Party Voting in Turkey - Melissa J. Marschall

 Party Voting in Turkey: Explaining the Durability of AKP Support Melissa Marschall Professor -­‐ Political Science Rice University [email protected] Abdullah Aydogan Ph.D. Candidate -­‐ Political Science University of Houston [email protected] Alper Bulut Ph.D. Candidate -­‐ Political Science University of Houston [email protected] Abstract In this paper we examine the rise and consolidation of the Justice and Development Party (Adelet ve Kalinma Partisi, AKP) by analyzing the AKP’s success in local elections from 2004-­‐2014. To explain the durability of the AKP, we develop a theoretical framework that takes into account both the clientelistic tendencies of Turkish parties and the effects of neo-­‐
liberal economic policies on the incentives and opportunities for patronage politics. We posit that a new form of clientelism, made possible by housing projects sponsored by Turkey’s Mass Housing Administration (TOKİ), not only solved the resource problem for but also allowed it to achieve efficiency gains in its clientelistic efforts. We test this hypothesis and others related to programmatic and more traditional clientelistic citizen-­‐
party linkages using data for 900 municipal districts in Turkey. Our results show that even controlling for traditional forms of clientelism and other contextual variables, TOKI housing has a significant impact on the durability of the AKP ’s electoral success. Prepared for presentation at the Political Psychology Perspectives on Participation and Protest in the Middle East, Sponsored by International Society of Political Psychology (ISPP) March 6, 2015 Yasar University, Izmir, Turkey. The rise and consolidation of the Justice and Development Party (Adelet ve Kalinma Partisi, AKP) since 2002 is a fascinating development in Turkish politics. In a country where parties tend to be characterized by short tenures and weak ideological foundations, the longevity of the AKP is a significant accomplishment. Indeed, since competitive elections were first introduced in Turkey in 1946, no other party has maintained its majority status and managed to successfully govern the country for more than a decade. While the initial success of the AKP is relatively easy to explain, the question of the AKP’s durability presents a puzzle not only to students of Turkish politics, but to comparative politics scholars as well. Given both the political challenges Turkey faced in the late 1990s and the inability of then governing coalition (DSP-­‐ANAP-­‐MHP) to steer Turkey away from the worst economic crisis in its history, the country was poised for change. The AKP was ideally positioned to capitalize on this political opening. With its pro-­‐EU stance, machine-­‐
like grassroots organization, and strongholds of support among both the urban poor and Turkey’s growing Islamic business sector (Ulusoy 2014), the AKP emerged at the right time with the right platform, leadership, and political structure. Consequently, the AKP’s strong showing in its first electoral contest in 2002—winning 34 percent of the votes and becoming the first party to govern without a coalition since 1991—is relatively straightforward to understand. On the other hand, the AKP’s consolidation of electoral power and its ability to sustain and grow its political base over the past dozen years is more difficult to explain. For example, after several years of strong economic growth, the Turkish economy went into a major recession in 2008-­‐09. The deterioration in the global economic environment led to greater uncertainty for the Turkish economy. In conjunction with competitiveness losses 1 before the peak of the crisis, this led to sharp declines in business and consumer confidence, which in turn amplified the exceptionally large foreign demand shock. Households cut consumption abruptly, while companies reduced their investment and greatly depleted inventories (Rawdanowicz 2010). While the literature on economic voting would predict significant losses for the incumbent party in the wake of an economic downturn like this (Başlevent et al 2004, 2005, 2009), in the AKP’s case it appeared to cause only a modest decline in its vote shares in the 2009 local elections. In addition to the economic downturn, the AKP also appeared to weather the string of stormy events it faced in 2013-­‐2014. From the Gezi Park protests and burgeoning anti-­‐government social movement, to the corruption scandals among top AKP government officials, to the rift between the AKP and the Gulen Movement, to a series of foreign policy challenges, the AKP has come through relatively unscathed. Not only did it improve its performance in the March 2014 local elections, but it also won an unprecedented outright majority in the country’s first-­‐ever popular election of the President in August 2015. In this paper we take a first step at solving the puzzle of the AKP’s durability by empirically analyzing the party’s success in local elections from 2004-­‐2014. So far, the literature on Turkish politics has heavily relied on the assumption that party competition in Turkey is programmatic and that voters and parties are connected through programmatic linkage mechanisms. In this paper we adopt a different approach by focusing more intently on clientelistic linkages between Turkish parties and voters. In addition, using a political economy framework we argue that economic restructuring in the 1980s and the intensification of neo-­‐liberal economic policy that ensued forced Turkish parties to search for new sources of patronage to fuel clientelistic relationships. We argue that 2 Turkey’s Mass Housing Administration (TOKİ) has played key role not only in providing these resources, but also in radically transforming the nature of patronage politics. With enhanced authority under the AKP government, TOKI has dramatically increased its involvement in the housing sector—providing between five and ten percent of all of housing in Turkey and becoming directly involved in the construction of social housing. Working in partnership with TOKİ, local governments have also assumed considerable discretion over large-­‐scale infrastructural projects. We posit that TOKI housing and the array of associated construction-­‐related industries that provide jobs, contracts, and other selective incentives are subject to distributive politics under clientelism. We test this hypothesis and others related to programmatic citizen-­‐party linkages and more traditional forms of clientelism by analyzing mayoral elections across 900 municipal districts in Turkey. Our data combine municipal district-­‐level election results with demographic, socio-­‐economic, and other contextual measures, including district-­‐level measures of TOKI housing units and costs. We estimate ordered-­‐logit models to predict the durability of popular support for the AKP, which we measure as the number of times district municipalities voted for AKP mayoral candidates in the 2004, 2009, and 2014 elections. Our results show that even controlling for traditional forms of clientelism and other contextual variables, TOKI housing has a significant impact on the durability of the AKP ’s electoral success. We begin by reviewing the literature on programmatic and clientelistic party systems and underline the differences between emerging and developed democracies. In this regard, we discuss the Turkish party politics literature and argue that it implicitly assumes the context of a programmatic setting. From here, we discuss the evolution of 3 clientelistic policies in Turkey, TOKI’s role in the housing sector and our neo-­‐liberal clientelism framework. Finally, we present our empirical analyses and conclude with some broader implications and next steps for this research. Clientelistic versus Programmatic Competition and Turkish Party Politics The dynamics of party politics and electoral competition in emerging democracies are quite different from the Western context. Mainwaring and Torcal (2005:1) underline the distinct characteristics of these two settings and point out the need for a new approach: Most theoretical works on voters, parties, and party systems implicitly assume the
context of the advanced industrial democracies, especially of the United States
and Western Europe.. . .the literature on the advanced industrial democracies
cannot account for important characteristics of party systems in democracies and
semi-democracies in less-developed countries. Voters, parties, and party systems
in less-developed countries are qualitatively different from those of the advanced
industrial democracies.
