Desastres naturales

Transcription

Desastres naturales

Disaster Risk and Poverty in Latin America:
The Peruvian Case Study
First Draft
Group of Analysis for Development-GRADE
Lima, 21th October 2008
CONTENT INDEX
1
2
3
Introduction............................................................................................................................. 5
Objectives and Methodology .................................................................................................. 6
Overview of Natural Hazards in Peru..................................................................................... 7
3.1
All the events: hydro-met and geological, intensive and extensive................................ 9
3.1.1 Number of events reported.......................................................................................... 9
3.1.2 Number of deaths caused by natural hazards............................................................ 12
3.1.3 Number of houses destroyed caused by natural hazards .......................................... 14
3.2
“Fenómeno del Niño”: frequency and magnitude ........................................................ 15
3.2.1 Number of events caused by ‘El Niño’..................................................................... 16
3.2.2 Magnitude of ‘El Niño’ in terms of deaths and houses destroyed............................ 18
3.3
Intensive Events ............................................................................................................ 19
3.3.1 Number of intensive events ...................................................................................... 19
3.3.2 Magnitude of the intensive events ............................................................................ 20
3.4
Extensive events............................................................................................................ 20
3.4.1 Comparative analysis of Hydro – meteorological and Geological events................ 21
3.4.1.1 By number of events reported........................................................................... 21
3.4.1.2 By number of deaths ......................................................................................... 21
3.4.1.3 By number of houses destroyed........................................................................ 22
3.4.1.4 Frequency of Hydro – meteorological events................................................... 22
3.4.1.5 Frequency of Geological events........................................................................ 23
3.4.2 Hydro – extensive events .......................................................................................... 23
3.4.2.1 Number of events.............................................................................................. 24
3.4.2.2 Number of deaths.............................................................................................. 25
3.4.2.3 Number of houses destroyed............................................................................. 26
3.4.3 Geo Extensive events................................................................................................ 27
3.4.3.1 Number of events.............................................................................................. 28
3.4.3.2 Number of deaths.............................................................................................. 29
3.4.3.3 Number of houses destroyed............................................................................. 30
4
Natural Hazards and the characteristics of affected areas .................................................... 31
4.1
Analysis of Bias in the Natural Hazards Reports of the DesInventar database ............ 31
4.1.1 Missing districts in the DesInventar Database.......................................................... 31
4.1.2 Comparison of Reports by district´s geo-political condition.................................... 32
4.2
Correlations between initial socio-economic conditions and Disaster´s loses at the
District level.............................................................................................................................. 35
5
The impact of Natural Hazards on Poverty indicators.......................................................... 39
5.1
Analysis at the District level ......................................................................................... 39
5.1.1 The effect of Natural Hazards on Poverty rates: Provincial Poverty Maps (19932005) 39
5.1.2 The effect of Natural Hazards on Children´s Malnutrition at the District level (19992005) 39
5.1.3 The effect of Natural Hazards on the Value of Agricultural Production at the District
level (1997-2006, yearly panel) ............................................................................................ 39
5.2
Analysis at the Household level.................................................................................... 40
5.2.1 Data and Descriptive Statistics ................................................................................. 40
2
5.2.2 Poverty matrix........................................................................................................... 42
5.2.3 Poverty transitions, multivariate household regressions........................................... 45
5.2.4 Change in per capita consumption............................................................................ 52
5.2.5 Analysis at the bottom of the distribution................................................................. 52
6
Conclusions........................................................................................................................... 54
References..................................................................................................................................... 55
Appendix 1: Descriptive statistics (Unbalance panel) ................................................................. 57
Appendix 2: Descriptive statistics (Balance panel) ..................................................................... 58
Appendix 3: Quantile regression, Dependent variable: (log) Monthly per capita consumption
(2006). Controls: 2002 .................................................................................................................. 59
GRAPHICS INDEX
Graph 1: Distribución temporal del número de eventos (All) ...................................................... 10
Graph 2: Distribución geográfica de la ocurrencia de eventos naturales (All)............................. 11
Graph 3: Distribución temporal del número de muertos por la ocurrencia de desastres naturales
(All) .......................................................................................................................................... 12
Graph 4: Distribución geográfica del número de muertos por la ocurrencia de desastres naturales
(All) .......................................................................................................................................... 13
Graph 5: Distribución temporal del número de casas destruidas por la ocurrencia de desastres
naturales (All)........................................................................................................................... 14
Graph 6: Distribución geográfica del número de hogares destruidos por la ocurrencia de
desastres naturales (All) ........................................................................................................... 15
Graph 7: Distribución temporal del número de eventos ocurridos por el fenómeno del niño...... 16
Graph 8: Comparación por tipo de evento del número de eventos ocurridos por el Fenómeno del
Niño.......................................................................................................................................... 17
Graph 9: Distribución geográfica del número de eventos ocurridos por el Fenómeno del Niño . 17
Graph 10: Distribución geográfica del número de personas muertas por el Fenómeno del Niño 18
Graph 11: Distribución geográfica del número de viviendas destruidas por el Fenómeno del Niño
.................................................................................................................................................. 18
Graph 12: Distribución temporal del número de desastres intensivos ocurridos (all intensive) .. 19
Graph 13: Número de eventos ocurridos, según event tipe .......................................................... 21
Graph 14: Número de muertes, según event tipe.......................................................................... 21
Graph 15: Número de viviendas destruidas, según event tipe ...................................................... 22
Graph 16: Eventos hydro, frecuencia............................................................................................ 22
Graph 17: Eventos geo, frecuencia ............................................................................................... 23
Graph 18: Distribución temporal del número de desastres ocurridos (hydro-ext)........................ 24
Graph 19: Distribución geográfica del número de desastres ocurridos (hydro extensive) ........... 24
Graph 20: Distribución temporal del número de muertes (hydro-ext) ......................................... 25
Graph 21: Distribución geográfica del número de muertes (hydro-ext)....................................... 26
Graph 22: Distribución temporal del número de viviendas destruidas (hydro-ext)...................... 26
Graph 23: Distribución geográfica del número de viviendas destruidas (hydro-ext)................... 27
Graph 24: Distribución temporal del número de desastres ocurridos (geo-ext) ........................... 28
3
Graph 25: Distribución geográfica del número de desastres ocurridos (geo-ext) ........................ 28
Graph 26: Distribución temporal del número de muertes (geo-ext) ............................................. 29
Graph 27: Distribución geográfica del número de muertes (geo-ext) .......................................... 29
Graph 28: Distribución temporal del número de viviendas destruidas (geo-ext) ......................... 30
Graph 29: Distribución temporal del número de viviendas destruidas (geo-ext) ......................... 30
Graph 30: Access to sewerage and deaths/population. Complete sample. ................................... 38
Graph 31: Access to sewerage and deaths/population. Q1. .......................................................... 38
Graph 32: Access to sewerage and deaths/population. Q4. .......................................................... 39
TABLES INDEX
Table 1: Number of events, deaths and houses destroyed by event type (percentages) ................. 9
Table 2: Number of events, deaths and housed destroyed by risk type (percentages) ................... 9
Table 3: Characteristics of Missing Districts in DesInventar ....................................................... 32
Table 4: Events reported by districts geo-political classification ................................................. 33
Table 5: Events reported by province geo-political classification................................................ 33
Table 6: Number of reported events and districts isolation variables (OLS) ............................... 34
Table 7: Comparison of INDECI and DesInventar databases ...................................................... 35
Table 8: Frequency of events and initial conditions, by quintiles ................................................ 37
Table 9: Profile of households, if whether they suffered a natural disaster (2002) ...................... 41
Table 10: Poverty matrix (percentages)........................................................................................ 42
Table 11: Profile of households, according to poverty status (Consumption, 2002-2006) (2002)44
Table 12: Shocks experienced by household in 2002................................................................... 45
Table 13: Multinomial regression. Dependent variable: Poverty transitions, consumption (20022006) ........................................................................................................................................ 47
Table 14: Multinomial regressions. Dependent variable: Poverty transitions, consumption
(2003-2006).............................................................................................................................. 49
Table 15: Multinomial regression, Dependent variable: Poverty transitions: assets (2002-2006)50
Table 16: Multinomial regressions. Dependent variable: Poverty transitions, assets (2003-2006)
.................................................................................................................................................. 51
Table 17: Quantile regression, Dependent variable: (log) Monthly per capita consumption (2006)
Controls: 2005.......................................................................................................................... 53
4
1
Introduction
Natural hazards, an increasingly important phenomenon, have a direct impact on the welfare of
regions and specific households. The growing incidence and persistence of natural events are
strongly linked to increasing vulnerability of households and communities in developing
countries. Previous socioeconomic vulnerabilities may exacerbate the impact of a specific event,
making more difficult the process of recovery (Vatsa and Krimgold, 2000). Thus, the impact of
such events could result in an immediate increase in poverty and deprivation, but they also have
long-term, permanent effects (Carter et al, 2007).
Vulnerability to natural disasters is a complex issue, as it is determined by several conditions
like economic structure, the stage of development, coping mechanisms available, risk
assessment, and frequency as well as intensity of disasters. In this sense, the impact on the poor
could be multidimensional.
Lindell and Prater (2003) outline the policy relevance of the issue. First, policy makers can
better understand the kind of external assistance that is more effective; second, specific
population groups can be identified as more vulnerable; and third, it may be also useful for
planning fast response-assistance for natural disasters to avoid long term consequences on
welfare. For example, De Janvry, et al. (2006) show that pre-existing conditional cash transfer
schemes function as a safety net for those exposed to the disaster. Alpizar (2007) also finds that
access to formal financial services mitigates the negatives effects from natural disaster shocks for
farmers in El Salvador, as it leads to more efficient production.
