous classes according to some statistical criteria. In particular, a-priori information is often used in order to characterise the typology of consumer. Differently, a market covering strategy is followed looking for an ideal product, which reflects the preferences of the most part of consumers. Factorial techniques allow obtaining a suitable synthesis of different preference structures and representing them on a common subspace defined in order to minimise the information loss. In this paper, we propose a strategy based both on a clustering method and a factorial approach to Conjoint Analysis called Factorial Conjoint Analysis (FCA) proposed by Lauro, Giordano, Verde (1998). Our strategy aims at defining market segments according to the preferences expressed by consumers and consists in successive and iterative steps of factorial syntheses and classification of the preferences. This approach allows to get, at the same time, the identification of an optimal preference model for each market segment and a common model as a synthesis of the individual preferences. We shall show, also by an application on a car tires preference data-set, the main results of our strategy that can be now synthesised in the following points: 1. the definition of an individual model synthesis; 356 2. the visualisation and interpretation of Conjoint Analysis models; 3. the identification of group models; 4. the definition of general hierarchical model for groups. From the Customer Satisfaction viewpoint, the proposed approach allows to define classes with homogenous preferences, hence specialised products can be offered in order to maximise satisfaction in each class. Furthermore, the common model is the best synthesis of consumer preferences as a whole and, consequently, an ideal product can be offered in order to satisfy the most part of the market. Keywords: Preferences analysis, Customer Satisfaction, Factorial techniques, Clustering methods The Typological Non Symmetrical Correspondence Analysis: Theory and Application for Social Research Marilena Fucili and Danilo Leone, D.S.A. S.r.l. Consulting and Training, Italy, [email protected], [email protected] Two-step procedures based on the sequential application of a factor analysis technique and a clustering algorithm have became commonly used in social studies for classification purposes (Greenacre, 1984). If it is available a priori information about the asymmetrical role played by different qualitative variables Non Symmetrical Correspondence Analysis (Lauro, D’Ambra, 1984) can be used, in the first step, in order to aggregate homogeneous units according with their particular structural dependence relationships. However such procedures hardly synthetize collective behaviours, beliefs or attitudes when data sets include several and heterogeneous sub-structures. The aim of our work is to propose an innovative approach to contextually identify and describe local structural relationships which are able to explain different aspects of the dependence of a set of categorical variables on another or more sets. We propose a Typological approach (Diday, 1980) focused on searching for the optimal separation of C sub-sets of units (local factorial sub-models) in such a way that the summation of their explicatory power measures is major than the global model’s one. In the first step of the iterative algorithm the units ui, with i=1,...,N are initially assigned to C groups in order to obtain the starting partition. Indicators of incremental advantage in reallocation are computed for each unit ui originally assigned to the cluster c, evaluating each potential transition from the cluster c to the clusters c’=1,...,C, with c’≠ c. The positive and major indicator will identify the cluster in which the unit has to be reallocated to improve the explicatory power of the so-defined sub-models. At each step it is reconstructed the partition according with the performed change. The convergence is obtained when no other transition is able to increase the heterogeneity explained by the dependence model synthetized by the local factorial models. Once obtained more sub-models it is possible to perform separate analysis of local factorial planes. Finally we suggest a method able to produce a global optimal compromise model in the style of Multiple Factor Analysis (Pages, Escoufier, 1998) Theory and advantages of the suggested method will be illustrated by means of examples on real data sets. Keywords: non symmetrical factor analysis, cluster analysis, typologies Some Different Typologies of Employment in Italy M. G. Grassia, F. Pintaldi and L. Quattrociocchi, ISTAT - Department for statistical production and technical-scientific coordination, Labour Force Survey, Via A. Ravà, 150 - 00142 Roma, Italy, Tel. +390654225049, Fax. +39065430853, [email protected], [email protected], [email protected] In order to comply with EUROSTAT standards, during the year 2002 the Italian National Institute of Statistics (ISTAT) started to design a new Labour Force Survey. After an experimental phase, the new survey is now carried on in parallel with the current one and will substitute it completely in June 2004. 357 Innovations in the new survey concerned both contents and technical aspects. In particular, in the collection of data, we passed from a PAPI technique to a mixed technique: it is used the CAPI technique for the first family interviewee and the CATI for the confirmation of the three following interviews. The questionnaire for the new survey is particularly complex and consists in a general opening part which collects information on the family’s demographic characteristics and 12 sections, 10 of which are repeated for each family member. The amount of information collected in the different waves is very big, and is very important exploit it to achieve other results a part from the employment rate. The aim of this paper is analyse all the collected information in order to investigate labour characteristics and identify several homogenous profiles of labour forces in Italy, not necessarily based on classical distinctions like, for example, employees and not employees or self- employees and vs. employees. In this way, will be possible to analyse the evolution of the labour market, verifying the changes of the profiles by years. The homogenous profiles are obtained on the basis of the microdata in each wave of the Survey using a multidimensional approach (Multiple Correspondence Analysis, Cluster Analysis and Symbolic Marking Technique). The analysis consist in four steps: 1) the use of multiple correspondence analysis (MCA) for data reduction of some categorical variables on the characteristics of work into a few principal dimensions; 2) the use of cluster analysis and symbolic marking technique in order to identify homogenous profiles; 3) the comparison of the profiles at different times/waves and the measure of the their variation; 4) ) a graphical representation of the profiles that permits to visualise the relationship between the profiles, their similarity or dissimilarity and their changes. Keywords: New Labour Force Survey, MCA, Cluster Analysis, Symbolic Marking. SESSION: IMPROVING SURVEY METHODOLOGY: THE EUROPEAN SOCIAL SURVEY Organizor: Ineke Stoop1 and Willem Saris2 Contact: 1Social and Cultural Planning Office of the Netherlands PO Box 16164 2500 BD The Hague The Netherlands, [email protected] 2 University of Amsterdam, The Netherlands, [email protected] Keywords: Cross-national research, optimal comparability, questionnaire design, translation, sampling, non-response Abstract In August 2003, the first data from the new, bi-annual European Social Survey (ESS) will be released and be freely available to researchers all over the world. This survey is jointly funded by the European Commission, the European Science Foundation and the National Science Foundations within 24 European countries. It focuses on changes in attitudes, values and behavioural patterns in the context of a changing Europe, and aims at achieving the highest standards of cross-national survey research, attempting to match the quality standards of the best national surveys. With the help of experts from different fields, a Central Coordinating Team has developed the survey design, the questionnaire and strict guidelines on issues such as random sampling, translation, response rates and fieldwork documentation. In each participating country a national coordinator and a local survey organisation have taken care of conducting the fieldwork. By August 2004 a wide range of quality assessments studies and methodological papers should be available, both by those who were involved in the project from the beginning, and by social research methodologist all over Europe. Further information on the ESS is to be found at: www.europeansocialsurvey.org. 358 Measuring Attitudes towards Immigration across Countries: Potential Problems of Functional Equivalence in the ESS Nina Rother, Center for Survey Research and Methodology (ZUMA), B 2.1, P.O. Box 12 21 55, D68072 Mannheim, Germany, Tel: +49/(0)621/1246-285, Fax: +49/(0)621/1246-100, [email protected], www.gesis.org/zuma This paper analyses the functional equivalence of attitudes towards immigration in internationally comparative research. Methods include secondary analysis of ESS 2002/2003 data (immigration module) and some preliminary cognitive tests. Data-analytical tools employed range from multidimensional scaling to variance-analytic procedures. The results show that some of the ESS measures might not be regarded as functionally equivalent. To know about attitudes towards immigration is essential not only for researchers but especially for politicians. Immigration plays a more and more important role in European societies nowadays, as can be seen in growing concerns of refugee issues or the importance to lower restrictions for highly skilled migrants that are needed in certain industries. In order to provide a knowledge base on attitudes towards immigration in Europe, a special module on attitudes towards immigration was included in the ESS round 