One key difference between the developing states and the developed Western countries involves the linkage strategies parties employ to connect with their voters. The literature identifies two linkage mechanisms: programmatic and clientelistic. Parties and party systems are considered programmatic if they compete for votes by offering alternative policy proposals that they promise to enact once they assume office (Schlesinger 1984; Kitschelt 2000). These policy proposals are collective in nature and aim to benefit society more broadly. Since politicians cannot determine the individuals and groups who voted for them under programmatic competition, they cannot offer targeted benefits to current or prospective supporters. However as Kitschelt (2000: 849) points out, political parties may offer substitute ‘products’ to voters; namely, in the form of clientelism. Under clientelism, party-­‐voter linkages are based on exchange—namely the distribution of selective benefits to individual voters or groups of voters in exchange for their votes. 4 It is widely accepted in the literature that programmatic citizen-­‐politician linkage mechanisms and electoral competition based on programmatic policy proposals deliver better results for representation and are more acceptable to electoral constituencies than clientelistic linkage mechanisms (Kitschelt et al. 2010, 29). Similarly, scholars have argued that the combination of at least a moderately stable party system with programmatic alternatives creates the basis for a durable democracy with mass support (Huntington 1968; Linz and Stepan 1996). Clientelistic linkage mechanisms, on the other hand, are usually considered detrimental to democratic representation. According to Stokes (2007:605), political clientelism slows down economic development by expanding the public sector and creating an environment where politicians prefer voters to be economically dependent on them. By allowing some voters to vote based on policy preferences and others to use their vote in exchange for side payments, it distorts the equality of the ballot and undermines democracy (Stokes 2007:605). When clientelistic politics becomes the norm, non-­‐policy selective benefits replace policy-­‐based collective benefits, which in turn insulates policymakers from the priorities of the general public. As underlined by Mainwaring and Torcal (2005), many studies in the party politics and voting behavior literature implicitly assume the context of developed Western countries (Downs 1957; Adams et. al. 2004; 2006). The literature on voting behavior—
including research on proximity and directional voting, studies on left-­‐right semantics (Fuchs and Klingeman 1990), social cleavage (Lipset and Rokkan 1967) and party realignments in the developed Western countries (Inglehart 1990)—has relied heavily on the assumption that voters and parties are programmatically connected. However, the meaning of ideological labels may also be quite different in the context of emerging 5 democracies. In particular, the literature identifies two components of the ideological labels: symbolic and substantive (Kitschelt et. al. (2010). In this respect, the usefulness of the left-­‐right semantics depends on the interconnectedness between parties’ or politicians’ policy stances on key issues, and their ability to place their party on the left-­‐right dimension. (Kitschelt et al. 2010: 62). When this linkage is strong, the programmatic content of the left-­‐right semantic will be more useful. If a symbolic component dominates, the utility of left right labels for guiding programmatic vote choice in that country is diminished; the use of left-­‐right would signal only a party name. In party settings that are programmatically structured, the substantive component is dominant whereas in clientelistic settings, the symbolic component is prevalent. In this regard, especially in the context of emerging democracies, the literature needs to take into account the non-­‐programmatic and non-­‐ideological motivations that voters might have (Kitschelt 2000). In particular, voters might make decisions on the basis of clientelistic rather than programmatic goods, thereby choosing politicians that are ideologically further away from their positions, but who can significantly advance their material interests. The general tendency in the literature is important for the Turkish case since most of these issues are also reflected in studies of Turkish elections and party politics. For example, earlier studies typically focused on the role of cleavages, especially the center-­‐
periphery divide (Mardin 1973). In his influential study on Turkish politics, Mardin describes Turkish politics in terms of a cleavage between the secular Kemalist, nationalist state elite in the center and the religious, ethnic, and traditional conservative groups of the periphery. Ozbudun (1976), on the other hand, argued that the modernization process in Turkey increased the autonomous, instrumental, class-­‐based participation in Turkey. In 6 this regard, the traditional center-­‐periphery approach was replaced by a class-­‐based cleavage (Ozbudun 1976). Building on Mardin’s work, Kalaycioglu (1995) argued that the party preferences of the Turkish voters are determined by the historical center-­‐periphery divide and rapid social mobilization. His analysis showed that voters associated with the center usually voted for liberal or left wing parties, whereas voters in the periphery usually voted for conservative or right-­‐wing parties (Kalaycioglu 1994). Research on Turkish politics and elections has witnessed a transformation in recent years thanks to the availability of reliable survey data and increased number of scholars utilizing quantitative data and methodologies. In this respect, more recent studies on electoral behavior have focused on ethno-­‐religious and economic voting theories and largely used individual-­‐level survey data to test the relevant hypotheses regarding the vote choices in general parliamentary elections. A strand of these studies highlighted the importance of individuals’ evaluations of the economy (Başlevent et. al. 2004, 2005, 2009; Çarkoğlu 2008, 2009, 2012); several others stressed the significance of ethnic background (Ekmekçi 2011; Sarigil 2010); whereas another vein underlined the relationship between religiosity and vote choice (Cesur and Mocan 2013; Çarkoğlu 2012; Gidengil and Karakoç 2014). Contrary to the studies utilizing individual-­‐level data, several scholars used province-­‐level data to explore the dynamics of voting behavior, in particular retrospective voting (Akarca and Tansel 2006; Akarca and Başlevent 2010). While most political science research on Turkish elections has assumed existence of programmatic party-­‐citizen linkages, this does not mean that clientelistic party-­‐voter linkages have been completely overlooked. Indeed, work by White (2002), Eligür (2010) and Sayari (2007; 2011) has made a significant contribution to our understanding of the 7 organization and functioning of patron-­‐client relationships in Turkey, particularly among pro-­‐Islamic parties over the past several decades. In addition, a handful of more quantitative studies have also appeared more recently (Aytac 2013; Çarkoğlu and Aytac 2014; Kemahlioğlu 2012). However rather focusing on how the distribution and targeting of patronage helps maintain and/or enhance the AKP (or the incumbent party’s) electoral support, these studies attempt to explain how the AKP or intra-­‐party dynamics influences the distribution of patronage. In sum, while studies of Turkish electoral behavior provide important insights regarding Turkish politics, for the most part they have been unable to account for the consolidation and durability of the AKP over the past twelve years. In particular, the literature on economic voting and ethnic/religious cleavages cannot explain how the AKP has increased its electoral strength given economic downturns and stalled progress on the expansion of rights and freedoms for ethnic minorities. Similarly, programmatic explanations fail to explain why voters have not defected from the AKP since it has reversed its position or performance on a number of policy issues (corruption, democratization, civil liberties) that were presumed to be central to AKP supporters initially. At the same time, studies adopting a clientelistic approach to Turkish parties and elections have not only been largely