Latin America is a region prone to natural disasters and the consequences are still to be
understood. Auffret (2003) found that the impact of natural disasters in Latin America and the
Caribbean is very significant, especially for the Caribbean, where the volatility of consumption is
higher than the one observed in other regions of the world, mainly due to inadequate riskmanagement mechanisms. The geographical conditions of the continent make it prone to the
occurrence of severe intensity events. Yet, part of the impact derives from the vulnerability
implied by low levels of socioeconomic development and inadequate risk management
(Charveriat, 2000). Such double-causality is extensively discussed in De la Fuente, et al. (2008).
Thus, in addition to the fact that the region has been constantly hit by several natural disasters, as
5
hurricanes, drought, floods and earthquakes, 1 the unfortunate fact that poverty and inequality are
high and persistent make of this area an interesting field for the analysis of welfare related issues
and their relation to disaster shocks.
The main objective of this study is to explore the relationship between natural hazards and
poverty in the Peruvian context. Peru is well known to have a high incidence of natural hazards
and disasters, apart from being one of the main ENSO centers in the region. Moreover, the lack
of formal insurance mechanisms for natural disasters in many areas of the country, particularly in
the poorest, as well as the tendency to establish new settlements in high-risk areas, increases the
probability of households constantly falling into poverty traps.
This document is divided in the following sections. Section 2 presents the principal
objectives and basic information on the data sources to be used. Section 3 uses the DesInventar
database to give an overview of the temporal and geographical distribution of natural hazards in
the country. In section 4 we begin to explore the links between natural hazards and district´s
socio-economic conditions. This section starts with an analysis of potential biases in the reports
of the DesInventar database, and assesses how could this bias affect the causal relationship
between natural hazards and poverty at the local level. This causal analysis at the local level will
be performed in section 5 [IF ALLOWED BY THE DATA AVAILABLE], where we will also
present the relationship at the household level of analysis. Section 6 is reserved for the
conclusions derived from this study.
2
Objectives and Methodology
The principal objectives of this study are the following:
a. To present a general overview of the different natural disasters that affected the country
in the past years, with a differentiation by type of disaster and regions affected.
b. To estimate the relation between natural disasters on social indicators at the local level
(districts or provinces), establishing a causal link whenever it is possible.
c. To determine the impact of natural disasters on social indicators -income, consumption,
assets- at the household level
For example Charveriat (2000) reports an average of 32.4 disasters per year in Latin America
and the Caribbean for the decade of nineties.
1
6
The main source of information for natural disasters will be the database of the DesInventar
program, containing reports at the district level for different sorts of natural hazards since 1970.
This database includes information on the occurrence of events as well as different types of
damages caused by them. Also at the district level, we count with information from the national
population census of 1993 and 2005, malnutrition rates for children between 5 and 9 years old
for the years 1999 and 2005 (Censo de Talla Escolar- Ministry of Education), and yearly
information on gross agricultural product since 1997. At the provincial level, we count with
information on poverty rates (fgt0) and per-capita consumption derived from poverty maps
elaborated using the 1993 and 2005 national population census and household surveys.
For the analysis at the household level we can rely on the National Household Survey
(ENAHO) of the National Institute of Statistics (INEI). With this survey it is possible to
ensemble a five-wave panel database for the period 2002-2006 with information for more than
two thousand households located in rural areas. As this survey is used to calculate and monitor
poverty in the country, it allows calculating household’ consumption levels as well as income,
including also several information on durable and productive assets. What is more, the survey
includes a question about the experience of a negative shock in the last 12 months (death of an
income’s provider, unemployment, natural hazard), and asks also about the consequences of that
shock and the strategies undertaken (depletion of assets, borrow money, etc.)
3
Overview of Natural Hazards in Peru
Peru is globally considered among the countries where ‘El Niño Southern Oscillation’
(ENSO) strikes harder. The Peruvian ocean is the scenario of the encounter of warm waters from
the Equator with the colder front coming from the extreme Southern Pacific (better known in
Peru as the ‘Humbold current’). The predominance of the colder waters from the south explains
how, despite being a ‘tropical’ territory, we find a much more moderate temperature throughout
the Peruvian coastal region, with very little precipitation and highly dependent on rainfall at the
Western slopes of the Andean highlands.
At the peak years along the ENSO cycles, popularly known as ‘El Niño years’, the classic
pattern of events is a combination of floods in the northern coast with extreme droughts in the
southern Andean highlands. The most recent ‘El Niño years’ have been 1972, 1983, and 1997-98
7
although the ENSO cycle is a dynamic climatological process and in the recent years the media
tend to grade every year having a more or less strong ENSO effect.
The following analysis provides an overview of natural hazards in Peru since 1970. The
assessment is focused both upon the frequency as well as the magnitude (measured in terms of
number of deaths and number of houses destroyed) of the events. The bottom line of the
assessment is to find a pattern of the most important hazards, their geographical location and
their long term cycles. However, since the objective of the study is to assess the relationship
between natural disasters and poverty, we are more interested not necessarily in the most
catastrophic and isolated events, rather than the most regular and predictable. The latter are
usually smaller events but their accumulation over time may cause huge material costs as well as
deaths.
We classify the events using two criteria: extensive – intensive events depending of the
number of deaths and houses destroyed (risk typology), and the division between geological and
hydro - meteorological events (event typology).
Risk type:
•
Intensive (Int): Any event with more than 50 death or 500 houses destroyed in any
district.
•
Extensive (Ext): The rest
Event Type:
•
Hydro Meteorological (Hydro)
•
Geological (Geo)
Tables 1 and 2 present the frequency distribution of events by type and risk in Peru (19702006). Table 1 shows that the hydro – meteorological events are the most frequent, representing
more than 90% of the total events reported in the period 1970 -2006. However, the number of
deaths caused by the hydro – meteorological events represent only 62.5% of the total number of
deaths. In general, the geological disasters, despite its low frequency, cause the most number of
houses destroyed, 66.4%, and an important percentage of the total number of deaths.
8
Table 1: Number of events, deaths and houses destroyed by event type (percentages)
Frecuency
Hydro - met
Geological
91.4
8.6
62.5
37.5
33.6
66.4
100.0
100.0
100.0
TOTAL
Deaths
Houses
Destroyed
Event type
In Table 2 we can see that natural hazards classified as extensive represent almost the total
(99.5%) number of events registered in the period. However, given the definition provided for
intensive events (with more than 50 deaths or 500 houses destroyed), although they represent
only 0.5% of the total events reported, they cause 87.6% of the deaths and 74.8% of houses
destroyed.
Table 2: Number of events, deaths and housed destroyed by risk type (percentages)
Risk
Type
3.1
Frecuency
Deaths
Houses
Destroyed
Extensive
Intensive
99.5
0.5
12.4
87.6
25.2
74.8
TOTAL
100.0
100.0
100.0
All the events: hydro-met and geological, intensive and extensive
3.1.1
Number of events reported
A continuación mostramos la distribución temporal de los distintos desastres naturales a
partir de 1970 hasta el año 2006. Se incluyen todos los tipos de desastres, tanto geológicos como
hidro-meteorológicos. También los eventos intensivos como extensivos, en cuanto a la magnitud
de impacto.
9
Graph 1: Distribución temporal del número de eventos (All)
En el gráfico Nº 1 apreciamos que existen picos notorios en la frecuencia temporal de los
desastres naturales. Los años en los que ocurrieron estos picos son 1970, 1972, 1983, 1994,
1997, 1998 y 2001. Los grandes eventos que pudieron repercutir en estos resultados son: la
ocurrencia del “Fenómeno del Niño” que tuvo sus efectos en los años 1972-1973, 1982-1983 y
1997-1998, o los terremotos de los años 1970 y 2001. Todos estos eventos coinciden con los
picos existentes en el gráfico anterior. El fenómeno del niño se relaciona directamente con la
ocurrencia de ciertos eventos naturales, principalmente con: alluvion, flash flood, flood, rains
and storm, además de muchos otros eventos. Repercute tanto en la intensidad como en la
frecuencia de este tipo de fenómenos.
Los grandes terremotos, de 1970 y del año 2001, (más fuertemente el primero de ellos)
también repercuten en la frecuencia de la ocurrencia de eventos. Directamente en los sismos,
creándose múltiples réplicas que son contabilizadas y pueden influenciar en esta información,
además de otros eventos relacionados, como por ejemplo los landslide. El pico que no es
10
explicado por este tipo de grandes eventos, es el del año 1994 que podría estar asociado a
efectos retrasados del niño del 93.
Graph 2: Distribución geográfica de la ocurrencia de eventos naturales (All)
En cuanto a la distribución geográfica de la ocurrencia de desastres naturales de todo tipo,
notamos la existencia de focos de ocurrencia de estos eventos. Sobre todo en Lima, Huancayo,
Arequipa y la zona norte (Piura, Chiclayo). Es interesante ver que en las provincias aledañas a
las capitales mencionadas también se registran altas tasas de ocurrencias de eventos naturales.
Esto se puede deber a un sesgo de información. Las principales fuentes que registran e informan
sobre los desastres naturales, están centralizadas en las capitales distritales, provinciales, y en
especial en las capitales departamentales. Entonces, la mayor ocurrencia de eventos naturales en
las zonas más urbanas puede estar explicada por motivos de mayor accesibilidad a las fuentes de
información. Por este motivo, para las zonas más aisladas, que frecuentemente son las más
pobres, pueden estar reportándose menores niveles de frecuencia de desastres naturales en
relación con las zonas menos aisladas, que son generalmente las capitales de cada nivel. Este
asunto lo analizaremos en la sección 4.
11
Luego de esta aclaración, dentro de esta perspectiva global, en la cual se cuentan todos los
tipos de eventos, examinaremos dos medidas de la magnitud de afectación de los desastres
naturales: muertes causadas por la ocurrencia de los desastres naturales y casas destruidas por la
misma causa.