1. However, attitudes towards immigration are not easy to measure and especially to compare across cultures. Different migration histories and policies in the different European countries make it hard to think of a shared understanding of what immigration and immigrants are. A cross-cultural comparison of attitudes towards immigration can only be done when functional equivalence is given, i.e. that the same construct is measured in all cultures. Hence, this has to be tested before analysing the data on the substantive level. Functional equivalence of data can be affected by three sources of biases: construct bias, method bias and item bias. In order to analyse the functional equivalence of ESS immigration attitudes measures two distinct analytical strategies are employed. The first is made up by a secondary analysis of the ESS 2002/2003 data. However, in many cases, evidence from secondary analysis is not conclusive and additional studies are needed, such as cognitive tests and split-ballot experiments. For this purpose, the results of some cognitive tests conducted in 2003/2004 are presented. Data-analytical tools employed range from multidimensional scaling to variance-analytic procedures. The results show that some of the ESS measures might not be regarded as functionally equivalent. Some possible solutions are outlined. Keywords: comparative research, cross-cultural survey research, ESS, functional equivalence, construct bias Quality Of Life in Europe: Comparing States using a Composite Index Jeroen Boelhouwer, Social and Cultural Planning Office, The Netherlands, [email protected] The Dutch Social and Cultural Planning Office (SCP) developed an integrative tool for measuring quality of life: the living conditions index (LCI). LCI combines eight domains of the living conditions into one single index (housing, health, mobility, leisure activities, consumer durables, social participation, sport and holiday activity). The index was developed in 1974 and results have been published ever since. As the index is based on Dutch data, developments for the country are followed over time and comparisons are made between social groups in Dutch society. However rich this information is, no comparison can be made with other countries. This means we cannot tell whether the situation in the Netherlands is (developing) better or worse than for example in Spain or Belgium. However, because of the growing integration within the European Union, the question of comparison between (member) states becomes more and more important. One of the main goals of the Maastricht Treaty was the improvement of living conditions in the member states. Therefore there is a growing attention for social monitoring and social reporting. Current reports, like The social situation in the European Union, comprise various indicators in social issues. They lack, however, an integrative measuring instrument. With the European Social Survey a great source for comparison is available. 359 The paper will seek for an integrative index for quality of life based on ESS data. SCP developed a conceptual model for measuring living conditions that will be used for the analysis. At the centre of this model are the living conditions, measured as a multi-dimensional concept. By doing so various domains of quality of life, like housing, leisure time and participation, are examined at the same time. Further, the model considers education, employment and income as individual resources for realising good living conditions. As against the actual situation, as measured by the composite index, another part of the model is made up of individual evaluations of their situation: are people satisfied and happy? The last two topics will be addressed describing the theoretical model in which the composite index is placed. As the main goal of the paper is to explore possibilities for constructing a composite index which is meaningful and useful for describing and comparing quality of life in different states of the European Union. Keywords: living conditions index, quality of life, monitoring, living conditions, the Netherlands, composite index Development of a Standardized Interviewer Questionnaire for International Research into Interviewer Effects on Nonresponse and Data Quality Joop Hox1 and Edith de Leeuw2 1 Department of Methodology and Statistics, Utrecht University, The Netherlands, [email protected] 2 Plantage Doklaan 40, NL-1018 CN Amsterdam, The Netherlands, tel + 31 20 622 34 38, cell phone: + 31 6 53 69 3815, fax + 31 20 330 25 97, [email protected] Interviewers play a key role in contacting and convincing potential respondents and recent research has focused on the role of interviewer attitudes and behaviour on nonresponse (Campanelli et al, 1997, De Leeuw et al, 1997), Groves & Couper, 1998; Stoop, 2003). As part of this effort, several interviewer questionnaires were developed and analysed (for an overview see Hox & de Leeuw, 1998; Hox, de Leeuw, et al, 2002), and the need arose for one standardized questionnaire to stimulate international research. Therefore, IQUEST was developed, a questionnaire for both face-to-face and telephone interviewers. IQUEST incorporated the accumulated knowledge of the earlier interviewer questionnaires, omitting questions that did not work based on both psychometric and substantial analysis. Also new questions were added based on new theoretical and empirical insights. A master version was developed containing context information about the questions; English was used as lingua franca (Harkness et all, 2003). This master questionnaire was checked for clarity and intercultural translatability, and made available to countries participating in the European Social Survey (ESS) 2002. As the ESS 2002 used highly standardized questionnaires and data collection procedures, including strongly standardized field methods and non-response registration, this is an ideal situation to study interviewer effects cross-nationally. In the present paper, we present the first results of this effort. We give an overview of available translations of the interviewer questionnaire and the results of psychometric analyses on the questionnaires cross nationally. Keywords: questionnaire development, psychometric properties, index construction, master questionnaire, available translations Improving Survey Methodology: The ESS. Index and Measurement: The Cultural Borders of Meaning Antonio Alaminos, Mathematical Sociology, University of Alicante, Spain, [email protected] ESS is, very probably, one of the most significant efforts to improve social survey methodology. No doubt, its aimed by the "best practices" spirit. In that sense, to put at work extensively the "manuals" of survey methodology is certainly a qualitative jump in quantitative research. The starting point is optimal to reach excellence. 360 This paper point out some comparability problems arriving from questions wording and translations, that emerges when testing measurement cross-culturally. Even with the best control on translation and wording, the invariance on theoretical meaning may not be reached. This is the case when questions define and stand for indicators of something different. For example, when measuring a latent variable. Then, questions and variables try to capture a different reality that the language suggest. For simplicity, we will deal with one of the most obvious: the concept of government from a cross-cultural approach. The conclusions will suggest some considerations to improve comparability. Response Propensity and Data Manfred Max Bergman, Nicole Schoebi and Dominique Joye, SIDOS: Swiss Information and Data Archive Service for the Social Sciences, Switzerland, [email protected] The response rate of a survey is an important quality criterion for making population inferences. These rates may depend in part on past experiences and propensities of the respondent, the skill of interviewers to create a positive rapport within the first few seconds of contact with respondents, the perceived sensitivity or complexity of the survey topic, question and questionnaire design, etc. Various techniques are employed to boost response rates in populations that often suffer from survey fatigue. However, such strategies may have detrimental effects on data quality. Repeated follow-up contact with initial non-respondents tends to increase the survey response rate although it is not clear to what extent the quality of response from follow-up contacts differs from other respondents. More precisely, it is unclear to what extent reluctant participants simply give in to social pressure by formally obeying instructions without necessarily committing themselves to the demands of the questionnaire, or whether they indeed comply after having been convinced about the value of their participation. Based on detailed event recording during data collection for the Swiss portion of the European Social Survey (2002), we will investigate the effects of various response propensities and degrees of response acquiescence on a number of data quality indicators. As participation and response style relates to various demographic and social variables, such as age, education attainment, and social position, we will include these in our analyses. More generally, we will explore interaction effects between response propensity, socio-demographic background, and data quality. Moreover, this survey included a methodological experiment, in which a sub-sample was contacted in a centralised manner, while the remaining sample was contacted directly by the interviewer. Results from this quasi-experiment will supplement our investigation on response propensity and acquiescence. Keywords: nonresponse, bias, survey quality Geographical Variations in Electoral Participation: Evidence from a Multilevel Analysis of the European Social Survey Ed Fieldhouse and Mark Tranmer, Cathie Marsh Centre for Census and Survey Research, University of Manchester, UK, [email protected], [email protected] Turnout at general elections across Europe is in decline. For example in the U.K. in 2001 turnout