descriptive and qualitative in nature, but as we discuss later, fail to explain how parties, including the AKP, have managed to generate the resources necessary to fuel patron-­‐client relationships in the current environment of globalization and neo-­‐liberal economic policy. We believe that the answer to our question regarding the AKP’s durability lies in both the party’s machine-­‐like organization and 8 historic clientelistic networks and in its strategic pursuit of new and more potent sources of patronage. The Emergence and Electoral Evolution of the AKP The AKP’s roots can be traced back to the National Outlook (Milli Görüş) Movement and the National Order Party (MNP, founded in 1970), which was considered the first Turkish party with clear Islamic credentials after the single party regime ended. Early on, the movement and the MNP experienced serious reactions from the hardcore secular bureaucrats and military generals. Consequently, the Constitutional Court banned not only the MNP, but also three other political parties affiliated with the movement (MSP, RP, FP). One important characteristic of the movement was its utilization of strong organizational networks that would, as Eligür (2010: 182) explains, “enable the party and its successors to frame the malfunctioning state in a manner that mobilized the electorate against the secular-­‐democratic state.” Operating largely under the radar given the media’s inattentiveness to the Islamist movement, in the mid-­‐1980s the movement concentrated its efforts on spreading its highly effective, hierarchical party structure across the country. The distinctive organizational structure featured a highly centralized and authoritarian decision-­‐making apparatus, several intermediary levels of party cadres, and a lower-­‐tier of foot-­‐soldiers who were rooted in villages and neighborhoods. At this lowest level were the dense networks of volunteers, many of whom were women and newly arrived migrants to Turkey’s rapidly growing urban centers. These foot-­‐soldiers went door-­‐to-­‐door spreading the party message of “Just Order” and providing material, emotional, and spiritual support in the form of food, financial assistance, solidarity, spiritual and emotional support (Atacan 2005; Eligür 2010; White 2002). 9 Unlike other parties in Turkey, the National Outlook-­‐affiliated parties engaged in face-­‐to-­‐face interactions with local residents, canvassing apartment buildings and neighborhoods year-­‐round rather than just before elections as most other parties tended to do. They effectively engaged in traditional machine-­‐style politics, supplying followers with more pre-­‐election incentives and post-­‐election services than their secular competitors. At the same time, these parties stridently promoted a pro-­‐Islamic, pro-­‐Ottoman (and thus anti-­‐secular and anti-­‐Western) message and culture (Akinci 1999). After the Constitutional Court banned its fourth pro-­‐Islamic party in 2001, the movement experienced a split: on one side were the “innovationists” (yenilikciler) lead by Erdoğan (former Istanbul mayor), and on the other the “traditionalists” (gelenekciler) who remained loyal to the core National Outlook ideas and principles. In 2001, Erdoğan and his reformist wing created the AKP. The AKP platform in its first electoral competition in 2002 sought to fight three “Ys”: yoksulluk (poverty), yolsuzluk (corruption), and yasaklar (bans on civil/individual liberties). In addition, the AKP staked out a staunchly pro-­‐Western agenda, strongly supporting Turkish integration into the global economy and full EU membership. This agenda helped the party make good on its pledge to be broad-­‐based, appealing to some secular citizens and large swaths of the middle class. The rebranding and reorientation of the AKP appeared to have struck a cord with the electorate. In the 2002 election and after only 15 months since its founding, the AKP managed to secure 34.3 percent of the popular vote, giving it two-­‐thirds of the seats in the Turkish Grand National Assembly. Even accounting for Turkey’s imbalanced election system and the high electoral threshold (10%), the 2002 election result was considered a 10 huge success for AKP. Indeed, the subsequent government formed by AKP was the first single party to govern Turkey since 1991 The initial electoral success of the AKP becomes relatively straightforward to understand after taking into account the historical context and the political and economic conditions leading up to the 2002 parliamentary elections. In particular, the 2001 economic crisis played a big role in the AKP’s success and the defeat of the previous ruling coalition parties, all of which failed to enter to the parliament in 2002. However, what came next would not necessarily have been easy to predict given the more enduring patterns of Turkish electoral politics and the Turkish party system. As Figure 1 shows, in the 2004 local elections, which also marked the AKP’s first appearance in mayoral and council elections in Turkey, it registered an extremely strong showing, improving its vote share by nearly ten percentage points and capturing 512 of the 914 (56%) district municipal mayoralties.1 [Figure 1 Here] This upward trajectory continued with the 2007 parliamentary elections, where the AKP came close to capturing a majority of the popular vote with 46.6 percent (341 seats) As noted in the introductory section, Turkey experienced a relatively severe economic recession in 2008-­‐09, and for the first time since it appeared on the electoral scene, the AKP witnessed a decline in its vote share. In addition, its share of mayoralties dropped by six percentage points (49.9%). However, the party rebounded in the 2011 parliamentary election, coming ever so close to the 50 percent threshold with 49.8 percent of the popular vote. The AKP also improved its showing in the 2014 local election, winning its largest 1 AKP was a majority winner in 177 of the 512 (35%) mayoral elections it won. 11 share of district mayoralties: 583 of 960 (60.7%). It finally crossed the majority threshold in the August 2014 presidential election, with Erdogan winning 51.8 percent of the popular vote in the first ever direct election of the Turkish President. AKP’s Neo-­‐Liberal Clientelistic Linkage Strategies and Significance of TOKI Housing The AKP did not simply benefit from the machine-­‐like organizational structure developed by the traditional National Outlook parties, but it also largely absorbed it as its own. Just as it was for the political bosses who governed American cities at the turn of the 20th Century, this structure was highly effective for mobilizing and rewarding party supporters. In the context of rapid urbanization, industrialization and infrastructural development, the pro-­‐Islamic parties relied heavily on patronage and clientelistic linkages. The party was expected to provide social and infrastructural services to individuals and communities in exchange for their votes. In urban settings, residents shared not only similar needs—for jobs, access to public services (utilities, roads, running water), information about how to navigate urban life—but also religion, religious brotherhoods, and himaye relations (White 2002). Their dense horizontal networks made the flow of information up the hierarchy much more efficient and the delivery of patronage resources and selective incentives more effective. As the tradition of clientelism dictated, voters expected the religious parties to provide social and infrastructural services in exchange for electoral support. However, as White (2002: 106) explains, the strong, horizontal networks of neighbors and the bonds of cultural and religious capital connecting them together enhanced voters’ ability to make parties more responsive and ultimately improved the performance and accountability of the pro-­‐Islamic parties. Sayari (2011) also underlines the success of pro-­‐Islamic parties in distributive politics and makes a similar point. 12 According to him, the success of these parties largely stems from their ability to replace vertical ties of clientelism with frequent face-­‐to-­‐face interaction between party workers and their neighbors (Sayari 2011: 13). This strong base of party workers, coupled with state resources, created a new network of clientelism, which played a major role in AKP’s success. Since the founding of the Turkish Republic in 1923, clientelistic linkages have been fostered by political and ideological principles that define the role of the state in more paternalistic terms. For instance, Kemalism, or the principle of statism, is one of the six core tenets of the Turkish Republic. It refers to the state’s responsibility to protect the economic well-­‐being of its citizens by both intervening in the economy and by developing social programs that provide a safety net for the needy. For much of its history, Turkey has had a very interventionist approach to its industrial development and economic growth. Nationalization and import-­‐substitution strategies contributed to a large public sector, which in turn provided the primary source of patronage for political parties. As Cizre-­‐