3.1.2
Number of deaths caused by natural hazards
Graph 3: Distribución temporal del número de muertos por la ocurrencia de desastres
naturales (All)
El gráfico Nº 3 nos muestra claramente que el número de muertes por desastres naturales en
el año 1970 es significativamente mayor que la de los demás años. El terremoto de ese mismo
año es lógicamente el que explica tal resultado. Como nuestro análisis quiere ir más allá del
examen de los “grandes eventos”, debido a que estos eventos muestran una intensidad sin
precedentes, que no es frecuente y que además afecta a todos prácticamente sin exclusión,
debemos concentrarnos principalmente en los eventos de menor intensidad, pero que afectan
permanentemente a ciertas regiones y pueden perpetuar el nivel de pobreza que tengan.
Una vez aclarado este punto, veamos cual ha sido la distribución provincial de la magnitud,
en cuanto a muertes, de los desastres naturales.
12
Graph 4: Distribución geográfica del número de muertos por la ocurrencia de desastres
naturales (All)
La distribución geográfica del número de personas muertas por la ocurrencia de desastres
naturales nos muestra que en el departamento de Huaraz, especialmente en las provincias de
Yungay, Santa y Huaraz, se concentra el mayor número de muertos. Esto tiene su explicación en
el terremoto del 70, cuyo epicentro se registró justamente en esa zona. Además, Lima y
provincias aledañas (como Huarochiri y Huaraz) al igual que en Arequipa y la sierra central
presentan una fuerte incidencia de muertes causadas por el impacto de desastres naturales.
13
3.1.3
Number of houses destroyed caused by natural hazards
Graph 5: Distribución temporal del número de casas destruidas por la ocurrencia de
desastres naturales (All)
El gráfico anterior nos muestra una mayor diversidad en la distribución temporal de la
magnitud de los desastres naturales. En este caso la magnitud del impacto de los desastres
naturales es medida por los daños materiales causados. El resultado encontrado muestra que
sigue siendo el terremoto del 70 el evento largamente más significativo, sin embargo también
sobresale, aunque en mucha mayor medida, la cantidad de casas destruidas en los años 1983 y
2001. Como sabemos, el año 2001 se produjo el “fenómeno del niño” y el año 2001 un
terremoto de gran magnitud en el sur del país.
14
Graph 6: Distribución geográfica del número de hogares destruidos por la ocurrencia de
desastres naturales (All)
En el gráfico Nº 6 apreciamos las zonas más perjudicadas, en un sentido material, por la
pérdida de viviendas. Sigue siendo Huaraz la zona más afectada, además la zona norte, en
especial Sullana, Piura y Lambayeque, áreas en las que el fenómeno del niño suele causar
grandes pérdidas Por último, Ica y Arequipa también han sufrido grandes pérdidas materiales en
el periodo de estudio.
Una vez analizado el panorama global en cuanto a frecuencia de eventos naturales y
magnitudes de estos, llama la atención el gran efecto de los terremotos de 1970 y del año 2001,
además de los “Fenómenos del Niño” de los años 1972-1973, 1982-1983, 1997-1998. En la
próxima sección analizaremos todos los eventos y efectos causados por el “Fenómeno del Niño”.
3.2
“Fenómeno del Niño”: frequency and magnitude
En esta sección analizaremos todos los eventos ocurridos por causa del Fenómeno del Niño.
El objetivo es analizar cual ha sido la magnitud de este fenómeno, su frecuencia y su ubicación
geográfica.
15
3.2.1
Number of events caused by ‘El Niño’
Graph 7: Distribución temporal del número de eventos ocurridos por el fenómeno del niño
El gráfico Nº 7 nos confirma las fechas de ocurrencia del “Fenómeno del niño”, excluyendo
el sucedido en el año 1972. La base de datos analizada, “Desinventar”, no registra eventos
causados por el “Fenómeno del Niño” del año 1972, a pesar que, tal como vimos anteriormente,
dicho año presenta elevados reportes de desastres. La posible causa de esta omisión es que todos
los desastres causados por el “Fenómeno del Niño” en dicho año, y que están reportados en la
base, no hayan sido clasificados según dicha causa. Por este motivo, la presente sección muestra
información de los “Fenómenos del niño” ocurridos solo para los años 1982-1983 y 1997-1998.
Pero dada la naturaleza de estos fenómenos, creemos que las principales conclusiones que
puedan extraerse de este análisis pueden ser extendidos para los años 1972-1973.
16
¿Cuáles son los principales desastres naturales que están vinculados a la ocurrencia del
fenómeno del niño?
Graph 8: Comparación por tipo de evento del número de eventos ocurridos por el
Fenómeno del Niño
El gráfico anterior nos muestra que los eventos rains, alluvion y flood son eventos ligados a
los “Fenómenos del Niño” y por tanto son los desastres naturales que afectan negativamente las
áreas en las cuales dicho fenómeno se presenta.
Graph 9: Distribución geográfica del número de eventos ocurridos por el Fenómeno del
Niño
17
En el mapa anterior podemos observar la ubicación de las provincias con mayor incidencia de
desastres generados por el “Fenómeno del Niño” (tanto para los años 1982-83 como para 199798). Aquí podemos observar en detalle que son las provincias costeras, en especial las de la zona
norte, las que mayor número de desastres registran.
3.2.2
Magnitude of ‘El Niño’ in terms of deaths and houses destroyed
Graph 10: Distribución geográfica del número de personas muertas por el Fenómeno del
Niño
Graph 11: Distribución geográfica del número de viviendas destruidas por el Fenómeno del
Niño
18
Los gráficos Nº 10 y Nº 11 nos muestran la magnitud de los desastres naturales causados por
el “Fenómeno del Niño”. El número de personas muertas es muy reducido y se concentra en la
costa norte del país. Las pérdidas materiales tuvieron un mayor efecto, pues ocasionaron grandes
pérdidas en las zonas afectadas, principalmente en las provincias Ica, Sullana, Trujillo y Satipo.
Una vez analizados los efectos del Fenómeno del Niño y su incidencia geográfica, pasaremos
a estudiar los desastres naturales según sus clasificaciones de event tipe y risk tipe.
3.3
Intensive Events
Esta sección tiene el propósito de identificar los años y lugares en donde ocurrieron los
desastres naturales de mayor impacto. Como mostramos en la tabla Nº 2, la ocurrencia de
desastres naturales clasificados como intensive es muy reducida, representando el 0.5% del total
de eventos ocurridos durante los últimos 36 años. Pero a pesar de esta esporádica frecuencia, son
los desastres cuyo impacto es el causante de cerca del 90% de las muertes y aproximadamente el
75% de las viviendas destruidas.
Veamos en que momento del tiempo se presentaron dichos desastres y que áreas geográficas
afectaron
3.3.1
Number of intensive events
Graph 12: Distribución temporal del número de desastres intensivos ocurridos (all
intensive)
19
El gráfico Nº 12 nos muestra claramente la relación directa existente entre los eventos
intensivos y los terremotos de 1970 y 2001. También se aprecia una relación, pero más leve,
entre los fenómenos del niño, en los años 1983 y 1998, y la ocurrencia de eventos intensivos.
3.3.2
Magnitude of the intensive events
Como vimos en el gráfico Nº 3 y Nº 5, las pérdidas humanas y materiales se concentran
fuertemente en el año 1970, y en una mucha menor magnitud, en el año 2001. Al tener estos
eventos un impacto tan grande, sobresalen en cuanto a su capacidad de afectación. Entonces, la
ubicación geográfica de las zonas en las cuales hubieron más pérdidas humanas o materiales por
la causa de los desastres naturales intensivos, no varían con respecto a las identificadas en los
gráficos Nº 4 y Nº 6. Por tanto la misma explicación se aplica para esta sección.
3.4
Extensive events
En adelante analizaremos los eventos extensivos, es decir, los múltiples desastres naturales
ocurridos durante todo el periodo estudiado (1970 – 2006). Como se mencionó anteriormente,
estos eventos son de mayor interés para nuestro análisis, pues es posible identificas regularidades
temporales, como geográficas, en cuanto a su afectación. Y la identificación de los elementos
mencionados puede ser de gran utilidad, pues permite planificar estrategias para prevenir este
tipo de efectos, como para identificar a las familias más vulnerables ante este tipo de
eventualidades de la naturaleza.
Dentro del los eventos extensivos, haremos una diferenciación entre los tipos de eventos
según su naturaleza: Hidro-meteorológicos y Geológicos.
20
3.4.1
Comparative analysis of Hydro – meteorological and Geological events
3.4.1.1 By number of events reported
Graph 13: Número de eventos ocurridos, según event tipe
Extensive Events: Number of events by event tipe
Geological ,
8%
Hidro-met ,
92%
3.4.1.2 By number of deaths
Graph 14: Número de muertes, según event tipe
Extensive Events: Number of deaths by event tipe
Geological ,
18%
Hidro-met ,
82%
21
3.4.1.3 By number of houses destroyed
Graph 15: Número de viviendas destruidas, según event tipe
Extensive Events: Number of houses destroyed by event
tipe
Geological ,
38%
Hidro-met ,
62%
Los gráficos Nº 13, 14 y 15 nos muestran que los eventos hydro son los más frecuentes (92%
del total), los que significativamente causas la mayor cantidad de pérdidas humanas, y las que
causan también la mayor cantidad de viviendas destruidas, aunque en este último caso la
proporción no es tan elevada con respecto a los efectos de los desastres geológicos.
Dentro de los desastres extensivos, ¿Cuáles son los desastres naturales que se relacionan con
los tipos de eventos?