declined to 59%, the lowest level since 1918. Reported turnout in the European Social Survey across the 21 countries for which data are currently available is 74% (though this is likely to be an overestimate due to non-response bias and misreporting of vote). This paper examines national and regional variations in turnout and uses multilevel models to understand contextual and compositional explanations of such variation. Contextual variations may arise, in part, because of election specific factors such as the closeness of the major parties in the polls, or the number of parties competing in an election. These variables are consistent with rational choice explanations of turnout which predict voting is a function of the costs and benefits of voting and also the likelihood of an individual electors 361 vote making any difference (i.e. where an election is not closely fought, the perceived likelihood of casting a decisive vote is reduced). However, previous studies have shown that such explanations are likely to provide only a partial insight into individual level variations in turnout. Thus we turn to two other major theories to explain these variations: social capital and civic voluntarism. In addition we explore how different key groups (e.g. ethnic minorities and young people) vary in their participation across different countries, and how well these differences between such groups are accounted for by these models. Finally, we look at wider measures of political activity, and consider how these are related to electoral participation. Keywords: European social survey, multilevel models, electoral participation. European Value Map. Based on ESS Data Hans Bay, SFI-SURVEY, Denmark, [email protected] The European Social Survey (the ESS) is a new, academic-driven social survey designed to chart and explain the interaction between Europe's changing institutions and the attitudes, beliefs and behaviour patterns of its diverse populations. The central aim of the ESS is to develop and conduct a systematic study of changing values, attitudes, attributes and behaviour patterns within European policies. Academically driven but designed to feed into key European policy debates, the ESS will try to measure and explain how people’s social values, cultural norms and behaviour patterns are distributed, the way in which they differ within and between nations, and the direction and speed at which they are changing. A two-dimensional map (European Value Map) is constructed upon 21 questions. The questions (named 21-item Basic Human Values Scale, developed by S. Schwartz) are grounded in theory and well tested internationally. The two dimensions in the map are: The individual possibilities and behaviourism in the society. These two dimensions can be found in all the countries, when the countries are analysed separately. It is demonstrated that the two dimensions are identical. Therefore the two dimensions are constructed from the total sampling. The total sampling covers 18 countries and 32,102 respondents. In the European Value Map 18 European countries are placed and thereby a comparison between the European countries can be made. Keywords: ESS, two-dimensional value map, factor analysis. Analysis and Relation between Social Variables and Economic Variables in the European Social Survey and World Bank and UNDP Reports: The Social Capital and Economic Capital at the European Countries Juan Sebastián Fernández Prados1, Jaime Andreu Abela2 and Antonia Lozano Díaz3 1 University of Almería, Spain, [email protected] 2 University of Granada, Spain 3 CentrA Foundation, Spain Our main goal is to confirm and to explore the hypothesis of the relation existing between the social capital and the economic capital or economic development of the countries. On one hand, we have highlighted three main components of the concept of social capital (participation in associations, civic and political engagement and interpersonal trust), from the definitions elaborated by different authors like Putnam, Coleman, Fukuyama, etc. We have analyzed the questionnaire of the European Social Survey and we have obtained three indicators for the study of the social capital in the different European countries: associations rate (percentage of population from one country that has taken part in at least one in the last 12 months), civic, social and political participation rate (percentage of population from one country that has carried out a civic or social action in the last 12 months) and level of interpersonal trust (arithmetic mean of 362 the following scales: "Most people can be trusted”, “most people try to take advantage of you” and “most of the time people helpful”). On the other hand, we have highlighted mainly two indicators utilized to measure and to rank the countries of the world regarding their economic development or economic capital, gross domestic product per capita PPP current international $ (GDP per capita, used by the World Bank) and the Human Development Index (HDI, created by the United Nations Development Programme). In the database of the World Development Indicators (2003) of the World Bank we can find more than 500 variables containing information of 208 countries of the world and, likewise, in the last Human Development Report of the UNDP (2003) we can also find hundreds of statistical and a dozen of indexes created ad hoc such as the gender -related development index (GDI), human poverty index (HPI), etc. of 175 countries including all the countries studied in the European Social Survey. The statistical treatment used to examine the relationships between the two groups of indicators, social capital and economic capital, was, firstly, to elaborate a description of all the social and economic variables of the countries; secondly, to obtain the correlation coefficient (r) and statistic of the regression (R-square) from the matrix of correlations of all the variables and from different linear regression models; finally, we carried out a series of Simple Scatter plots of the most important results. Keywords: Social Capital, Economic Development, Human Development Index, European Social Survey. Influence of Media Reported Events on Attitudes and Opinions Jurjen Iedema and Ineke Stoop, [email protected] In 2002 and 2003 the first round of the European Social Survey was held in a number of European countries. Between 1500 to 2500 respondents per country were questioned and this usually took several months to accomplish. During this period important events on a national or worldwide scale could influence respondents’ answers. During that time the (coming) war in Iraq was such an event and, for example, in Spain soldiers were sent to Iraq by the government whereas a large majority of the Spanish people were opposed. Consequently, one might expect that Spanish respondents would show a decrease in trust in government. In this way we looked at the influence of events in each country on a number of questions, especially trust in political institutions. On some of these opinions large variation in time was observed and occasionally one could summarize these as linear or curvilinear trends. The predictability of these trends from events that took place in the country varied. Keywords: contextual information, external validity Some Methods for Investigating Validity and Reliability of Questions in the European Social Survey A. J. Kutylowski, Buskerud College, Oevrefoss 8d/11, 0555 Oslo, Norway, tel. +(47) 21910596, [email protected] The core module of the European Social Survey questionnaire is supplemented with a replication module containing a number of questions which purport to be indicators of the same kind as questions in the core module yet differ in one or more ways as to the manner they were asked. Purpose of the presentation is to investigate efficacious statistical modelling of the information as to the agreement between two or more manners of asking the question. The issue of interest deals with the basics of survey instrumentation in the European Social Survey, namely whether a change in the format of the questioning leads to change in the substantive interpretations of the finding, i.e. with the validity and/or reliability of the measurement instruments. Publications of the ESS coordinating team appear to favour MTMT “approach” to resolving these issues. Here a different approach will be presented, based on direct modelling of joint multivariate distributions of alternative question wordings and/or formats. Log-linear models will be used for 363 structural representation of the distributions and special parametrizations of the models, either previously proposed in the statistical literature or entirely new will be employed to formulate and test hypotheses inherent in the design of the replication module. A general symbolic notation will be used to express the structural parametrizations, so as to make it straightforward for researchers to carry out similar analyses in the context of their own investigations based on the study’s data. With respect to some questions a latent variable approach will be appealing: here we shall show how to link relevant graphical log-linear models for ordinal variables with the nonparametric latent factor modelling. Examples will presented showing how the proposed methods help in an investigation with respect to whether some response categories can be collapsed; whether there is a “middle” category; whether it is feasible to proceed under the assumption of approximately equal intervals between categories of the response. Keywords: log-linear models, agreement, opinions, attitudes, latent variable, ordinality, European Social Survey 2002. Modelling Interviewer-Effects in the European Social Survey Geert Loosveldt and Michel Philippens, Centre for Survey Methodology, Centre for Data collection and Analysis, Department of Sociology, Catholic University Leuven, [email protected] Survey researchers are since long concerned with the fact that interviewers can introduce an additional source of variation in survey-estimates. It is assumed that this is most