Sakallioglu and Yeldon (2000: 481) note, “In the historical context of a state-­‐dominated economy, politics was understood and defined as a strategy to build and sustain power by distributing material benefits generated by the state through clientelistic channels of interest mediation, with political parties and corporatist unions being two prominent organizations.” However these economic strategies proved unsustainable as inflation and balance of payments deficits led the Turkish economy into full-­‐blown crisis and eventual default on its debt servicing in the late 1970s. Under the auspices of the World Bank and the 13 International Monetary Fund, Turkey adopted a package of reforms that included privatization, liberalization, and greater export-­‐oriented development. Since these policies targeted the primary sources of patronage, political parties found it increasingly difficult to manage their resource problems. In particular, as public sector jobs became scarcer, the challenges of balancing electoral demand made party politics considerably less stable and contributed to the volatility that characterized Turkish politics in the 1990s. Under these conditions, it is logical to ask whether clientelism survived and if so, how? Kemahlioglu (2012) addresses the very question of how clientelism was affected by economic liberalization in Turkey. She argues that instead of shifting away from clientelistic politics toward responsible party government, Turkish parties continued to rely on jobs to maintain party support but adopted new, more effective strategies for distributing them. As she explains, rather than meting out jobs to ordinary citizens in exchange for their votes before elections, in the post-­‐neoliberal economic reform era, politicians began to allocate jobs in a direct and personal manner to active supporters already situated within the party structure. Kemahlioglu claims that this strategy increased the impact of the selective incentive since rewarding party activists with public sector jobs encouraged them to step up their efforts in ways the directly benefited the electoral prospects and career ambitions of the politicians awarding the jobs. In particular, after receiving a public sector job, party activists were more likely to increase their campaign activities, mobilize more voters, and secure more campaign contributions on behalf of the patron. Thus, rather than simply securing the votes of the client and his or her family 14 members, this strategy presumably produced a much greater return on the investment with hundreds or perhaps thousands of votes. Apart from Kemahlioglu’s work, other research suggests that the advent of neoliberal reforms led parties to substitute public sector jobs with other material rewards that were of smaller value (Stokes 2005). For example, appliances, coal, food baskets, and even transit tickets are all highly valued commodities that parties could still access and distribute to ordinary citizens in exchange for their support at the ballot box (Eligur 2009) In addition, other scholars have focused on governmental programs that distribute public benefits and social assistance to low-­‐income or otherwise disadvantage populations as potential sources of patronage. Conditional Cash Transfer (CCT) programs, like the one adopted in Mexico in 1997, are a case in point. In Turkey, a CCT program was first piloted in 2002 and then expanded nationwide in 2004. The program involves cash payments by the government to poor households, but requires recipients to make investments in human capital (health care, education) in return. In a recent study, Aytac (2013) investigates the distributive features of Turkey’s CCT program under the AKP government. He argues that because there is room for discretionary enrollment by the CCT program’s executive committee (Aytac 2013:9) the program is subject to patronage politics. Although we do not doubt that Turkish parties, the AKP included, have relied on the alternative sources of patronage identified in the literature to maintain clientelistic networks and relationships, we are skeptical that these resources have been sufficient. In the Turkish case, the more programmatic platform of the AKP, particularly in the initial years, may have reduced voters’ demands and expectations for selective incentives. However, there is no credible evidence that the AKP sought such a radical break with 15 entrenched Turkish political traditions. Indeed, the AKPs organizational structure and strong stores of cultural and religious capital made it ideally suited for patron-­‐client linkages. So, what else could be fueling clientelism under the AKP? We believe the answer lies in the massive housing and construction projects undertaken by the AKP government under the auspices of TOKI. As we briefly explain below, the unprecedented activity and investment in local housing markets by the central government were the direct result of the economic restructuring initiated decades earlier. However, what has been overlooked in the much of the existing literature on electoral politics in Turkey is the effect that these policies had on transforming the urban landscape and in turn, patron-­‐client linkages. Specifically, the commodification of land drove out the less productive and lower value land uses—squatter settlements, irregular housing, vacant land and green space—in favor of higher-­‐value, capital intensive commercial and residential developments. The AKP eventually capitalized on the opportunity this created. The Mass Housing Law of 1984 created the Mass Housing Fund, which became the Mass Housing Development Authority (TOKI) in 1990. While the Fund and TOKI were established to provide financing and services both to meet Turkey’s housing needs at a national level and to oversee an orderly process of urban development (Gunay, Koramaz & Ozuekren 2014), under the AKP it assumed a much more significant role in the direct provision of housing. Indeed, from the inception of the Housing Fund in 1984 to the election of the first AKP government in 2002, roughly 43,000 housing units were produced by TOKI (Karatepe 2013). In contrast, over 450,000 units of housing—more than 90 percent of all housing constructed by TOKI—were built between 2003-­‐2010 (TOKI 2012). The AKP increased the power and autonomy of TOKI itself, moving it directly under the 16 authority of the Prime Ministry, and amending the Public Management and Control Law (No. 5018) to exempt TOKI from the internal auditing conducted by the Turkish Court of Accounts (Karatepe 2013). With the enhanced authority under the AKP government, TOKI has assumed important responsibilities as regulator and investor. It plays an important role not only in the overall housing market, providing between five to ten percent of all of the housing need, but also in the construction of social housing. Since the mid-­‐2000s, TOKI and the district municipalities have had the authority to make decisions related to the size, type (including target income groups), and siting of housing projects. In other words, both central and local governments have considerable discretion over huge investment and infrastructural projects that directly affect the housing, employment, and amenities of local residents. Theoretical Framework and Hypotheses With this political economy framework in mind, the present research seeks to explore the relationship between clientelism and the electoral success of the AKP. In this study we focus on two types of clientelistic behavior that can be empirical measured. The first is the Conditional Cash Transfer (CCT program), which we believe taps more traditional sources of patronage. It is a benefit program targeting needy families and distributed through local agencies appointed by the central government. Based on extant research, we would expect a significant relationship between district-­‐level CCT expenditures and and support for the incumbent party. The second type of patronage, what we refer to as ‘neo-­‐liberal clientelism’ is TOKI housing projects. Taking into account the broader political economy in which Turkish parties and voters are situated, we hypothesize 17 that TOKI housing projects represent a more attractive and effective source of patronage than traditional forms, and thus will be a stronger predictor of AKP performance in mayoral elections. In Figure 2 we illustrate the causal pathway for both hypotheses. [Figure 2 Here] In the case of neo-­‐liberal clientelisim, decisions for the allocation of resources are given by the central government, here TOKİ administration officials. TOKİ contracts out the actual construction to firms, which have in many cases been identified as mostly pro-­‐