3.4.1.4 Frequency of Hydro – meteorological events
Graph 16: Eventos hydro, frecuencia
22
Del cuadro Nº 16 se desprende que los eventos más característicos del tipo hydo son: flood,
rains, alluvion and landslide.
3.4.1.5 Frequency of Geological events
Graph 17: Eventos geo, frecuencia
En el caso de los eventos geo, prácticamente los únicos eventos significativos son los
earthquake, con una representación de más del 95%.
3.4.2
Hydro – extensive events
Pasaremos a detallar la distribución temporal y geográfica de la ocurrencia de este tipo de
eventos. Además de las variables para medir la magnitud de estos. Estos tipos de eventos, los
hydro-ext, son los más frecuentes.
23
3.4.2.1 Number of events
Graph 18: Distribución temporal del número de desastres ocurridos (hydro-ext)
El gráfico Nº 18 nos muestra que los años en los que se registró un mayor número de
desastres naturaleza de este tipo son los años 1972, 1973, 1983, 1997, 1998 (Años del
“Fenómeno del Niño”) y en el año 1994. Como en este análisis solo consideramos los eventos
extensive, los años 1970 y 2001 de ocurrencia de terremotos, no aparecen tan significativos
como anteriormente, a pesar de seguir siendo años importantes en cuanto a la ocurrencia de
desastres. Además, la inspección visual nos muestra la ocurrencia de ciclos de desastres
naturales, temporadas bajas y altas.
Graph 19: Distribución geográfica del número de desastres ocurridos (hydro extensive)
24
En cuanto a la distribución geográfica de este tipo de desastres (Hydro-Ext), mediante el
gráfico Nº 19 notamos que las zonas que presentan un mayor índice de ocurrencia de eventos
son: Arequipa, Huancayo, Lima, Huarochiri, además de ciudades del norte, como Piura, Trujillo,
y del centro como Satipo, Chanchamayo y Oxapampa. Es interesante apreciar que las áreas más
pobladas y concurridas sean las que mayores eventos presentan, una vez más la idea de que
exista un sesgo en la información de ocurrencia de desastres se presenta.
3.4.2.2 Number of deaths
Graph 20: Distribución temporal del número de muertes (hydro-ext)
En el gráfico Nº 20 observamos que el mayor número de muertes se presentan en los años
1982, 1983, 1994, 1996 y 2001. Y las fechas en las que ocurren menos de estos eventos son en
los años 1976, 1989 y 1992.
25
Graph 21: Distribución geográfica del número de muertes (hydro-ext)
El gráfico Nº 21 nos muestran que las zonas que han tenido mayores pérdidas humanas por
desastres naturales del tipo Hydro-Ext, son Lima y Chanchamayo, seguidos por Huancayo, La
convención, Puno y Arequipa. De estas seis zonas, cuatro son capitales departamentales, y las
dos restantes son zonas con alta densidad poblacional.
3.4.2.3 Number of houses destroyed
Graph 22: Distribución temporal del número de viviendas destruidas (hydro-ext)
26
El gráfico Nº 22 muestra información interesante, pues muestra un cambio en el patrón de
viviendas destruidas por causa de desastres naturales a partir del año 1994. Según esta
información, los desastres naturales pequeños (extensivos) han tenido una mayor magnitud en
cuanto a pérdidas materiales, a partir del año 1994.
Graph 23: Distribución geográfica del número de viviendas destruidas (hydro-ext)
La distribución geográfica del número de viviendas destruidas es variada, pero llama la
atención que sea la provincia de Lima, la que muestra un mayor índice de viviendas destruidas.
3.4.3
Geo Extensive events
Los eventos Geo Extensive son mucho menos frecuentes que los Hydro, y se compone
básicamente de temblores y terremotos. En esta sección examinaremos este tipo de eventos, en
cuanto a los considerados extensive, es decir, los de magnitud moderada o pequeña.
27
3.4.3.1 Number of events
Graph 24: Distribución temporal del número de desastres ocurridos (geo-ext)
En el gráfico Nº 24 observamos que la frecuencia de estos eventos es muy alta en los años en
que ocurrieron terremotos de gran magnitud, los del 70 y del 2001. Por tanto los eventos
registrados en esta sección están relacionados con los terremotos de aquellos años, pues pueden
ser réplicas o el mismo terremoto registrado como temblor en una zona alejada del epicentro.
Graph 25: Distribución geográfica del número de desastres ocurridos (geo-ext)
28
La zona más afectada, en cuanto al número de desastres de este tipo ocurridos, es la del sur,
en especial Arequipa y Caylloma. Esto está asociado directamente al terremoto del año 2001.
3.4.3.2 Number of deaths
Graph 26: Distribución temporal del número de muertes (geo-ext)
Del gráfico anterior llama la atención el hecho que en el año 1990 se registre una gran
ocurrencia de muertes por desastres naturales del tipo indicado. Este resultado se explica por la
ocurrencia de un desastre aislado, el terremoto producido en San Martin el año 1990, el cual
afectó a las provincias Rioja, Moyabamba y al departamento de Amazonas, en especial a la
provincia de Rodríguez de Mendoza.
Graph 27: Distribución geográfica del número de muertes (geo-ext)
29
El gráfico Nº 27 grafica lo dicho anteriormente. Además de que la ocurrencia de muertes esté
asociada a los lugares donde ocurrieron los mencionados terremotos del 70 y 2001, está asociado
al terremoto que afectó los departamentos de San Martín y Amazonas.
3.4.3.3 Number of houses destroyed
Graph 28: Distribución temporal del número de viviendas destruidas (geo-ext)
En cuanto a las viviendas destruidas, en el año 2001 se registra el mayor número. Esto claro,
excluyendo los desastres más catastróficos, considerados intensive. De nuevo, este hecho esta
directamente asociado al terremoto del 2001 en el sur del país.
Graph 29: Distribución temporal del número de viviendas destruidas (geo-ext)
30
El gráfico Nº 29 nos brinda la misma información que la del gráfico Nº 27, la mayor cantidad
de casas destruidas por desastres geo-ext, son causa del terremoto del 2001, afectando a la zona
sur del país.
4
Natural Hazards and the characteristics of affected areas
4.1
Analysis of Bias in the Natural Hazards Reports of the DesInventar database
The information on natural hazards in the DesInventar database relies on reports that appear in
newspapers of the national capital of Peru, Lima. Therefore, it was expected that natural hazards
events which occurred in somehow more “important” districts would have a higher probability of
being reported in these newspapers than other events happening in more “marginal” districts of
Peru. The objective of this section is to analyze the extent of this potential bias in the
DesInventar database.
For that matter, we will use different measures of district´s importance and relate them to
the number of reported events in the database. Our main comparison will be between districts
that are capital of a province and the rest of the districts in the same province. A similar analysis
will be done with provinces that are capital of a department and the rest of the provinces in the
same department.
4.1.1
Missing districts in the DesInventar Database
First of all, it is important to notice that the number of districts for which the complete
DesInventar database (1970-2006) contains positive reports on natural hazards (Extensive and
Intensive events of both, Hydro-meteorological and Geological type) is only an 80% (1,477) of
the total number of districts in Peru. In the following table we compare some characteristics of
the districts with a positive number of reports in the DesInventar database with the districts that
do not present any information.
31
Table 3: Characteristics of Missing Districts in DesInventar
Variables (at 1993)
Population Variables
Total population
Urban population
Rural population
Access to public services
Percentage with electricity
Percentage with cces
Percentage with sewerage
Other variables
Altitude
Index of ccessibility (less is more accessive)
Surface
Population pressure
Districts capital of province (1 yes, 0 no)
With reports
Without reports T-Test (p-value)
14,284.94
10,343.75
3,941.19
3,378.82
901.09
2,477.73
0.00
0.00
0.00
29.41
36.77
33.88
12.12
27.41
26.09
0.00
0.00
0.00
2,161.02
4.12
266.26
499.19
0.13
2,559.59
5.30
190.93
41.00
0.01
0.00
0.00
0.00
0.00
0.00
District´s without any natural hazard reported in the database are less populated and more
rural than the ones with reports. Moreover, districts excluded from the DesInventar database
present on average worse living conditions than the ones with at least one report. Variables
related to district´s isolation, like mean altitude and accessibility (index derived from type of
principal road), confirm the idea that more marginal districts have being excluded from the
database.
The last variable included in the table, reveals that almost all districts that are capital of a
province present at least one report in the database. As we mentioned before, one feature that
could better capture the district´s importance is the condition of being a province´ capital. In the
following section we will use this classification to explore the difference in the number of reports
between these districts and the rest.
4.1.2
Comparison of Reports by district´s geo-political condition
We compare the difference between the number of reports registered in each district capital of
province with the mean number of reports registered for the rest of the districts of that same
province. For that matter we use two different categories of events: all events reported, and
extensive events excluding fire.
32
Table 4: Events reported by districts geo-political classification
T-TEST
Mean events reported
Geo-political
classification
All events
Events Extensive
no-Fire
Provincial capital
Other districts
20.5
4.2
17.3
3.9
P-value
0.0
0.0
At a 99.9% significance level, the number of reports in capital districts are higher than the
average number of reports in other districts.
Continuing with this idea, now we scale up the previous comparison to the provincial
level. We proceed by calculating the difference between the number of reports in the districts
capital of provinces and the mean number of reports for the rest of the districts in the same
province. Then, we identify the provinces that area capital of a department, and compare the
difference found before, between these provinces and the rest of the provinces in the same
department. We expect that the difference in number of reports between districts that are capital
of a province and the rest of the districts in the same province, is higher for provinces that are
capitals of a department than for provinces that are not.