likely to occur when interviewer behaviour departs from standardized procedures. In this paper we argue that the frequency of idiosyncratic interviewer behaviour and accordingly the size of interviewer-variance are mainly a function of the quality of the survey organisation (e.g. quality of interviewer training) and the amount of problems respondents encounter when answering questions. In order to test this proposition we used data from several countries participating in the European Social Survey (ESS). By comparing the size of interviewer-variance across countries we hope to learn more about the factors and conditions that influence the presence of interviewer-effects. In a first section of this paper we will test whether countries differ significantly with respect to the size of interviewer-variance on a set of items designed to measure ethnocentrism attitudes. This will be done by modelling interviewer-variance as a function of the respondent’s country in a multilevel regression model. Next, we will introduce to our model a number of variables associated with the answering behaviour of the respondent - i.e. degree of understanding questions, motivation and reluctance (as evaluated by the interviewer) and presence of item-nonresponse. This will allow assessing whether and to which extent country-differences of interviewer-variance can be attributed to differences in problematic respondent behaviour. Finally, we will try to explain the remaining country differences in interviewer-variance by comparing the quality of survey organisation across countries. Keywords: ESS, interviewer effects, multilevel modelling, survey quality International Benchmarking: Are we Comparing Apples to Oranges? Frank Stroobandt, Mariska Storm and Hans Waege, University of [email protected], [email protected], [email protected] Ghent, Belgium, International benchmarking is a process by which an organisation compares its results or performance indicators with organisations in foreign countries. Public authorities increasingly use this instrument to plan, evaluate, and improve their policies. Although benchmarking can be a valuable tool for policymakers, there are some risks involved. In this paper we focus on the difficulties and pitfalls regarding the selection of countries or regions that are included in an international benchmarking project. Our aim is to develop a methodology which assesses the relevance of comparing a given 364 country or region with other countries or regions. This methodology is tested in the field of cultural policy. It is a remarkable fact that the selection of countries or regions in international benchmarking is usually taken for granted. The decision to choose certain countries or regions is generally based on practical criteria such as the availability of data and the belonging of a country to a supranational entity, e.g. the European Union. When the selection process is not methodologically substantiated, benchmarking may lead to irrelevant comparisons and distort the policy decisions that are based on it. The methodology we developed consists in matching the countries or regions that are considered for benchmarking on a series of contextual indicators. The latter were defined and selected during focusgroups with Flemish policymakers. The indicators that are involved in the matching procedure are parameters of the economic, social, demographic, juridical, political, and financial factors that operate in a given country or region. Since the methodology is tested in the cultural field, we also defined measures that control for the contextual factors, e.g. subsidy structure, with regard to this domain. The matching procedure was applied on an international cultural benchmarking project that was initiated by the Flemish government. The results show that some countries and regions such as The Netherlands and France are more relevant as reference point for Flanders than others European countries. We note that certain non-European regions, for instance Québec, match better with Flanders than regions that are situated in Europe. We conclude that the quality of an international benchmarking process is also determined by the initial stages in this process, namely the selection of the countries or regions that are benchmarked. Using a matching process with regard to benchmarking projects in other policy fields is recommended. Keywords: international benchmarking, comparability, quality, methodology, cultural policy. Modeling Events for ESS: Toward the Creation of an Autonomous Tool for Survey Research Theoni Stathopoulou, National Centre for Social Research, Greece, [email protected] The European Social Survey has supplemented its questionnaire with the creation of the event database in order to methodologically integrate the survey responses. The aim of this database is to record and evaluate the impact of historical circumstance, as reported by the media, in the shaping of attitudes. While event data analysis is a powerful methodological tool in the field of international relations for crisis monitoring, its application in the case of the survey has presented some difficulties, particularly concerning the combination of the macro and micro levels of analysis. In order to overcome these difficulties, it is proposed here that event data be approached as an autonomous tool. Such independence empowers and broadens our analytical capabilities, allowing the better evaluation, monitoring and utilisation of the events. This approach would better serve the needs of this particular survey in the long run, and may also be of use to surveys of a similar nature. The proposed model is based on the simultaneous analysis of the three main components of the event database: time, space (the total countries), and events. Using multi-dimensional analysis, and taking into account the serious homogenisation problems apparent in the data, we will attempt to highlight: the continuous flow of events; the possible impact of the events within each country and across countries diachronically; the shifting of the thematic event categories within countries; and the appearance of new thematic categories. To this end, lexicographic analysis with parsing techniques will be carried out on the total available event reports of each country. Subsequently, cluster analysis will be used to associate the countries with the thematic event categories in order to reveal the “event identity” of each country, which affects the attitudes of the respondents, and to understand the processes involved in the development of these identities. This model aims at the creation of a dynamic (the entry of new events each month), diachronic (analysed by month) system, which will monitor, map, and compare the evolution of the events in time and space. Keywords: events, text-mining, ESS, multidimensional analysis. 365 STREAM: INTERNATIONAL COMPARATIVE RESEARCH SESSION: COMPARABLE DATA ACROSS COUNTRIES Organizor: Willem Saris1 and Ineke Stoop2 Contact: 1University of Amsterdam, The Netherlands, [email protected] 2 Social and Cultural Planning Office of the Netherlands PO Box 16164 2500 BD The Hague The Netherlands, [email protected] Keywords: comparative research , cross cultural, reliability, validity, sampling theory, survey research, non-response, functional equivalence Abstract In the design of the European Social Survey (ESS) that is going on in 23 countries of Europe a maximal effort has been made to obtain comparable data across countries. This means that: the samples have been made maximally comparable strict rules have been formulated for the fieldwork rules have been formulated for the translation of the questionnaire in different languages in all countries extra questions have been asked to estimate the reliability and validity of the questions in order to correct for measurement error event data have been collected to understand possible differences due to specific events In this session different members of the Central Coordinating team of the ESS will present the procedures they have developed to maximize comparability across countries Pursuing Equivalence in Cross-National Surveys Roger Jowell Comparability in cross-national attitude surveys has always been notoriously difficult to achieve. Cultural incompatibilities and methodological variations affect questionnaire design, sampling, fieldwork, response rates and coding. Problems of ‘non-attitudes’ and social desirability biases abound. The political or social context varies by nation and across time. The 22-nation European Social Survey – started in 2001 and funded by the European Commission, the European Science Foundation and the academic funding agencies of all participating nations – tries to mitigate these problems by means of a wide range of innovative methods. Thus even in its first round the ESS achieved rigorous probability samples of a similar ‘effective’ size and quality in all participating nations, as well as achieving unusually high standards of translation and fieldwork and higher response rates. Accompanying the main dataset is a meta dataset containing details of questionnaire development and testing, contact and calling strategies, a separate file of event and context information, and a comprehensive and transparent technical report. It is therefore also a rich resource for methodological research. The ESS’s fully-documented dataset was made available on-line in September 2003, only months after the end of fieldwork. Its quality and outreach suggest that the detailed organisational networks created to implement the ESS, together with meticulous design and documentation protocols, embedded expertise and ambitious targets, added up to a winning formula for ensuring impressive cross-national equivalence. Keywords: Comparative Attitudinal Research, Survey Methods, Equivalence Comparability across Countries of Responses in the ESS Willem E. Saris, University of Amsterdam, The Netherlands In cross-cultural research a problem is that the questions have to be translated in different languages. In each language each question has a specific