government companies (Kitschelt et al 2010). Since the construction industry represents a meaningful share of the labor market, TOKI housing projects involve a sizable number of jobs that can potentially be distributed to AKP supporters. In return for the award of TOKI contracts, construction companies have more incentive to support and finance party activities. This is true for subcontractors that provide a host of other construction related products and materials (e.g., furniture, iron-­‐steel, cement, banking, insurance, and plastic). In general, the investment in large housing projects by the central government acts as a significant stimulus on the local economy and the improvement in the local economic fortunes will directly influence the tax revenues of the local municipalities. Presumably, this will in turn increase the quality of municipality services and improve the mayor’s performance. Thus, ‘neo-­‐liberal’ clientelism operates on AKP electoral fortunes not only by providing housing to local residents in need, but also through the ‘multiplier effect’ that large construction projects have on the local economy and the performance of the local municipality. We therefore hypothesize that municipal districts with larger numbers of 18 TOKI units and greater TOKI expenditures (per capita) will be more steadfast in their electoral support for the AKP. Empirical Analysis To empirically investigate our research questions regarding the AKP’s durability in Turkish electoral politics, and more specifically, the extent to which citizen-­‐party linkages in the form of neo-­‐liberal and/or more traditional forms of clientelism are important explanatory factors in the AKPs ascendancy, we focus on municipal rather than national elections. First, we believe the impact of housing is more clearly observable in voting behavior during local elections. Second, TOKI investments in construction-­‐related industries are more directly connected to both the local economy (in the form of jobs, wages, standard of living, etc.) and the fiscal and governance capacity of local municipalities. Local elections therefore provide a more appropriate test of both programmatic and clientelistic linkages since issues of service delivery, jobs, and local economic conditions are salient for both parties and voters. Local elections in Turkey take place every five years at scheduled intervals. Our analysis includes all mayoral races in which the AKP has participated since its inception: 2004, 2009, and 2014. Turkish local elections also follow the “first-­‐past-­‐the-­‐post” method, meaning that the candidate who receives the plurality of the total valid votes wins the election. Thus, municipal mayors often win with less than a majority of the vote, sometimes with a vote share as low as 20 percent. 19 The unit of analysis in this study is the municipality or municipal district.2 In 2004 there were 915 districts, however a 2008 Law (No. 5747) added new metropolitan district municipalities, which increased the total number of municipal districts to 957 (Kapucu and Palabiyik 2008). Further, in 2012, Law No. 6360 added thirteen new metropolitan municipalities, and more district municipalities, putting the total number of municipal districts at 970. Our dataset includes all districts that consistently existed from 2004 to 2014 (N= 900 districts).3 The dependent variable is the durability of AKP support, which we operationalize as the number of times AKP won the district mayoral election over last three local elections (0 ≤ Y ≤ 3). Based on data from Turkstat (TUIK) and YSK (Higher Election Council), there are 137 municipal districts (15%) where the AKP never won the mayoral election and 221 districts (25%) where it won all the three elections (there are 249 and 293 districts where AKP won once and twice respectively). Figure 3 identifies municipal districts by the dependent variable spatially (see also Appendix A). [Figure 3 Here] To operationalize our measure of neo-­‐liberal clientelism we construct district-­‐level variables for the number of TOKİ housing units and the total cost of TOKI housing projects per 1,000 district residents.4 The total number of TOKI housing units was 432,079 and the 2 Given the tier system of local government in Turkey, there are several types of municipalities (see Kapucu and Palabiyik 2008). Our analysis is based on metropolitan district municipalities, provincial municipalities, and district municipalities (and excludes metropolitan municipalities, first-­‐lower tier municipalities, and subdistricts (townships)). 3 In seven of the new metropolitan municipalities, old Merkez (central) districts were only renamed. For example in Aydin, Merkez district was renamed Efeler. This does not create a problem regarding consistency in data. In six of the new metropolitan municipalities, on the other hand, old Merkez districts divided into two such as Yunusemre and Şehzadeler in Manisa. This creates a problem since we cannot match both districts to a single district in prior election years. Thus, we omit these Merkez districts as well as the new districts of these types of metropolitans. 4 https://www.toki.gov.tr/illere-­‐gore-­‐uygulamalar 20 total cost was 26,068,000,000 Turkish Lira (approximately $11,114,000,000 as of January 20, 2015).5 We include a measure of traditional clientelism as well. Borrowing from Aytac (2013), we operationalize this as the per capita average conditional cash transfer (CCT) funds appropriated to the municipal district between 2005 to 2008. There are six district-­‐level control variables in our model. The first is a dummy variable indicating whether the AKP won the mayoral election in 2004 with a majority of votes. We also include a second dummy variable to indicate if a party other than the AKP was a majority winner in the 2004 mayoral election. We expect districts where the AKP garnered majority support in its first mayoral competition would be more likely to continue voting for the AKP, whereas districts where voters endorsed other parties with a majority of votes would be less likely to vote AKP in future elections, compared to districts where no party won with a majority in 2004. We also include two variables to capture the level of development of each district: infant mortality rate and a composite development index (see Dinçer and Özaslan 2004), which measures development using 32 indicators (such as infant mortality, consumption of electricity, literacy rate, etc). Finally, we also include district-­‐level variables to measure urbanization (percent living in urban areas) and socio-­‐economic status of residents (percent of district residents with at least a high school diploma in 2013). Our model also includes additional control variables measured at the provincial-­‐