Table 5: Events reported by province geo-political classification
T-TEST
Mean difference between # of reports in districts that
are provincial capitals and the mean # of reports in
districts that are not provincial capitals
All Events
Events Extensive No-fire
Provinces that are departmental
capitals
64.5
49.3
Other provinces
8.1
7.4
P-value
0.0
0.0
The results of this table confirm our hypothesis. The bias on the number of reports is even
stronger for districts capitals of a province that is in turn the capital of a department.
33
Finally, when we look at the relationship between the number of reports by district and
different measures of district´s isolation, we find a significant coefficient for most of them.
Table 6: Number of reported events and districts isolation variables (OLS)
Number of reports
event_extnofire
event_inten
Variables
event_all
Altitud
0.00127***
(-5.21)
-0.000872***
-0.0000118***
(-4.12)
(-3.57)
Surface
0.00378**
-2.84
0.00435***
-3.75
0.000000188
-0.01
Accesibility Index
-1.147***
(-8.04)
-1.033***
(-8.34)
-0.00435*
(-2.25)
Distance to
provincial capital
-0.0158**
-0.0140**
-0.000113
(-3.05)
(-3.10)
(-1.60)
13.60***
-15.62
1541
0.094
11.45***
-15.15
1541
0.092
0.0787***
-6.66
1541
0.017
_cons
N
adj. R-sq
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001
In conclusion, there seems to be a strong bias in the information on natural hazards in the
DesInventar database. First, more isolated districts do not count with any report on natural
hazards in the past 36 years. Second, districts of higher rank or importance in terms of geopolitical classification systematically present a higher number of reported events than the rest,
even when compared to their neighbor districts. Given that the main scope of this study is to
assess the relationship between natural hazards and welfare indicators, the fact that districts with
better socio-economic conditions (like provincial capitals) tend to have a higher number of
reported events in the DesInventar database due to its method for data collection, will seriously
limit the possibility of using this information in our analysis.
34
NOTE TO FIRST DRAFT: The Instituto Nacional de Defensa Civil-INDECI has information
on the occurrence of natural hazards and its loses at a national level for the period 2003-2007.
Even though their website just report this information at the national level by year and type of
event, it might be possible to obtain the raw datasets with information at the district level.
Contrary to DesInventar, INDECI collects information on natural hazards from its decentralized
offices located in almost all provinces of Peru. A basic comparison of these databases highlights
the extreme difference in the number of reports:
Table 7: Comparison of INDECI and DesInventar databases
# of events reported
2003
2004
2005
2006
Type of event
INDECI
DesInv
INDECI
DesInv
INDECI
DesInv
INDECI
DesInv
Strong Rains
388
15
426
17
391
7
738
27
Floods
470
83
234
49
134
8
348
41
Earthquackes
35
12
11
1
261
14
32
15
Freeze
73
42
438
28
296
2
177
12
Landslides
138
46
100
29
99
11
158
14
5
3
215
39
224
5
74
0
Droughts
4.2
Correlations between initial socio-economic conditions and Disaster´s loses at the
District level
The main idea of this section is to provide a basic socio-economic profile of the districts that
typically suffer the most disasters. Even though the frequency of occurrence (and some
measurements of intensity like ml of rainfall or scale of an earthquake) of natural hazards can be
considered “exogenous” to the socio-economic characteristics of a district, the value of loses
caused by those natural hazards (called susceptibility index in the background paper) might be
correlated to those characteristics. So, if we have a group of districts with different initial socioeconomic conditions, and similar patterns of occurrences of natural hazards, we can run
correlations between the values of loses caused by those hazard and the initial conditions to see
what type of districts are the ones more affected.
One potential problem with this formulation however, is that we are assuming that the initial
socio-economic conditions of districts are somehow exogenous to previous occurrences of
35
shocks. Given the availability of long time series data on natural hazards, we can try to control
this problem in the following way:
•
Use the information from the 93 census as initial socio-economic conditions.
•
Create an index of “frequency” of events for the previous 10 years (1982-1992), or even
since the start of the DesInventar database. This can be just the total number of events for a
particular district, or the total number of certain type of events only.
•
We can then classify districts, in quintiles for example, using the distribution of that index.
We will expect some variation in terms of socio-economic conditions between groups
(districts in the first quintile of this distribution-less affected by natural hazards- should
present on average better socio-economic conditions than districts in the last quintile), but
also enough variability of these conditions within districts in the same group.
•
Next, we calculate the value of loses by district for a period after initial conditions (19932000 for example). These loss values should be normalized by total district population per
year per 10,000 habitants, using inter-census population growth rates. The variables of loses
to be used in this exercise could be either houses destroyed or the number of deaths.
Finally, we will run correlations between initial socio-economic districts´ characteristics
(measured as percentages of people with access to public services) and the normalized value of
loses, for districts in the same “quintile”. Note that we are not considering in this analysis the
fact that a difference in loses between two districts for the period after 1993 can be due to the
occurrence of more events in one of them and not necessarily due to different initial conditions.
But we think that the classification of districts by quintiles according to the “history of natural
hazards” will be already helping to control for this problem.
We perform this exercise using the total number of extensive no-fire events for the period
1982-1992 to create the frequency index and its quintiles, and then run the correlations between
the percentage of population with sewerage connected to the public network and the number of
(deaths/population) *10,000 for the period 1993-2000.
The next table gives us an idea of the extent to which the bias reported before in the
information of the DesInventar database can be affecting this type of analysis.
36
Table 8: Frequency of events and initial conditions, by quintiles
Q0
Q1
Q2
Q3
Q4
Number of districts
1,147
299
144
83
158
Events extensive no-fire 82-92
0
1
2
3
9
Total population 1993
6,439
11,375
27,016
17,762
38,921
% with electricity 1993
19
32
39
45
48
% with water 1993
30
38
42
44
55
% with sewerage 1993
27
35
44
44
50
deaths 93-00
1.0
1.2
1.7
2.3
4.8
deaths/pop 93-00
1.54
1.09
0.62
1.31
1.24
First, we can observe that a big majority of districts in the database did not have any
report of natural hazards for the period 1982-1992. These districts were classified as a category0,
and the rest were grouped in quartiles according to the frequency index. Second, we find again,
as we showed in the analysis of bias, a very strong and positive relationship between population
or access to different types of public services and the total number of events reported. Finally,
while the number of deaths caused by natural hazards on the period 1993-2000 increases for
districts with higher index of occurrence of events (1982-1992), when this variable is normalized
by population (deaths/pop 93-00) the relationship changes completely. This difference is
principally due to the positive correlation between number of reported events and district´s
overall population, which is a direct result of the bias in the DesInventar database.
Using these results, the correlations between initial district´s characteristics and loses by
natural disasters will be the following:
37
0
20
40
60
80
100
Graph 30: Access to sewerage and deaths/population. Complete sample.
0
50
100
150
(max) death_extnofire_pob_93_00
p_con_desague93
200
250
Fitted values
Correlation coefficient: -0.058
0
20
40
60
80
100
Graph 31: Access to sewerage and deaths/population. Q1.
0
20
40
60
p_con_desague93
Fitted values
38
80
100
0
20
40
60
80
100
Graph 32: Access to sewerage and deaths/population. Q4.
0
10
20
30
p_con_desague93
40
Fitted values
But as we can derived from the previous analysis, this relationships are only hidden a bias in the
report of events in the DesInventar database.
5
The impact of Natural Hazards on Poverty indicators
5.1
Analysis at the District level
5.1.1
The effect of Natural Hazards on Poverty rates: Provincial Poverty Maps (19932005)
5.1.2
The effect of Natural Hazards on Children´s Malnutrition at the District level (19992005)
5.1.3
The effect of Natural Hazards on the Value of Agricultural Production at the
District level (1997-2006, yearly panel)
39
5.2
Analysis at the Household level
5.2.1
Data and Descriptive Statistics
The quantitative analysis is based on the national household survey ENAHO, conducted by the
National Institute of Statistics (INEI). It has been possible to ensemble a five-wave unbalanced
panel database for the period 2002-2006 with information for 2,091 households at rural level.
However, most of these households do not have information for all the five periods of the survey.
Some households were not encountered again, while others were not included in one year, but
appear again in another wave. Due to these problems, the balance panel database just includes
831 households.
ENAHO is used to calculate and monitor poverty in the country, consequently it allows
calculating household’ consumption levels as well as income. Furthermore, it includes valuable
information regarding durable and productive assets and access to public services. The survey
also includes a question about the experience of a negative shock in the last 12 months (death of
an income’s provider, unemployment, natural hazard), and asks also about the consequences of
that shock and the strategies undertaken (depletion of assets, borrow money, etc.)
Table 9 shows the average of the most important variables use in our analysis for the year
2002. Those are the “initial conditions” that characterized the households of our sample (see
Appendix 1 and Appendix 2 for a full report of the descriptive variables for the unbalance and
balance panel, respectively). There are statistically significant differences between the
households that report having experienced a natural disaster in 2002 and the ones that did not.
The human capital variables show more positive results for the households that experienced a
shock. By contrast, those households had less access to piped water, electricity and fixed
telephone. A lower percentage of those households received income from renting private
properties, in comparison with the households that did not experience a shock. Furthermore,
households that experienced a shock in 2002 were less integrated to the market, since a lower
percentage of their total income came from monetary sources. This is consistent with the fact
they had a higher percentage of income that came from agricultural activities. These results
could be signaling some bias in the report of natural disasters by households more involve in
traditional agriculture and with less access to market and services. Notice that, those households
were poorer in 2002, but the result is not statistically different from the poor rate of the
40
households that did not experience a shock. The analysis of the impact of natural disaster will
take into account this feature of the sample.