quality expressed in a reliability and validity coefficient. It has been shown that the data quality has a very strong effect on the relationships between variables: 366 the quality of a question can decrease relationships but can also incorrectly increase correlations. In cross-cultural research the problem is that , even if the questions have been translated perfectly, the quality can vary from country to country and this means that the relationships between the variables in the different countries are incomparable. In the paper I will present evidence of this phenomenon on the basis of data of the ESS. I will also show how one can take these differences in data quality into account so that the relationships between variables are comparable across countries. Trapped in Translation? ESS Translation Protocols provide a Key Janet Harkness, ZUMA, Mannheim, Germany, [email protected] The ESS translation context is complex: multiple languages, a source questionnaire finalized before translation begins, numerous countries sharing a language (e.g., French and German), and language minorities over 5% provided with their own translations, country budgets already stretched by other requirements for ESS fielding. The EU funded a work package to provide translation guidelines and requirements for countries participating in the ESS. The guidelines provided are based on ongoing basic and applied research in survey translation on how to implement survey translation and assessment through a holistic, teambased process. The commonly used and originally envisaged procedure of back translation was successfully replaced with TRAPD – a translation, review, approval(adjudication), pre-testing and documentation protocol. Results and feedback from the first round of the ESS are positive and an increasing number of international projects are incorporating the basic idea of TRAPD in their translation production processes. At the same time, basic research within the ESS framework and in other studies suggests that questionnaire design is the ultimate issue. The paper outlines the theory and practice of TRAPD and using data from the ESS and other studies demonstrates how translators can help improve questionnaire designs. Sampling for the European Social Survey Seppo Laaksonen1, Siegfried Gabler2, Sabine Häder2 and Peter Lynn3 1 University of Helsinki, Finland 2 Zentrum für Umfragen, Methoden und Analysen (ZUMA), Germany 3 Institute for Social and Economic Research, University of Essex, UK The ESS has used innovative methods to ensure that the samples interviewed in round 1 are of the highest possible quality and are comparable between countries. In this presentation, we will describe why it was important for the ESS to pay close attention to sample design and outline some key achievements of the ESS with respect to sample design. There is considerable variation between countries in the constraints upon sampling (existence of and access to population registers and other sampling frames) and in usual survey practice. ESS has tried hard to minimise the impacts of this variation on the representativeness of the samples. The aim was to achieve samples in each country that represent the same population (all ages 15+, all de facto residents – not just citizens) and provide a similar level of statistical precision. A precondition of comparability is that probability sampling must be used in each country. In addition, the ESS specified the level of precision that should be achieved, in terms of a statistical concept, the effective sample size. Countries were provided with guidance on how to apply this concept to their situation and how to interpret the implications by a central team of sampling experts. This team also helped the national teams to identify and implement appropriate sampling frames and sampling methods and good ways to maximise response rates. A fundamental belief of the ESS is that this co-operative approach with 367 constant multi-way dialogue is more likely to achieve functional equivalence than the application of standard procedures. We will describe the key aspects of the sample specification and how the support and assistance was given to each country. We believe that these processes represent a significant advance on the implementation of other cross-national surveys and have resulted in a higher quality survey product. Keywords: cross-national surveys, sample designs, functional equivalence Data Quality Assessment in ESS Round 1: Between Wishes and Reality Jaak Billiet and Michel Philippens, K.U. Leuven, Belgium The quality assessment of the obtained data from first round of ESS is based on a conceptual framework of a pragmatic approach to data quality assessment1. It combines aspects of the total survey error approach that are focused on output evaluation, and the total quality management approach, which is concentrated on process evaluation. It was intended to cover the process and output aspects of both the sample obtained, and the registered responses. For practical reasons, not all elements of the full pragmatic approach could be addressed in the first round of ESS, but the most important ones are covered. The utilisation of extensive uniform contact forms was crucial. Most of the countries performed well on this. Documenting the process of contacting the sample units within several sampling designs and contacting strategies, obtaining information about the reasons of non-contacts and refusal, constructing comparable non-response rates over countries, and obtaining information about co-operative respondents, reluctant respondents, and non-respondents were the main functions of the contact forms. The paper evaluates the performance of the participating agencies, and discusses the possibilities of improvement. The study is mainly based on the analysis of contact forms data, but also on substantial data from the main questionnaires. 1 Loosveldt, G., A. Carton & J. Billiet (2004), “Assessment of survey data quality: a pragmatic approach focused on interviewer tasks”. International Journal of Market Research, vol. 46 (1): 65-82. Context, Events and Attitudes Jurjen Iedema and Ineke Stoop In the early stages of the preparation of the European Social Survey, during the final years of the previous century, when the construction of an event data inventory was planned as a corollary to survey data collection, the most disruptive events most people had in mind that could influence individual reactions to certain questions were national elections. The 9-11 terrorist attacks in the USA in 2001 and its aftermath showed that events could have an infinitely more far-reaching effect on values and opinions. When designing the actual data collection for the event inventory early 2002, it was still assumed that similar high impact events would occur only very sporadically. Now, with a war in Iraq that seems to be still going on, the importance of collecting event data during fieldwork seems more pertinent than ever before. The fact that the fieldwork of the first wave of the ESS took place in period when war was being prepared with uncertain consequences, when the influence of international organisations was under discussion and when European countries were seriously divided among themselves, is something future users of the ESS should be aware of, especially as the study has been set up as an important asset for historical micro analysis. The impact of the war in Iraq on the social and political fabric of Europe is something substantive researchers should especially take into account as this international event translated itself or was related to many distinctive national events, such as attitudes towards the Blair and Aznar government in the UK and Spain, the outcome of the Dutch and German national elections, the enmity between the USA and France and Germany, the involvement of the Polish army, 368 public demonstrations and anti-war rallies, etc. The single event of the war in Iraq thus may have influenced societies in a way which was not uniform across countries To pinpoint the impact of the national ‘Iraq’ events, the involvement of local reporters is required who can relate events to the contents of the ESS-questionnaire. During fieldwork they forwarded perceptive overviews of what kept people – or more precisely – newspapers busy during the final months of 2002 and the first of 2003. From the media reported event inventory of the national reports it can be seen that the ESS 2002/2003 was conducted in turbulent times: reports mention regime changes, terrorist attacks, major floods, heated public discussions on immigration and asylum seekers, political wranglings, financial scandals, planned revisions of social protection regimes, economic decline and national elections. When analysing attitudes, the event inventory should provide the necesary background. The complete overview of the events that took place in the participating countries during fieldwork is publicly available at the ESS-websites, together with an overview of websites with country data, containing both aggregate statistics and system descriptions. SESSION: METHODOLOGICAL ISSUES IN DESIGNING IMPLEMENTING CROSS-CULTUREL SURVEYS AND Organizors: Janet Harkness1 and Edith D. de Leeuw2 Contact: 1ZUMA, Postfach 12 21 55, D-68072 Mannheim, Germany, [email protected] 2 Edith de Leeuw, MethodikA, Plantage Doklaan 40, Nl-1018 CN Amsterdam, The Netherlands, [email protected] Keywords: Cross-cultural methodology, multi-cultural survey design and implementation, multipopulation sampling, multi-cultural survey instruments, quality control, non-response and survey errors Abstract Cross-Cultural Surveys have become increasing important both from a theoretical and an applied policy view. Understanding complex realities and generating and testing social theories is one of the main issues in social science. Going from the limitations of one culture, nation, or group to comparing groups, cultures and countries, enables researchers to distinguish between ‘local conditions’ and ‘universal regularities.’ As our world grows from a local to a global one, policy makers and economists have an urgent need for comparative high quality international data. A Process Quality Approach to the Testing and Evaluation of Cross-Cultural Surveys Johnny Blair and Linda Piccinino, Abt Associates Inc., USA, [email protected], [email protected] Instrument development for cross-cultural surveys often must contend with problems that have not been widely addressed in the literature. Survey methodology has not focused much on these issues, though it has some tools and theory to do so (Schwartz, 2003; De Leeuw & Hox, 2003). Using recent examples from health surveys in Africa and Asia we present development of processes to ensure quality in survey instruments used across purposes. Survey instruments designed for one cultural context sometimes are not easily adapted to fit additional contexts. Often much of the design of instruments involves technical staff who are not native to the target country, or instruments that are developed for one country must be modified for use in another. Client demands and resource constraints, however, might make using the same instrument for more than one culture a necessary, although often not an optimal, solution. Evaluation of surveys that are intended for other cultural contexts should rely on processes to test whether the index survey and its 369 proposed modifications are suitable to the next context. The need for comparability of selected questionnaire items and the preservation of time trend data add another level of complexity to the task. We develop a process or series of stages to show the types of problems that can be addressed by alternative testing approaches such as focus groups, cognitive testing and conventional pre-testing methods. The stages of testing or pre-testing are comprehensive, starting from testing of all sections of the current instrument through testing on target populations, and additional testing of flawed or significantly changed questions. The type of pre-testing or testing method, whether focus groups, behavior coding, or conventional testing, should be conducted in stages to (1) identify problems and (2) verify that revisions made have addressed the problems. Expert review, while beneficial, should not be viewed as a substitute for these methods of testing. This paper describes a process approach for the combined use of these multiple tools for instrument development in cross-cultural surveys. Keywords: pre-testing, questionnaire design, measurement error Language, Usage, and Cultural Context: Challenges to “Functional Equivalence” in Questions Michael Braun and Janet Harkness, European Center for Cross-Cultural Surveys, ZUMA, P.O. Box 12 21 55, D - 68072 Mannheim, Germany, [email protected], [email protected] In cross-lingual, cross-cultural studies, appropriate language and translations often play a key role in securing intercultural comparability. Sometimes language as language seems to be the foremost issue, sometimes more a blend of language, language use, and culture. At other times, problems of “functional equivalence” have less to do with language and how it is used than how cultural contexts frame respondents’ understanding of questions. We explain how--against the backdrop of specific cultural contexts--cognitive and communicative processes can trigger different interpretations of survey items. The paper provides illustrations of each of these challenges to “equivalence” and demonstrates how cultural contexts—linked to language or not—relate to respondent perceptions in ways similar to question contexts; these last play a prominent role in the social-cognition approach to survey research. Often enough, the effects of language system and usage and cultural context may be hard to disentangle. The paper explores the relationship between language-anchored features and nonlinguistic aspects of survey questions in context that create problems for “equivalence” or comparability. Selected examples illustrate the usefulness of an integrated framework in trying to come to terms with social science research across cultures and language. Keywords: Boundary Crossing Data Use, Measuring Accuracy in Complex Surveys Developing a Low-Cost Technique for Parallel Cross-Cultural Instrument Development: The Question Appraisal System (QAS-04) Elizabeth Dean, Rachel Caspar, Georgina McAvinchey, Rosanna Quiroz and Leticia Reed, RTI International, USA, [email protected] Many different approaches are used to prepare instruments for multicultural administration, depending on the scope, schedule and budget of the study. Sequential questionnaire development, the most common approach to developing cross-cultural instruments, is also the most affordable. Using the sequential approach, instrument designers formulate and pretest an instrument in the source language, then translate the final version into the target language(s) using as much culture-specific tailoring as is possible. On the other hand, the parallel approach to multilingual questionnaire development incorporates target cultures into the design and pretesting process. The disadvantage to parallel development is that it is expensive, time-consuming, and subject to version control problems (Harkness, Van de Vijver, and Johnson 2003). As researchers design a growing number of instruments 370 across languages and cultures, the need arises for low-cost techniques for parallel instrument development. RTI’s Question Appraisal System (QAS) is a coding tool used for pretesting instruments. The QAS is supported by an item taxonomy that describes the cognitive demands of the questionnaire and documents the question features that are likely to lead to response error. These potential errors include comprehension, task definition, information retrieval, judgment, and response generation. This appraisal analysis is used to identify possible revisions in item wording, response wording, questionnaire formats, question ordering, and instrument flow. As a tool designed to address potential measurement problems during the development phase of instruments, the QAS is eminently suited to incorporate language issues and provide a relatively low-cost methodology for identifying potential problems due to cross-cultural and cross-linguistic application of instruments. This paper will present a revised and expanded QAS that enables survey developers to assess linguistic and cultural issues within instruments, including: errors likely to be introduced during translation, differential levels of comprehension across cultures, idiomatic vocabulary, potential complications to programming, such as gender specification and sentence structure around filled-in text, and culturally inappropriate references. When the new QAS taxonomy is finalized, it will be used to evaluate an instrument that is currently in the field in English, Spanish, Mandarin, and Cantonese. To validate the revised QAS, the results will be compared to item nonresponse data from questionnaire’s field administration and to data collected during in-depth interviews with interviewing staff. This paper will present the final revised QAS and an assessment of its effectiveness in predicting cultural problems in the field. Keywords: Cross-cultural methodology, multi-cultural survey design, forms appraisal coding References Harkness, J., Van de Vijver, F., and Johnson, T., (2003), Questionnaire Design in Comparative Research, in Harkness, et. al, (eds), Cross-Cultural Survey Methods, New York: Wiley. – Pages 19 – 34. Test of Anchoring Vignettes in a Low Income Population Patricia M. Gallagher and Floyd J. Fowler, Center for Survey Research, University of Massachusetts Boston, USA We currently are fielding a survey designed to test an approach to enhancing the cross-cultural comparability of Consumer Assessment of Health Plans (CAHPS) instruments that follows a model proposed by Gary King, Christopher Murray, et al. (2003). The technique involves the use of "anchoring" vignettes that present hypothetical examples of health care at various levels of efficacy. Respondents are asked to assess the care presented in the hypothetical scenarios. They are also asked to use the same response choices presented for the vignette assessments to self-assess the domain of care described in the vignettes. The responses to the vignette items can be used to create comparable inter-rater measurements to adjust self-assessments. The study involves surveying parallel samples (total n=1200) of Latino, African American, and white members of MassHealth, the State of Massachusetts medical insurance (Medicaid) program for lowincome residents. The first step in this test was to develop the vignettes. The development work involved two rounds of focus groups (n=6) with members of the target populations and the translation of the vignettes into Spanish. The sample members each receive a Canadian-style dual-language (English/Spanish) instrument that includes the anchoring vignettes developed in stage one and a subset of core CAHPS items. Standard mail survey protocols are being followed, with an initial mailing of the questionnaire packet, followed by a reminder/thank you postcard, and the mailing of a second questionnaire packet to all nonrespondents. 