level since not all indicators of interest are available at the district level. These variables capture local economic conditions, religiosity, and key demographic changes.6 First, we 5 There are 142,338 more TOKİ units. Due to the recent redistricting and lack of information about the specific location of each project, we were not able use these units in this analysis. 6 All these province level data were obtained from Turkstat website. 21 include a variable measuring the average unemployment rate from 2008 to 2012. This measure tests for economic voting and is expected to negatively affect AKP durability. Second, we include a measure for the province’s average net migration from 2008 to 2013 (positive values represent more migration into than out of the province). Since provinces with increasing populations are expected to have greater need for housing, this measure may serve as a proxy for local housing supply. To control for religious cleavages in the population and the possibility that districts with more religious populations are more likely to vote AKP, we include a measure for the average number of mosques (from 2009 to 2013) per 1,000 citizens in the province. Finally, we include a region dummy for Southeast Anatolia as a control variable. Analysis and Results To test our hypotheses regarding the relative effects of programmatic versus clientelistic explanations of the AKP’s durability, we estimate our model using ordered logit. Given the possibility that districts within provinces share characteristics that may affect their propensity to vote AKP, we estimate robust standard errors, clustering on province.7 In Table 1 we report estimates from a series of models that gradually introduce our control variables. Regardless of specification, the results provide very convincing evidence that the key independent variables, TOKİ units and TOKİ costs, are statistically significant predictors of AKP durability. In districts where the TOKİ investments are greater, the likelihood of AKP winning more elections is higher. That these effects obtain even after controlling for a host of variables identified in the literature as key explanations 7 In some provinces (e.g.,Kayseri, Rize, Kahramanmaras, Istanbul, Bayburt, and Kocaeli) AKP won more than half of the districts for 3 times, whereas in other provinces (e.g., Diyarbakır, Mersin, İzmir, Muğla, Mardin, Hakkari, Kırklareli, Şırnak) AKP never won in more than half of the districts (see Table 1). 22 for the AKP’s electoral success provides support for our hypothesis regarding the role of the neo-­‐liberal clientelism mechanism. On the other hand, when it comes to traditional sources of clientelistic support, our empirical results are much less encouraging. The coefficient on the conditional cash transfer (CCT) variable is statistically insignificant across the two models where it is included, suggesting that the durability of AKP support is not related to the distribution of CCT spending across districts. [Table 1 Here] To illustrate the substantive effects of TOKI investments, Figure 4 displays predicted probabilities for the number of AKP victories based on the total cost of TOKİ projects per 1,000 district residents (in millions of Lira). The first panel shows that the probability that AKP never wins is about 12 percent when TOKİ makes no investments in the district. When TOKİ investments equal 8 million Lira, the probability of never winning decreases to approximately 2 percent. On the other hand, the bottom right panel shows that the probability of winning three elections is about 20 percent if there is no TOKİ investment in the district. This probability climbs to over 80 percent if TOKİ spends 8 million Lira. The confidence interval of the predicted probability for the lowest level of spending does not overlap with the confidence interval of the highest level of spending in these two graphs, indicating that these effects are statistically significant. [Figure 4 Here] A similar pattern emerges when we examine the substantive effects of TOKİ housing units. As Figure 5 shows, the predicted probability of the AKP winning three times is statistically and substantively higher in districts where TOKİ constructed more housing 23 units. Overall, the effects of TOKİ investments in the form of housing units and expenditures are both robust and striking. [Figure 5 Here] Turning now to the effects of the variables that tap aspects of programmatic linkages we find more mixed results. On the one hand, the measure of religiosity—the average number of mosques per 1,000 citizens in the province—is not statistically significant. This is contrary to religious based voting theories, though tests of these theories have relied on individual-­‐level data (e.g. Gidengil and Karakoç 2014). Similarly, our measure of regional, ethnic difference (the dummy variable for the southeastern Anatolia region), is also statistically insignificant. Given the concentration of Kurdish voters in this region, we would have expected a negative relationship between region and AKP durability. However, in this region the AKP won more than 34 percent of the districts in at least two of the last three elections. Thus, even though there is a strong popular support for the Kurdish parties (BDP and others), the AKP is not totally absent from the competition. On the other hand, our results do show some significant effects for urbanization. In districts with larger shares of urban residents, the AKP is more likely to achieve multiple electoral successes. However, these effects are weaker and substantively rather small. When it comes to our measure of economic voting we do find strong and consistent effects. Across all of our specifications, the coefficient on the unemployment variable is negatively signed and statistically significant, indicating that in districts with higher levels of unemployment, the electoral prospects of the AKP are reduced. In Figure 6 we illustrate the substantive impact of unemployment on the AKP’s durability. The top-­‐left panel shows that the predicted probability of the AKP never winning is about 5 percent when the 24 average unemployment rate is zero. However, this probability increases to 26 percent if the unemployment rate increases to 20 percent. On the other hand, the probability of winning the last three elections is 40 percent when the unemployment rate is zero, whereas it is about 9 percent if the unemployment rate increases to 20 percent (bottom-­‐right panel). Again the confidence intervals at the extreme points do not overlap, which indicates the statistical significance of the impact. [Figure 6 Here] Beyond these effects, our empirical results suggest that other covariates also help explain variation in AKP durability across municipal districts. In particular, we find the majority winning status of both the AKP and other parties in the 2004 mayoral election to be strong and consistent predictors of the durability of AKP support. Specifically, if the AKP won the district with a majority in 2004, it was more likely to win subsequent elections compared to districts where no party won with a majority in 2004. The opposite is true for districts where another party achieved a majority win in 2004. Here the AKP was significantly more likely to lose in subsequent elections as well. Thus, all things equal, the strength of party support is a good predictor of the party’s future electoral success. We also find strong negative effects for education, which was operationalized as the percent of high school graduates and above. In districts with more educated residents, the AKP less likely to achieve successive electoral victories. In addition, the composite index for district-­‐level development is positively associated with multiple AKP victories. This indicates that in districts with higher levels of development, the AKP is more likely to have won all three mayoral contests. 