Table 9: Profile of households, if whether they suffered a natural disaster (2002)
Natural
Natural
Difference
disaster
disaster
(p-value)
Variable
(NO)
(YES)
PERCENTEGES
Human capital
Gender of the hh (woman)
14.13
7.01
**
HH is literate
54.52
82.68
***
At least one children don't go to school
4.27
0.00
***
Characteristics of the dwellings
Low quality of dwelling's materials
22.38
19.68
Owner of house
84.39
92.59
Water: access to public network
42.21
18.08
***
Sewerage connected to public network
57.07
47.49
Electricity as lightning source
37.07
20.56
*
Telephone (fixed)
0.36
0.00
*
Welfare indicator
Poor [consumption]
63.97
75.44
Poor [assets]
40.69
28.56
Risk management and coping indicators
Receive income from renting private properties
9.86
2.58
**
Remittances
Receive local remittances
27.58
18.38
Receive international remittances
0.62
1.31
Remittances (from at least one source)
28.14
19.69
Food assistance (at least one member)
Glass of milk
42.20
58.37
Popular dinning room
7.22
1.53
**
Scholar breakfast
20.39
52.74
***
Other program
7.00
19.95
Proportion of beneficiaries (as a proportion of total members)
24.97
40.43
***
Welfare indicators
Monetary expenses (as % of total expenses)
59.10
57.38
Monetary income (as % of total income)
59.39
48.70
***
Participation in agricultural activities
Percentage of members that have as main activity agriculture
42.78
48.94
Percentage of members that have as secondary activity
agriculture
8.31
12.09
Percentage of income from agricultural activities
34.89
42.15
*
41
AVERAGES
Human capital
Age of the hh
Average years of education of the members of the household
Total years of education of the members of the household
Average years of education of the hh
Welfare indicators
Number of members per worker
Assets
Livestock (on sheep equivalences)
Vector of assets
Local Remittances (Yearly amount)
International Remittances (Yearly amount)
*** 1% significance ** 5% significance *10% significance
Source: ENAHO 2002-2006. Balance Panel
5.2.2
48.11
4.32
20.00
4.46
46.30
4.41
21.24
5.04
2.95
3.19
18.30
769.51
30.23
511.30
223.96
8.56
358.07
9.67
**
Poverty matrix
Table 10: Poverty matrix (percentages)
Year
2006
Poor
Non-Poor
Overall
2002
Poor
46,14
18,05
64,19
Non-Poor
9,38
26,43
35,81
Overall
55,51
44,49
100,00
Source: ENAHO 2002-2006. Unbalance Panel
Probability that a non-poor household in 2002 becomes poor in 2006: 0.26
•
(Non Poor(2002) and Poor(2006))/Total Non Poor in 2002
Probability that a poor household in 2002 becomes non poor in 2006: 0.28
•
(Poor(2002) and Non-Poor(2006))/Total Poor(2002)
We obtain four categories from analyzing poverty transitions from 2002 to 2006. A
household is classified as “Never Poor” if it has never fallen under the poverty line in the five
periods of the survey. Conversely, it is classified as “Always Poor” if it has been poor in every
wave of the survey. Households are can be also classified as “Several episodes” if it has been
42
poor more than two times but less than five times, between 2002 and 2006. Finally, a household
that has fallen under the poverty line just once is classified in the category “One episode”.
The four categories obtain from the construction of the poverty matrix are used in Table
11 to draw a new profile of the households in the sample. In addition, a mean analysis has been
included to show if the differences between the households classified as “Never poor” and
“Always poor” are statistically different. As expected, the households that never experienced an
episode of poverty were better endowed than the households classified as “Always poor” in
terms of human capital, assets and access to services. The former are also more integrated to the
market, which is reflected in their higher percentage of monetary income and expenses. Notice
that there seems to be a positive correlation between the proportion of income generated from
agricultural activities and the number of poverty episodes experienced by a household in the
rural area. This reflects the presence of a more traditional agriculture in this area. Since
chronically poor households heavily rely in agricultural income -that in turns is heavily affected
by natural hazard- a higher impact of natural disasters is expected for them.
43
Table 11: Profile of households, according to poverty status (Consumption, 2002-2006)
(2002)
Several
Diff 1/
Never
One
Always
episode
(ppoor
episode
poor
Variable
s
value)
(1) vs
(1)
(2)
(3)
(4)
(4)
PERCENTEGES
Human capital
Gender of the hh (woman)
13.36
22.02
15.01
9.57
**
HH is literate
64.70
56.26
56.16
53.22
At least one children don't go to school
0.00
1.00
4.15
6.64
***
Low quality of dwelling's materials
15.48
21.52
26.37
19.88
Owner of house
82.38
86.80
82.79
86.76
***
Water: access to public network
58.48
55.28
34.73
34.16
**
Sewerage connected to public network
66.23
58.87
61.28
44.81
**
Electricity as lightning source
68.83
61.58
30.21
18.19
***
Telephone (fixed)
0.60
0.00
0.31
0.36
*
Welfare indicator
Poor [assets]
26.24
33.70
38.36
51.29
***
18.05
11.81
10.11
3.40
**
Remittances
Receive local remittances
23.65
27.11
31.87
24.55
Receive international remittances
0.34
0.00
1.46
0.00
Remittances (from at least one source)
23.65
27.11
33.33
24.55
0.00
0.00
0.00
0.00
Food assistance (at least one member)
Glass of milk
20.38
30.18
43.69
58.58
***
4.40
7.98
6.34
9.89
Schoolar breakfast
6.18
16.05
20.14
34.67
***
Other programm
4.14
2.76
7.65
13.09
***
Proportion of beneficiaries (as a proportion of total
14.49
21.30
26.30
33.68
***
members)
Welfare indicators
Monetary expenses (as % of total expenses)
70.85
64.17
56.94
53.86
***
Monetary income (as % of total income)
74.36
66.72
56.39
51.56
***
Participation in agricultural activities
Percentage of members that have as main activity
40.97
41.96
45.06
41.34
agriculture
Percentage of members that have as secondary activity
agriculture
5.97
8.07
9.87
8.65
Percentage of income from agricultural activities
25.15
26.54
36.09
41.06
***
1/ Mean differences are calculated by comparing column (1) with column (4)
44
In the ENAHO interview, households are asked to report if the have experienced any of the
shocks included in 8 different categories (see Table 12). Most of the households do not report
having experienced any negative episode, and these results are similar for the four categories of
the poverty transitions. Nonetheless, having suffered a robbery or an assault and a natural
disaster is significantly different for households classified as “Never Poor” in comparison with
households classified as “Always poor”. The latter tend to recall more being hit by a natural
disaster. This is consistent with the fact that these households obtain a higher proportion of their
income from agricultural activities.
Table 12: Shocks experienced by household in 2002
Never
poor
One
episode
Several
episodes
Variable
Shock: Loss of job
1.55
0.69
0.86
Shock: Bankruptcy of family business
0.60
2.40
0.49
Shock: Death of an income perceiver
0.00
0.00
0.96
Shock: Sickness or accident of a household
4.20
3.48
3.42
member
Shock: Abandonment of the head of the
household
0.00
0.00
0.29
Shock: Fire housing/business
0.00
0.00
0.29
Shock: Robbery, assault
5.50
3.03
7.23
Shock: Natural disaster
5.54
2.66
2.60
Shock: Other
0.00
1.73
0.52
Shock: None
84.17
87.40
83.57
1/ Mean differences are calculated by comparing column (1) with column (4)
5.2.3
Always
poor
Diff
(pvalue)
0.00
0.28
0.96
1.85
0.00
0.00
4.20
11.41
0.69
80.61
Poverty transitions, multivariate household regressions
A first approach to estimate the impact of hazard over poverty is to use the categories obtained
from the analysis poverty transitions as dependent variable. One can model probabilities of
entering; exiting, remaining or staying out of poverty based on status regression and then
establish whether a hazard may have a differentiated impact depending on the poverty transition
45
**
*
in turn. All the multinomial regressions estimate in this section use as poverty base status the
category “Never poor”.
For this specification we have estimated poverty transitions using two different measures.
First, we rely on the official estimates of poverty, following the INEI methodology that compares
the real monthly per capita consumption of each household with a predetermined poverty line
that is calculated valuating a basket of goods. Second, we measure a vector of assets by adding
the different number of durable goods (e.g. radio, TV, car, trunk) that households possess. We
use the median of the reported price of each item –in 2006- as a weight to be able to sum these
different items. In addition, a factor of depreciation is included to account for the age of the
objects, information that is also reported in the survey. A household is considered poor by assets
if the value of this vector is below the median value for all the rural households included in the
ENAHO in 2006 (not just the panel observations).
A. Poverty, measured as monthly per capita consumption
Table 13 shows the odd ratios of the multinomial regression for three different models that
include different controls. All models include controls for demographic composition (not
reported). These odd ratios –also know as risk ratio- are the ratio between the probability to
belong to each category and the probability to belong to the base category, given a unit increase
in the corresponding explanatory variable.