371 Analysis of the data collected will allow the identification and quantification of differences in the ways that different cultural groups use CAHPS response scales. In addition, we will have the tools to calibrate self-assessments to adjust for interpersonal differences in the use of the response scales. References King, Gary; Christopher J.L. Murray; Joshua A. Salomon; and Ajay Tandon. ``Enhancing the Validity and Cross-cultural Comparability of Measurement in Survey Research,'' American Political Science Review, Vol. 97, No. 4 (December, 2003), 567-584; reprinted, with printing errors corrected, Vol. 98, No. 1 (February, 2004): 567-583. Disambiguating Culture: The Cultural Element of Survey Response in Perspective Eleanor Gerber, US Census Bureau, Statistical Research Division, [email protected] Interest in examining surveys in a cross-cultural perspective has risen as surveys have been fielded in complex, multi-national and multi-ethnic contexts. Survey response and respondent behavior in general is understood to vary by cultural, ethnic and linguistic factors. The aim of this paper is to examine some key concepts by which these behavioral differences may be understood. An understanding of these factors is useful both in predicting where differences may occur and in guiding future research. The data on which this paper are based are drawn from qualitative work over more than a decade surrounding the US decennial census. For this discussion, culture is regarded as a set of mental constructs; definitions, assumptions, values and beliefs by which people construct and interpret the world. It is the central insight of anthropology that these beliefs differ by culture. However, this single conceptual insight does not do justice to the complexity in respondent behavior to surveys nor to the range of differences to which survey designers should attend. These factors include: Conceptual differences: An example is the race categories in the US census, which are problematic for persons from Latin America, who often do not find acceptable categories for themselves. Social adaptation: Some relevant differences are not codified in belief, but represent common structural adaptations to external circumstances. Thus, immigrant households from many different systems of belief adapt to high rents by sharing space, regardless of varied ideals of household structure. In addition, wide similarities in attitude towards government data collection are shared by diverse groups with similar histories of conflict with the Federal Government. These social adaptations can create wide similarities in behavior among very diverse groups. Assimilation: Despite very different beliefs, it often does not take immigrants long to learn what they consider to be the expected response. Thus, although “Hispanic” is an unfamiliar concept, many immigrants adopt it as a self-identification in the US context. Different understandings of respondent behavior. The survey itself is a cultural representation. Respondents have a set of cultural expectations about what surveys are and how to participate, which vary by previous experience. These factors affect survey design and procedures, and should be the subject of further research. Keywords: Cross-cultural methodology, multi-cultural survey design, nonresponse and survey errors Question Design Features and Cultural Variability in Comprehension Timothy Johnson, Young Ik Cho, Allyson Holbrook, Diane O’Rourke and Richard Warnecke, Survey Research Laboratory, University of Illinois, Chicago, USA Although cultural variations in survey question interpretation have now been well documented, along with cultural differences in survey-related behaviors such as nonresponse, extreme-responding and acquiescence, little guidance is available to researchers regarding specific question design strategies that might be useful in minimizing variations in interpretation. This paper will address this issue by first reviewing current conventional wisdom regarding cross-cultural question design strategies. It will then present original behavioral coding data from 350 interviews with respondents representing four distinct cultural subgroups in the United States (non-Hispanic White, African American, Mexican 372 American and Puerto Rican). Using survey questions as the unit of analysis (n = 14,000), nested within survey respondents, hierarchical linear modeling (HLM) is employed to examine the effects of five questionnaire design features on cultural variations in question interpretation difficulties. The five design features examined include (1) type of response options used, (2) level of question abstraction, (3) question length, (4) question reading level, and (5) question type (behavior vs. opinion). The main findings of these analyses indicate that the use of vague quantifier response options disproportionately increase the question comprehension difficulties of minority respondents, relative to non-Hispanic Whites. Based on these and other findings, we offer a revised set of question design recommendations for the development of questionnaires to be used in cross-cultural surveys. Keywords: Cross-cultural methodology, multi-cultural survey design, multi-cultural survey instruments Socio-Cultural Factors in the Question-Response Process: A Comparative Coding Analysis of Latino and Anglo Cognitive Interviews Kristen Miller1 and Gordon Willis2 1 National Center for Health Statistics, 3311 Toledo Rd; Hyattsville, MD 20782, USA, 301-458-4625, [email protected] 2 Concepcion Trevino Eason This paper will describe project, conducted by the U.S. CDC National Center for Health Statistics and the NIH National Cancer Institute, in which nearly 70 cognitive interviews on health topics were conducted among Latino and Anglo participants (in both English and Spanish) in rural, urban and suburban locations in the United States. Questions were asked on issues that perennially produce problems in cognitive testing, including reports of chronic health conditions that involve technical terms, and of cancer screening tests. Further, questions were included on disease risk factors that we suspected may show cross-cultural variation, such as diet, nutrition, and physical activity. Finally, we included a question on perceived health status, in order to assess potential between-group conceptions of health in general. Cognitive interviews were conducted using intensive verbal probing techniques, and interviewers recorded subject responses to a series of targeted probe questions designed to elucidate the comprehension and further cognitive processing of the survey items. Further, somewhat unique to this study, interviews were coded, allowing for both quantitative and qualitative comparative analyses, to examine the ways in which socio-cultural factors (e.g. language, cultural values and tradition, socio-economic status and geographic region) might impact the survey response process. Codes were based on a question-response model, depicting (1) the respondent's interpretations of key terms, (2) the process for making judgments and decisions, and (3) the presence of response error and response error type. The paper will present findings related to overall trends concerning identified problems, and in particular, will determine the extent to which variations in responses point to differential item functioning across the range of groups studied, whether defined according to culture, socio-economic status, geography, or some other critical variable(s). Based on the obtained results, we will discuss implications for survey research conducted within a multi-cultural context, and for the practice of questionnaire development, evaluation, and pre-testing. Finally, the discussion will point toward specific directions for future methodological investigations that are critical to the development of international and other cross-cultural surveys, especially those that make use of systematic forms of coding of cognitive interview outcomes. Keywords: question response, question design, socio-cultural factors What are we talking about ?Interviewing and Interpretation for a Biographical Survey in Sub-Saharan Africa. Cross-cultural/ Cross-National Survey Research Session Martine Quaglia, INED, National Institute for Demographic Studies, France, [email protected] To conduct a survey in a foreign country requires some adaptations to the context in which the survey is to be made. One of the main questions has to do with how the research is going to fit into a national 373 or international demand for reliable data on the subject and therefore combines with those conducted by national scientists and institutions directly concerned and interested in the results. But furthermore, it has to deal with survey methodology , questionnaire design, and interviewing related to the sociological and cultural context in which the survey is taking place. What are the basic methodological requirements to be fulfilled in order to get in touch with the population we have to meet ? Adaptation also has to do with questionnaire organization and formulation. The questions have to fit into an environment which is not always familiar to the research teams. Although interviewers are generally recruited locally, they are not always aware of the researcher cultural point of view, and its consequences on the questionnaire formulation and organization. More over, when the interviewer doesn’t speak the language of the respondent, an interpret will assist the interviewer. Interviewers and interprets have to adjust their own perception of familiar and perfectly known concepts to the new environment. Words refer to many different concepts depending on how, where, or/and when they are used. Every social science researcher has once been confronted to the different meanings of a question, according to the country or environment in which the question is asked. No survey can be done (and no reliable data can be found) without an interviewer or an interpret who knows what he is talking about, that is, what the survey is looking for, and what the researcher and respondent linguistic and therefore cultural codes are. A close look at a training session for a biographical survey conducted in two villages of the Boo country in Mali (Hertrich 1996, Quaglia 2004) should underline some of the questions raised from a methodological, and ethical point of view. Keywords: cross-cultural survey methodology, cross-cultural interviewing, cross-cultural questionnaire design References Véronique Hertrich, Permanences et changements de l’Afrique rurale. Dynamiques familiales chez les Bwa du Mali , Paris, Centre français sur la population et le développement, XXII + 548 p. (Les études du CEPED, n°14) 1996. Martine Quaglia, Rapport de mission, Mali, décembre 2003-janvier 2004. « Mise à jour des enquêtes chez les Bwa du Mali », Paris, Ined, mars 2004. Challenges in Implementing a Multi-National Tobacco Use Survey Jutta Thornberry1, Michele