25 While the results provide compelling evidence linking TOKİ housing to AKP durability, it is fair to ask why this relationship implies clientelistic linkages rather than government responsiveness. In other words, why don’t greater district-­‐level investments in TOKİ housing simply reflect greater housing needs? The counter argument would be that if these housing needs are in met with government supplied housing, district voters are simply rewarding the ruling party for its responsiveness by voting for the AKP in subsequent elections. In fact, our model includes variables that partly control for housing need/demand—net migration, urbanization, level of district development. Based on policy responsiveness, TOKİ housing would be more concentrated in districts with more disadvantaged populations, increasing populations, and perhaps, larger concentrations of urban residents. If TOKİ housing were perfectly distributed based only on these factors, presumably our measures of TOKİ investments would be highly correlated with these measures of housing demand. Under conditions of severe multicollinearity, our models would be unable to distinguish the independent effects of the TOKİ housing on AKP durability. The same would be true of the other independent variables. However, our models find statistically significant effects, suggesting that at least part of the relationship between TOKİ housing and AKP support is independent of housing need and demand. Another way to look at this might be to see if the correlation between these variables is relatively similar in districts where AKP always or mostly wins (AKP wins ≥ 2) compared to districts where AKP never or almost never wins (AKP wins ≤ 1). In Table 2 we present partial correlations between the number of TOKİ units per 1,000 residents and the variables tapping housing need/demand: net migration, development, urbanization, and the region dummy (Southeast). In the first column we report partial correlations for 26 districts where AKP won at most once, and the second column includes partial correlation coefficients for districts where AKP won at least twice. [Table 2 Here] In districts where the AKP typically loses, the number of TOKI units per 1,000 residents is positively and significantly correlated with net migration. Specifically, controlling for level of development, urbanization and region, district population growth and TOKI housing units are correlated at 0.14. This correlation is positive as we would expect, since increasing population suggests increasing housing need and demand and thus more housing supply on the part of TOKI. No other indicators are significantly correlated with TOKI housing units in districts where AKP typically loses. When it comes to districts where AKP typically wins, net migration is negatively and significantly correlated with TOKI housing units per 1,000 residents, indicating that TOKI housing is more prevalent in districts that are losing population. This correlation is counter to expectation and suggests that factors beyond housing need and demand drive TOKI housing investments in these districts. TOKI housing is also significantly and negatively correlated with the Southeast region in mostly winning districts. Since the partial correlation controls for level of development, SE region here is mostly a proxy for the Kurdish presence in this region. This implies that in districts with similar levels of housing need/demand, those with large Kurdish populations will have significantly less TOKI housing compared to those with smaller Kurdish populations. Finally, in districts where AKP mostly wins, TOKI housing is also positively and significantly associated with urbanization. 27 Overall, the partial correlations presented in Table 2 provide further evidence in support of our hypothesis. In particular, in districts where AKP mostly wins, it appears that with the exception of urbanization, factors other than those associated housing need and demand play a significant role in TOKI housing investments. This pattern is more consistent with clientelism than government responsiveness. The opposite is true in districts where AKP mostly loses. Here it appears that TOKI housing investments are correlated mostly with population growth—a key indicator of housing need/demand. In one final test, we compare difference of means (t-­‐tests) for all three measures of clientelism (TOKI housing units, TOKI costs, and CCT costs) across these two groups (AKP win ≤1 vs AKP win ≥ 2). As reported in Table 3, we find significantly higher mean TOKI units and costs in districts where AKP won at least twice compared to those it won at most once. On the other hand, we found the opposite pattern for CCT spending: significantly higher mean spending per 1,000 residents in districts where AKP mostly loses compared to districts where AKP mostly wins. This analysis does not control for other factors, but does show that as a group, AKP winning districts receive significant highly investments in TOKI housing—a pattern that is consistent with clientelistic linkages. [Table 3 Here] Conclusions and Implications In this study we took a first step at solving the puzzle of the AKP’s durability by focusing on the local elections between 2004 and 2014. The results of our analysis support our theory of neo-­‐liberal clientelistic linkages and their effects on the electoral success of the AKP in Turkey. TOKİ district-­‐level investments in the form of housing units and associated expenditures are a significant determinant of the number of times the AKP won 28 mayoral elections from 2004-­‐2014. This finding holds up even after controlling for many other covariates, including the majority winning status of AKP in 2004. Our results also show that district-­‐level CCT spending is not significantly related to AKP’s electoral success, at least in mayoral elections. The effect of traditional clientelistic resources seems to be limited whereas neo-­‐
liberal clientelism has a significant impact on AKP’s success. Together these results indicate that clientelistic linkage mechanisms are crucial in explaining party voting and especially incumbent party support in Turkey. In addition our analysis confirms that the nature of clientelism in Turkish politics has evolved and the sources of patronage have changed. In the current environment of neo-­‐liberal economic policies and land-­‐use priorities, it appears that TOKİ housing projects and the multiplier effects these projects have on construction-­‐
related industries located within municipal districts fuel patronage networks and also play a key role in explaining how the AKP has so successfully maintained its hegemonic position in the electorate. Our findings have several implications. First, since clientelistic policies are considered to undermine democracy (Stokes 2007), the continuation and perhaps intensification of clientelism in Turkish politics would seem to have negative effects on democratic consolidation. Our findings have also broader implications for the for the democratization literature, especially for the democratic movements in the Middle East. For instance, many observers considered AKP’s initial years a great success and consequently, the party was frequently held up as a model for other Islamic parties in the region. However, as AKP began consolidating its power, it appears to have moved towards 29 more clientelistic policies. Future research should investigate whether other Islamic parties such as the Nahda in Tunisia might following a similar trajectory. 30 Works Cited Akarca, Ali T., and Aysit Tansel. 2006. "Economic performance and political outcomes: An analysis of the Turkish parliamentary and local election results between 1950 and 2004." Public Choice 129 (1-­‐2): 77-­‐105. Akarca, A. T., & Başlevent, C. (2011). Persistence in regional voting patterns in Turkey during a period of major political realignment. European Urban and Regional Studies, 18(2), 184-­‐202. Akinci, Ugur. 1999. “The Welfare Party's Municipal Track Record: Evaluating Islamist Municipal Activism in Turkey.” Middle East Journal 53 (1): 75-­‐94 Aytaç, S.E. (2014), ‘Distributive politics in a multiparty system: the conditional cash transfer program in Turkey’, Comparative Political Studies 47(9): 1211–1237. Atacan, Fulya, 2005. “Explaining Religious Politics at the Crossroad: AKP-­‐SP.” Turkish Studies 6 (2): 187–199. Başlevent, Cem, Hasan Kirmanoğlu, and Burhan Şenatalar. 2009. "Party preferences and economic voting in Turkey (now that the crisis is over)." Party Politics 15(3): 377-­‐
391. Başlevent, Cem, Hasan Kirmanoğlu, and Burhan Şenatalar. 2005 "Empirical investigation of party preferences and economic voting in Turkey." European Journal of Political Research 44 (4): 547-­‐562. Baslevent, Cem, Hasan Kirmanoglu, and Burhan Senatalar. 2004. "Voter profiles and fragmentation in the Turkish party system." Party Politics 10 (3): 307-­‐324. Cesur, Resul Naci H. Mocan. 2013. “Does Secular Education Impact Religiosity, Electoral Participation And The Propensity To Vote For Islamic Parties? Evidence From An Education Reform In A Muslim Country.” National Bureau of Economic Research, Working Paper 19769. http://www.nber.org/papers/w19769 Cizre-­‐Sakallioglu, U., & Yeldan, E. (2000). Politics, society and financial liberalization: Turkey in the 1990s. Development and Change, 31(2), 481-­‐508.Çarkoğlu, A. (2008). Ideology or economic pragmatism? Profiling Turkish voters in 2007. Turkish Studies, 9(2), 317-­‐344. Çarkoğlu, A. (2012). Economic evaluations vs. ideology: Diagnosing the sources of electoral change in Turkey, 2002–2011. Electoral Studies, 31(3), 513-­‐521. Çarkoğlu, Ali. 2009. "Turkey’s local elections of 2009: Winners and losers." Insight Turkey 11.2 (2009): 1-­‐18. Çarkoğlu, Ali. 2013. “Economic Evaluations vs Ideology: Diagnosing the sources of electoral change in Turkey, 2002-­‐2011. Electoral Studies 31: 513-­‐21. Çarkoğlu, Ali and S. Erdem Aytaç. 2014. Who gets targeted for vote-­‐buying? Evidence from an augmented list experiment in Turkey. European Political Science Review Dinçer, B., & Özaslan, M. (2004). İlçelerin sosyo-­‐ekonomik gelişmişlik sıralaması araştırması [Socioeconomic development ranking of districts]. Ankara, Turkey: T.C. Başbakanlık Devlet Planlama Teşkilatı. Ekmekci, Faruk. 2011. “Understanding Kurdish ethno-­‐nationalism in Turkey: socio-­‐