46
Table 13: Multinomial regression. Dependent variable: Poverty transitions, consumption (2002-2006)
Variables
One episode
Several episodes
Always Poor
Model 2 Model 3 Model 1 Model 2 Model 3
Model 1
Model 2
Shock: Natural disaster (yes=1)
0.526
0.508
1.170
1.224
1.098 2.443*** 2.696***
(0.225)
(0.216)
(0.359)
(0.388)
(0.345)
(0.772)
(0.893)
Total years of education (all members)
1.003
1.006 0.980***
0.985*
0.990 0.963***
0.976**
(0.008)
(0.009)
(0.008)
(0.008)
(0.009)
(0.008)
(0.009)
Gender of the hh (woman=1)
0.823
0.782
0.796
0.685
0.565
0.380** 0.304***
(0.392)
(0.363)
(0.295)
(0.263)
(0.225)
(0.171)
(0.138)
Main activity: agriculture 1/
1.524
1.400 5.774*** 5.128*** 3.471*** 14.998*** 10.570***
(0.706)
(0.683)
(2.499)
(2.225)
(1.553)
(7.487)
(5.372)
Secondary activity agriculture 1/
0.837
0.836
4.124*
3.863
2.953
3.543
3.134
(0.645)
(0.656)
(2.835)
(2.797)
(2.366)
(2.808)
(2.594)
Agricultural income (proportion of total income)
1.873
1.777
1.780
2.195
1.851
2.783
3.911*
(1.129)
(1.073)
(0.907)
(1.152)
(0.997)
(1.538)
(2.235)
0.990** 0.989**
0.993* 0.992**
0.988***
(0.005)
(0.005)
(0.004)
(0.004)
(0.004)
Vector of assets
1.000
1.000
1.000** 1.000***
0.999**
0.000
0.000
0.000
0.000
0.000
Water: access to public network (yes==1)
0.652
0.543***
(0.170)
(0.119)
Telephone (fixed) (yes==1)
0.000
1.972
(2.955)
Electricity as lightning source (yes==1)
0.787
0.303***
(0.263)
(0.083)
Region fixed effects (domain)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Demographic controls
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
N
4086
3865
3865
4086
3865
3865
4086
3865
Exponentiated coefficients, Standard errors in parenthesis
Source: ENAHO 2002-2006: Balanced panel
* p<0.10, ** p<0.05, *** p<0.01
1/ Number of individuals as a proportion of total members
Model 1
0.478*
(0.208)
0.998
(0.007)
0.924
(0.421)
1.517
(0.694)
0.864
(0.645)
1.314
(0.777)
47
Model 3
2.285**
(0.745)
0.983*
(0.010)
0.230***
(0.109)
6.582***
(3.517)
2.060
(1.839)
3.067
(1.798)
0.986***
(0.005)
1.000**
0.000
0.412***
(0.102)
11.293*
(16.546)
0.170***
(0.054)
Yes
Yes
3865
According to model 3, the probability of being “Always Poor” is 2.29 times the
probability of being “Never poor”, given that the household experienced a natural disaster.
Similarly, the probability of being “Always Poor” is 6.58 times the probability of being “Never
poor”, given a unit increase in the proportion of member of the household that have as
agriculture as main activity. The variable shock is not statistically significant for any of the other
two categories.
By contrast, an increase in livestock possessions, as well as, the access to piped water and
electricity makes a household more likely to be “Never Poor” than to be “Always Poor”. The
same results hold if we compare households that have been classified as “Never Poor” with
households that have experienced “Several episodes” of poverty.
Use as a dependent variable a set of categories that does not change over time imposes
some restrictions for the estimation of a multinomial regression using a panel database. Under
this specification the data is pooled and the time dimension is not taking into account. A
household classified as “Never poor” is considered as five different observations (one for each
year) with the same value of the dependent variable. To circumvent this restriction, we reshape
the data in order to estimate a multinomial regression that considered the information for each
household in each year. Then, we control the four categories of poverty transitions by having
experienced a shock in different years and some additional controls that capture the initial
conditions of our sample. Notice that we exclude the year 2002 for the estimation of the poverty
transitions in order to use the information of that year as the initial conditions.
The coefficients shown in Table 14 are consistent with the results report in Table 13. In
this specification, the probability of being “Always Poor” is 4.75 times the probability of being
“Never poor”, given that household experienced a shock in 2004. Conversely, a household is
more likely to be “Never poor” –than to be “Always Poor”- given an increase in livestock and in
the access to public services (piped water, telephone and electricity).
[Note: This specification can be modified to replace the shock dummies for each year for a
variable that captures the number of years that a household report having experienced a shock]
48
Table 14: Multinomial regressions. Dependent variable: Poverty transitions, consumption
(2003-2006)
Variables
Shock: Natural disaster (yes=1) [2002]
Total years of education of the members of the household [2002]
Gender of the hh (woman=1) [2002]
Number of plots [2004]
Main activity: agriculture [2002] 1/
Secondary activity agriculture [2002] 1/
Agricultural income (proportion of total income) [2002]
Livestock (on sheep equivalences) [2002]
Vector of assets [2002]
Water: access to public network (yes=1) [2002]
Telephone (fixed) (yes=1) [2002]
Electricity as lightning source (yes==1) [2002]
N
* p<0.10, ** p<0.05, *** p<0.01
One
episode
0.192**
(0.158)
0.985
(0.650)
1.356
(1.044)
0.393
(0.238)
0.360
(0.237)
1.001
(0.010)
0.801
(0.389)
1.231
(0.182)
1.045
(0.691)
2.335
(2.543)
1.878
(1.344)
0.994
(0.006)
1.000
0.000
0.517 *
(0.187)
0.000***
0.000
0.432**
(0.153)
771
Several
episodes
0.214**
(0.165)
1.546
(0.866)
3.123
(2.223)
0.921
(0.460)
0.621
(0.300)
0.983*
(0.009)
0.607
(0.267)
1.382***
(0.168)
1.247
(0.771)
4.120
(4.439)
1.611
(0.950)
0.992
(0.005)
1.000
0.000
0.531*
(0.164)
0.000***
0.000
0.334***
(0.170)
771
Always
poor
0.423
(0.311)
3.069*
(1.784)
4.753**
(3.543)
1.312
(0.632)
1.422
(0.720)
0.960***
(0.011)
0.274***
(0.137)
1.250*
(0.156)
2.197
(1.681)
2.683
(3.189)
1.914
(1.306)
0.985***
(0.005)
1.000**
0.000
0.419**
(0.147)
0.000***
0.000
0.163***
(0.630)
771
B. Poverty, measured as possession of assets
As a second approach, we estimate the same specification report in section A, but using the
possession of assets as a measure of poverty. Similarly to the prior models, all specifications
include controls for demographic composition (not reported). In both specifications the variable
of natural disaster does not have a statistically significant effect over any of the categories of the
poverty transitions (see Table 15 and Table 16).
49
Table 15: Multinomial regression, Dependent variable: Poverty transitions: assets (2002-2006)
Variables
One episode
Several episodes
Always Poor
Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Shock: Natural disaster (yes=1)
1.086
1.102
0.971
0.912
0.942
0.792
0.699
0.746
0.599*
(0.260)
(0.274)
(0.248)
(0.180)
(0.187)
(0.165)
(0.175)
(0.193)
(0.164)
Total years of education (all members)
0.988*
0.990
0.991 0.955*** 0.957*** 0.960*** 0.899*** 0.899*** 0.905***
(0.007)
(0.007)
(0.007)
(0.008)
(0.008)
(0.008)
(0.012)
(0.012)
(0.012)
1.279
1.253
1.209
1.668
1.600
1.529 3.595*** 3.477*** 3.335**
(0.528)
(0.521)
(0.526)
(0.568)
(0.555)
(0.558)
(1.368)
(1.345)
(1.390)
Main activity: agriculture 1/
2.278*
2.208*
1.691 6.197*** 6.480*** 4.600*** 8.835*** 8.668*** 5.951***
(0.976)
(0.954)
(0.739)
(2.056)
(2.180)
(1.651)
(3.411)
(3.416)
(2.489)
Secondary activity agriculture 1/
3.727*** 3.726***
2.710*
1.920
2.009
1.323
1.590
1.572
0.913
(1.863)
(1.899)
(1.405)
(0.759)
(0.810)
(0.553)
(0.741)
(0.752)
(0.462)
Agricultural income (proportion of total income)
2.046* 2.466**
2.216*
1.230
1.449
1.220
1.845* 2.346**
1.854
(0.862)
(1.077)
(0.958)
(0.411)
(0.493)
(0.421)
(0.682)
(0.921)
(0.750)
0.995
0.995
0.995*
0.994*
0.993
0.992
(0.003)
(0.003)
(0.003)
(0.003)
(0.005)
(0.006)
Water: access to public network (yes==1)
0.930
0.587**
1.016
(0.187)
(0.105)
(0.240)
Telephone (fixed) (yes==1)
1.737
3.604
0.000***
(2.779)
(6.451)
0.000
Electricity as lightning source (yes==1)
0.381***
0.329***
0.108***
(0.100)
(0.070)
(0.035)
Region fixed effects (domain)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Demographic controls
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Number of observations
4086
3865
3865
4086
3865
3865
4086
3865
3865
* p<0.10, ** p<0.05, *** p<0.01
50
Variables like the participation of the members in agricultural activities and access to
public services have the same direction found in the specifications that use consumption to
measure poverty. One can argue that natural disaster affect these households through its negative
effect over the agriculture activity, affecting the level and variability of their income, but not
their possessions of durable goods.
Table 16: Multinomial regressions. Dependent variable: Poverty transitions, assets (20032006)
Variables
Total years of education of the members of the household [2002]
N
* p<0.10, ** p<0.05, *** p<0.01
51
One
episode
0.846
(0.465)
1.578
(0.845)
1.193
(0.509)
0.720
(0.283)
0.499
(0.218)
0.985*
(0.009)
1.353
(0.683)
1.089
(0.085)
1.496
(1.043)
1.112
(1.025)
1.732
(0.936)
0.995
(0.004)
0.756
(0.222)
11.643**
(14.116)
0.258***
(0.087)
771
Several
episodes
0.800
(0.383)
1.799
(0.769)
0.991
(0.380)
0.811
(0.363)
0.760
(0.310)
0.962***
(0.011)
2.182*
(0.934)
1.105
(0.082)
3.900**
(2.231)
1.284
(0.893)
0.686
(0.329)
0.996
(0.004)
0.399***
(0.108)
0.000***
0.000
0.454***
(0.133)
771
Always
poor
0.769
(0.586)
1.304
(0.690)
1.025
(0.426)
0.446*
(0.192)
0.549
(0.248)
0.919***
(0.013)
4.138***
(1.881)
1.035
(0.081)
6.013***
(3.702)
0.742
(0.607)
0.795
(0.444)
0.989**
(0.006)
1.059
(0.339)
0.000***
0.000
0.085***
(0.032)
771
5.2.4
Change in per capita consumption
[TO BE DONE]
5.2.5
Analysis at the bottom of the distribution
As we have mentioned lines above, it is possible that the report of natural disasters is biased to
households that are poorly endowed and less integrated to the market. To circumvent this
problem and to analyze the impact of natural disaster ate the bottom of the income distribution
we have estimated a Quantile regression. We use as dependent variable the (log) monthly per
capita consumption in 2006. We add dummies for the shock reported in each year as well as
some additional controls of the prior period. This model also use controls for demographic
composition (not reported). In addition, we have included the variable “plots” that captures the
number of plots worked by a household. This variable was included to the ENAHO
questionnaire just in 2004.