Bloch2, Robert Goldenberg3, Norman Goco1, Don Jackson1 and Nancy Moss4 1 RTI International, USA, [email protected], [email protected], [email protected] 2 National Cancer Institute, USA, [email protected] 3 University of Alabama at Birmingham, USA, [email protected] 4 National Institute for Child Health and Human Development, USA, [email protected] Globally, tobacco use ranks among the leading causes of preventable, premature death. In most highincome nations, tobacco use is stable or decreasing, due to decades of research and program and policy interventions. Conversely, in developing nations, tobacco use is increasing, and may continue to do so among women who have traditionally had low rates of tobacco use. Understanding the knowledge, attitudes, and behaviors regarding tobacco use of pregnant women in the developing world would contribute to developing interventions aimed at (1) maintaining and supporting women’s and girls’ non-use of tobacco products and (2) helping users to quit. However, there are many challenges in collecting common data across different countries and cultures. These challenges are magnified when research capacity building is also a goal, and novice researchers must be taught standardized survey techniques. The Global Network for Women’s and Children’s Health Research was established by the National Institute of Child Health and Human Development (NICHD) to respond to the most critical existing and emerging health problems of women and children, globally. International, multidisciplinary teams of investigators, organized into research units, work collaboratively to answer scientific and/or public health questions to improve health, and to prevent premature disease and death among women and 374 children in developing countries. Currently, the Network is composed of 10 research units, located in Southern Asia, Central and South America, and Africa. The Tobacco Use Survey is the first common protocol to be implemented by the Network. For the survey, a convenience sample of 750 women, 18-46 years of age and at least 3 months pregnant will be identified in hospitals and clinics in each of 10 countries: the Congo, Zambia, Uruguay, Argentina, Brazil, Ecuador, Guatemala, India, Pakistan, and Tibet. A face-to-face interview, ranging from 20-40 minutes, will be administered. Each site will pretest the questionnaire before the instrument is finalized, train data collectors using standardized training materials, and collect the data using standardized oral procedures in the respondent’s language. This paper will discuss the following challenges of implementing a common survey in 10 countries: Developing a questionnaire with common, yet country-specific content. Conducting quality control measures to check language translations; Incorporating feedback from the pretest; Ensuring standardized training in data collection procedures; Conducting quality control measures to verify data collection procedures. Keywords: Cross-cultural methodology, multi-cultural survey design and implementation, multicultural survey instruments, quality control Attitude Scale Development in Countries which send Illegal Immigrants Kees van der Veer1, Reidar Ommundsen2, Hao Van Le3, Krum Krumov4 and Knud S. Larsen5 1 Vrije University Amsterdam, Faculty of Social Sciences, Department of Social Research Methodology, De Boelelaan 1081c, NL-1081 HV Amsterdam, The Netherlands, tel: +31 20 4446866 (office) or +31 6 29070050, [email protected] 2 University of Oslo, Norway 3 National Institute of Psychology,Vietnam 4 University of Sofia, Bulgaria 5 Oregon State University, USA This study reports on the utility of a Likert type scale measuring attitudes toward illegal immigrants in two samples from sending nations. The purpose of the current study is to investigate attitudes toward illegal immigration in net sending countries; i.e. in this case in samples from Bulgaria and Vietnam. What items of the illegal immigration scale (Ommundsen & Larsen, 1999) are the most and least significant contributors to attitudes in those net sending countries, based on item analysis? Previous research indicated that it is not possible to develop unidimensional Mokken scales which are robust and economical in samples from nations which have positive net illegal migration rates. The attitudes toward illegal immigrant scale was administered to 219 undergraduates from Sofia University in Bulgaria, and 179 undergraduates from Hanoi State University in Vietnam. Results yielded a scale with no gender differences, and acceptable alpha coefficients. Item analysis identified the most contributory and least contributory items, with considerable overlap in the two divergent samples. A varimax factor analysis produced a complex, but meaningful structure. Keywords: cross-cultural methodology; multi-cultural survey instruments; scale development Behavior Coding of Multilingual Survey Interviews: What Can We Tell by Systematically Listening? Gordon Willis1, Elaine Zahnd2, Sherm Edwards3, Stephanie Fry3, David Grant4, Nicole Lordi2 1 National Cancer Institute, NIH, USA 2 Public Health Institute, USA 3 Westat, USA 4 University of California at Los Angeles, Center for Health Policy Research, USA Assessment of the cross-cultural equivalence of survey questions poses a particularly vexing problem for multilingual surveys, due to the complexities that survey administrators face in evaluating the 375 quality of interviews conducted in languages they do not speak. For interviewer-administered questionnaires, investigators cannot simply listen to interviews (as through telephone monitoring), and may not be able to explain findings such as widely discrepant mean interview durations across languages. More generally, we normally lack systematic opportunities for judging how questionnaires function when translated. As a potential solution to this problem, the current project applies behavior coding, first developed by Charlie Cannell and colleagues at the University of Michigan, to the crosscultural, multilingual domain. This methodological study utilized the California Health Interview Survey (CHIS), a bi-annual telephone-administered survey conducted by the University of California at Los Angeles, in partnership with the Public Health Institute and the California Department of Health Services, and fielded by Westat. The project uses the 2003 CHIS as a vehicle for developing means for assessing the cross-cultural equivalence of survey questions through the application of behavior coding across English, Spanish, and Korean interviews. Trained, bilingual coders analyzed digitized recordings of over 300 fielded interviews. Coders listened for instances in which questions exhibited problems associated with the behavior of interviewers or respondents, including interviewer failure to read questions as written or to probe appropriately, and respondents’ requests for the question to be reread, requests for clarification, failure to provide codeable responses, expressions of uncertainty in their responses, extraneous commenting, and so on. In a general attempt to distinguish effects of culture from those due to language translation, coding was done for a set of English-language interviews of Hispanics and Koreans, and for another interview set conducted in either Spanish or Korean. All bilingual (Spanish-English and Korean-English) coders conducted coding in both their native language and in English, so that their English language coding performance could be assessed by the investigators. Through the quantitative aggregation of results of these behavior-coded interviews, the researchers will determine whether the targeted questions function as intended in each targeted language, and across cultures. Further, where overt evidence of difficulties arises, we will attempt to judge whether this is a general problem with the question, is due to mistranslation, is culturally specific, or might be due to some other confounding factor. Finally, we will discuss the relative merits of behavior coding as a means for evaluating cross-cultural equivalence, in a way that considers the operational as well as the scientific aspects of the endeavor. Keywords: Cross-cultural equivalence, behavior coding, translation Comparability of Attitude Measures towards Social Inequality. The Level of Equivalence in the ISSP 1999, its Implications on Further Analysis and the Presentation of Results Mag. Vlasta Zucha, Institute for Social Research and Analysis (SORA), A- 1150 Wien, Linke Wienzeile 246, [email protected] The International Social Survey Programme (ISSP) is a large-scale survey program conducted since 1983, which is covering special topics over time and different nations. The data is available for scientific research and offers extensive opportunities for the analysis of different social phenomena. In cross-national research high comparability of results is necessary and therefore researchers should pay attention to several dimensions of equivalence analysing this cross-national data. Yet testing of several forms and levels of equivalence is very time- and cost-intensive. Multiple methodologies must be applied to cross-national and cross-cultural data to assure comparability of measures and test comparability of data in the phase of data analysis. By analysing the ISSP 1999 (special topic: “social inequality”) the optimal level of equivalence for testing theories and understanding complex social realities will be addressed. The research focuses on the construct equivalence in cross-national research and examines whether the theoretical construct is universal for the countries under investigation or not. Depending on the purpose of an empirical study and research design, for most studies in social sciences structural equivalence might be adequate and sufficient. A secondary analysis of the ISSP-data from 1999 was conducted, in which equivalence of attitudes towards social inequality was tested for three countries – for Austria, Czech Republic and Germany. 376 After choosing the theoretical framework several techniques and statistical methods were