economy, religion, and politics.” Ethnic and Racial Studies, 34(9), 1608-­‐1617. Eligür, Banu. 2009. “Turkey’s March 2009 Local Elections.” Turkish Studies 10 (3): 469-­‐96. 31 Eligür, Banu. 2010. The Mobilization of Political Islam in Turkey. NY: Cambridge University Press. Erdogan , Mustafa. 1999. “Islam in Turkish politics: Turkey's quest for democracy without Islam”, Critical Middle Eastern Studies, 8(15): 25-­‐49. Fuchs, Dieter and Hans-­‐Dieter Klingemann. 1990. “The Left-­‐Right Schema”. In M. Kent Jennings, Jan van Deth, et al., Continuities in Political Action: A Longitudinal Study of Political Orientations in Three Western Democracies. Berlin; New York: Walter de Gruyter Gidengil, E., & Karakoç, E. (2014). Which matters more in the electoral success of Islamist (successor) parties–religion or performance? The Turkish case. Party Politics, 1354068814549341. Gunay, Z., Koramaz, T., & Ozuekren, A. S. (2014). From squatter upgrading to large-­‐scale renewal programmes: housing renewal in Turkey. Renewing Europe's housing, 215.Huntington, Samuel. 1968. Political Order in Changing Societies. New Haven: Yale University Press. Inglehart, R. (1990). Culture shift in advanced industrial society. Princeton University Press. Kalaycioğlu, E. (1994). Elections and Party Preferences in Turkey Changes and Continuities in the 1990s. Comparative Political Studies, 27(3), 402-­‐424. Kapucu, Naim and Hamit Palabiyik. 2008. Turkish Public Administration. Ankara: International Strategic Research Organization. Karatepe, Ismail D. 2013. Islamists, State and Bourgeoisie: The Construction Industry in Turkey World Economics Association Conferences, Neoliberalism in Turkey: A Balance Sheet of Three Decades Kemahlioğlu, Ö. (2012). Agents Or Bosses?: Patronage and Intra-­‐party Politics in Argentina and Turkey. ECPR Press. Kitschelt, Herbert. 2000. “Linkages Between Citizens and Politicians in Democratic Polities,” Comparative Political Studies, 33: 845-­‐79. Kitschelt, Herbert, Hawkins, Kirk, Luna, Juan Pablo, Rosas, Guillermo, and Zechmeister, Elizabeth (2010) Latin American Party Systems, February, 2010, Cambridge University Press. Kitschelt, Herbert and Matthew Singer (2011). “Do Everything (DoE) Parties: When Can Politicians Combine Clientelistic and Programmatic Appeals?”, (Working Paper). Retrieve: https://web.duke.edu/democracy/papers/3.3.Kitschelt_Singer.pdf Linz, Juan J. and Alfred Stepan. 1996. Problems of Democratic Transition and Consolidation: Southern Europe, South America, and Post-­‐Communist Europe ,Johns Hopkins University Press. Lipset, Seymour and Stein Rokkan. 1967. Party systems and voter alignments: Cross-­‐
national perspectives. Toronto: The Free Press. Mainwaring, Scott, and Mariano Torcal. 2005. Party System Institutionalization and Party System Theory After the Third Wave of Democratization. Working Paper 319, Kellogg Institutefor International Studies. Mardin, Ş. (1973). Center-­‐periphery relations: A key to Turkish politics?. Daedalus, 169-­‐
190. Özbudun E (1976) Social Change and Political Participation in Turkey. Princeton, NJ: Princeton University Press. 32 Pope, Nicole and Hugh Pope. 1997. Turkey Unveiled: Ataturk and After. London: John Murray. Rawdanowicz, Łukasz. 2010. “The 2008-­‐09 Crisis in Turkey: Performance, Policy Responses and Challenges for Sustaining the Recovery,” OECD Economics Department Working Papers, No. 819, OECD Publishing. http://dx.doi.org/10.1787/5km36j7d320s-­‐en Sarigil, Zeki. 2010. “Curbing Kurdish ethno-­‐nationalism in Turkey: an empirical assessment of pro-­‐Islamic and socio-­‐economic approaches.” Ethnic and Racial Studies, 33(3), 533-­‐553. Sayari, Sabri. 2007. “Towards a New Turkish Party System?” Turkish Studies 8 (2): 197-­‐210 Sayarı, Sabri. 2011. “Clientelism and Patronage in Turkish Politics and Society.” In The Post Modern Abyss and the New Politics of Islam: Assabiyah, Revisited Essays in Honor of Şerif Mardin, eds. Binnaz Toprak and Faruk Birtek. Istanbul Bilgi University Press 355. Schlesinger, Joseph (1984) “On the Theory of Party Organization”, Journal of Politics, 46: 369–400. Stokes, S. C. (2007). Political clientelism. In C. Boix & S. C. Stokes (Eds.), Oxford handbook of comparative politics (pp. 604-­‐627). New York, NY: Oxford University Press. TOKI. 2012. Due Diligence Document. Republic of Turkey, Prime Ministry. Housing Development Administration. Ulusoy, K. (2014). Elections and Regime Change in Turkey: Tenacious Rise of Political Islam. Elections and Democratization in the Middle East: The Tenacious Search for Freedom, Justice, and Dignity, 179. White, Jenny. 2002. Islamist Mobilization in Turkey: A study in Vernacular Politics. Seattle: University of Washington Press. Yavuz, M. H. (1997). Political Islam and the welfare (Refah) party in Turkey. Comparative Politics, 63-­‐82. 33 Tables and Figures Figure 1: Turkish Elections, 2002-­‐2014 60 50 ** Vote Share 40 30 20 10 0 2002 2004* 2007 AKP CHP 2009* MHP 2011 3/2014* 8/2014** DEHAP/DTP/BDP/HDP Note: * indicates Local elections; ** indicates Presidential Election with İhsanoğlu, as Independent or CHP/MHP coalition candidate. 34 Figure 2: Traditional Clientelism versus Neo-­‐Liberal Clientelism Mechanisms 35 Figure 3: Number of AKP Wins in Mayoral Elections, 2004-­‐2014 36 Figure 4: Predicted Probabilities for TOKİ Costs per 1000 citizens (in million Liras) 37 Figure 5: Predicted Probabilities for Number of TOKİ Units for 1000 Citizens 38 Figure 6: Predicted Probabilities for Unemployment 39 Table 1: Multivariate Ordered Probit Results: Durability of AKP Support 40 Table 2: Partial Correlations Between TOKI Housing Units and Indicators of Housing Need/Demand: Low vs. High AKP Winning Districts AKP Wins ≤1 AKP Wins ≥2 Net Migration 0.142*** -­‐0.086* Development -­‐0.045 -­‐0.030 Urbanization 0.039 0.170*** SE Region -­‐0.066 -­‐0.106** N 365 491 ***p ≤ .01 **p ≤ .05, *p ≤ .10 Table 3: Difference of Means Tests for Clientelism Indicators: Low vs. High AKP Winning Districts AKP Wins ≤1 AKP Wins ≥2 t-­‐value Mean TOKI Units per 1,000 4.52 7.66 -­‐3.86*** Mean TOKI costs per 1,000 0.24 .400 -­‐3.17*** Mean CCT costs per capita 36.96 26.09 3.65*** ***p ≤ .01 **p ≤ .05, *p ≤ .10 41 Appendix A: Summary Statistics for Dependent Variable at Province Level Province AKP win=0 AKP win=1 AKP win=3 AKP win=3 Total districts Adana 1 6 5 1 13 Adiyaman 0 3 5 1 9 Afyonkarahisar 1 5 9 3 18 Agri 3 2 1 2 8 Aksaray 0 2 3 2 7 Amasya 1 1 3 2 7 Ankara 1 7 6 10 24 Antalya 4 6 2 1 13 Ardahan 1 3 1 1 6 Artvin 1 5 2 0 8 Aydin 8 6 3 0 17 Balikesir 1 5 7 5 18 Bartin 1 2 1 0 4 Batman 2 1 3 0 6 Bayburt 0 0 1 2 3 Bilecik 1 3 2 2 8 Bingol 0 3 3 2 8 Bitlis 0 2 5 0 7 Bolu 0 2 3 4 9 Burdur 0 4 4 3 11 Bursa 1 1 7 8 17 Canakkale 3 4 3 2 12 Cankiri 1 3 6 2 12 Corum 1 3 5 5 14 Denizli 0 8 7 4 19 Diyarbakir 7 6 0 0 13 Duzce 0 0 5 3 8 Edirne 4 4 1 0 9 Elazig 1 5 5 0 11 Erzincan 0 3 2 4 9 Erzurum 1 4 6 6 17 Eskisehir 0 1 10 1 12 Gaziantep 0 0 4 4 8 Giresun 0 4 9 3 16 Gumushane 0 3 2 1 6 Hakkari 3 1 0 0 4 Hatay 1 2 6 2 11 Igdir 1 1 2 0 4 Isparta 1 7 3 2 13 Istanbul 8 4 0 19 31 Izmir 16 10 2 0 28 Kahramanmaras 1 2 1 5 9 Karabuk 0 1 4 1 6 Karaman 0 3 2 1 6 42 Kars 1 3 3 1 8 Kastamonu 2 2 13 3 20 Kayseri 0 3 4 9 16 Kilis 0 2 1 1 4 Kirikkale 0 1 5 3 9 Kirklareli 6 1 0 1 8 Kirsehir 0 3 3 1 7 Kocaeli 0 1 1 4 6 Konya 0 4 15 12 31 Kutahya 1 0 8 4 13 Malatya 2 3 1 6 12 Manisa 3 2 6 4 15 Mardin 7 2 0 1 10 Mersin 5 3 1 0 9 Mugla 8 3 0 1 12 Mus 3 2 0 1 6 Nevsehir 1 2 3 2 8 Nigde 1 1 2 2 6 Ordu 0 4 6 8 18 Osmaniye 1 4 2 0 7 Rize 0 1 4 7 12 Sakarya 0 0 6 6 12 Samsun 0 2 7 5 14 Sanliurfa 3 5 2 0 10 Siirt 0 4 1 1 6 Sinop 1 5 1 2 9 Sirnak 6 0 1 0 7 Sivas 1 4 8 4 17 Tekirdag 4 4 1 0 9 Tokat 0 3 6 3 12 Trabzon 0 5 6 6 17 Tunceli 4 3 1 0 8 Usak 0 4 2 0 6 Van 1 7 2 1 11 Yalova 0 2 1 3 6 Yozgat 0 3 7 4 14 Zonguldak 0 3 2 1 6 137 249 293 221 900 Total 43