The reported coefficients show the median of each variable in the corresponding
percentile. That is why the column that shows the results for the whole sample is equal to the
column shows the results for the 50th percentile. The constant term captures the median of the
dependent variable if all control variables are set to 0. This constant term is use to compare the
coefficients corresponding to each the explanatory variable. For instance, the variable “Shock:
Natural disaster” in 2002 decrease the median monthly per capita consumption in 0.28 logarithm
points. In other word, having experienced a shock in 2002 reduces the monthly per capita
consumption of the bottom 25th of the distribution in 3.85% 2 . It also reduces the monthly per
capita consumption of the 50h of the distribution, but in a lower percentage (2.68%).
2
This result is obtained after exponentiating the value in logs.
52
Table 17: Quantile regression, Dependent variable: (log) Monthly per capita consumption
(2006) Controls: 2005
Variables
Total
Constant
3.853***
(0.170)
-0.236***
(0.072)
0.134**
(0.053)
-0.108**
(0.046)
-0.091*
(0.048)
-0.226***
(0.051)
0.006***
(0.001)
0.264***
(0.059)
-0.024**
(0.010)
-0.169**
(0.067)
-0.280***
(0.086)
-0.205***
(0.069)
0.002***
(0.001)
0.000***
Total years of education (all members) [2005]
N
Standard errors in parenthesis
* p<0.10, ** p<0.05, *** p<0.01
53
0.079**
(0.034)
0.872***
(0.179)
0.138***
(0.034)
771
25th
Percentile
3.538***
(0.165)
-0.283***
(0.077)
0.041
(0.056)
-0.048
(0.048)
-0.191***
(0.047)
-0.091*
(0.049)
0.008***
(0.001)
0.103*
(0.057)
-0.031***
(0.009)
-0.216***
(0.071)
-0.185*
(0.099)
-0.14*
(0.072)
0.002***
0.000**
50th
Percentile
3.853***
(0.170)
-0.236***
(0.072)
0.134**
(0.053)
-0.108**
(0.046)
-0.091*
(0.048)
-0.226***
(0.051)
0.006***
(0.001)
0.264***
(0.059)
-0.024**
(0.010)
-0.169**
(0.067)
-0.280***
(0.086)
-0.205***
(0.069)
0.002***
(0.001)
0.000***
75th
Percentile
4.126***
(0.188)
-0.291***
(0.077)
0.058
(0.064)
-0.119**
(0.054)
-0.138**
(0.054)
-0.177***
(0.057)
0.003***
(0.001)
0.363***
(0.063)
-0.030***
(0.011)
-0.144*
(0.076)
-0.186*
(0.099)
-0.008
(0.087)
0.003***
(0.001)
0.000***
0.013
(0.034)
0.437***
(0.138)
0.187***
(0.036)
187
0.079**
(0.034)
0.872***
(0.179)
0.138***
(0.034)
381
0.129***
(0.038)
0.584***
(0.136)
0.155***
(0.039)
582
To sum up
•
There is a higher report of natural disasters from households that have less access to public
services, less integrated to the market, and with a higher proportion of agricultural income.
•
Two different specifications of multinomial regressions show that households are between
2.3 and 4.8 times more likely to be “Always Poor” than to be “Never Poor” given that they
have experienced a natural disaster. This results only hold if consumption, rather than assets,
is use to measure poverty.
o
One can argue that natural disaster affect these households through its negative
effect over the agriculture activity, affecting the level and variability of their
income, but not their possessions of durable goods.
•
Having experienced a natural disaster in the past have a negative effect over the monthly per
capita consumption. This result is stronger when analyzed at the bottom of the distribution.
6
Conclusions
54
References
Alderman, H., Hoddinott, J. and Kinsey, B. H (2006). Long term consequences of early
childhood malnutrition. Oxford Economic Papers, 58(3), 450-474.
Alpizar, C. A (2007). Risk coping strategies and rural household production efficiency: quasiexperimental evidence from El Salvador. PhD Thesis, Ohio State University.
Arellano, M., and S. Bond. (1991). “Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations.” Review of Economic Studies 58: 277–97.
Auffret, P. (2003). High consumption volatility: The impact of natural disasters. World Bank
Policy research Working Paper 2962. World Bank, Washington.
Benson, C. and Clay, E.(2003). Economic and financial impact of natural disasters: An
assessment of their effects and options for mitigation. London, Overseas Development Institute.
Blundell, R., and S. Bond. (1998). “Initial conditions and moment restrictions in dynamic panel
data models.” Journal of Econometrics 87: 11–143.
Carter, M. R., Little, P., and Mogues, T. (2007). Poverty traps and natural disasters in Ethiopia
and Honduras. World Development, 35(5), 835-856.
Crowards, T. (2000). Comparative vulnerability to natural disasters in the Caribbean. Caribbean
Development Bank Research Paper 1/00.
Dercon, S. (2004). Growth and shocks: evidence from rural Ethiopia. Journal of Development
Economics, 74(2), 309-329.
Dercon, S., Hoddinott, J. y Woldehanna, T. (2005). Shocks and consumption in 15 Ethiopian
villages 1999-2004. Journal of African Economies, 14(4), 559-585.
De la Fuente, A. et al. (2008). Assessing the Relationship between Natural Hazards and Poverty:
A Conceptual and Methodological Proposal. Document Prepared for ISDR-UNDP Disaster RiskPoverty Regional Workshops in Bangkok, Thailand (22-24 April 2008) and Bogotá, Colombia
(10-11 June, 2008)
De Janvry, A., Finan, F., Sadoulet, E (2004). Can conditional cash transfers serve as safety net to
keep children at school and out of the labor force?. Paper 99, Dept of Agricultural Economics,
University of California Berkeley.
55
De Janvry, A., Sadoulet, E., Salomón, P., and Vakis, R. (2006). Uninsured risk and asset
protection: can conditional cash transfer programs serve as safety nets? SP Discussion Paper No
0604. World Bank, Washington.
Heger, M., Julca, A. and Paddison, O. (2008). Analysing the impact of natural hazards in small
economies. UNU-WIDER Research Paper 2008/25. UNU-WIDER, Helsinki.
Jaramillo, C (2007). Natural disasters and growth: evidence using a wide panel of countries.
Documento CEDE 2007-14. Bogotá.
Lindell, M. K. and Prater, C. S. (2003). Assessing community impacts of natural disasters.
Natural Hazards Review, 4(4), 176-185.
Moser, C. (1998). The assets vulnerability framework: reassessing urban poverty strategies.
World Development, 26(1), 1-19.
Sawada, Y. (2006). The impact of manmade disasters on household welfare. Paper presented at
the International Association of Agricultural Economists Conference, Australia.
Sawada, Y. and Shimizutani, S. (2004). How do people cope with natural disasters Evidence
from the Great-Hanshin-Awaji earthquake. ESRI Discussion Paper 101. Tokyo.
Toya, H. and Skidmore, M. (2005). Economic development and the impact of natural disasters.
University of Wisconsin Whitewater Working Paper 05-04. Wisconsin.
Vatsa, K. and Krimgold, F (2000). Financing disaster mitigation for the poor. In A. Kreimer, and
M. Arnold (eds), Managind Disaster Risk in Emerging Economies. World Bank, Washington.
56
Appendix 1: Descriptive statistics (Unbalance panel)
Variable
Human capital
Age of the hh
Education of the hh: equal or lower than complete primary (yes=1)
HH is literate (yes=1)
At least one children don't go to school (yes-=1)
Low qualitity of dwelling's materials (yes=1)
Owner of house (yes=1)
Number of rooms use to sleep
Water: access to public network (yes=1)
Sewerage connected to public network (yes=1)
Electricity as lightning source (yes=1)
Telephone (fixed) (yes=1)
Welfare indicators
Poor [consumption] (yes=1)
Poor [assets] (yes=1)
Monetary expenses (as proportion of total expenses)
Monetary income (as proportion of total income)
Received credit from any source (year 2004-2006) (yes=1)
Remittances
International Remittances (yes=1)
Local Remittances (yes=1)
Remittances (at least one source)
Food assistance (at least one member, yes=1))
Glass of milk
Schoolar breakfast
Other programm
Assets
Vector of assets
Number of plots (2004-2006)
57
Obs
Mean
Sd
Min
Max
8411
8400
8410
8411
8411
8411
4105
8411
49.58
0.28
4.45
19.31
4.52
0.16
0.57
0.11
16.25
0.45
3.13
15.63
4.65
0.36
0.49
0.31
14
0
0
0
0
0
0
0
96
1
17
131
18
1
1
1
8369
8410
4373
8411
8411
8411
8411
0.15
0.85
1.60
0.41
0.59
0.37
0.01
0.36
0.35
1.13
0.49
0.49
0.48
0.09
0
0
0
0
0
0
0
1
1
7
1
1
1
1
8382
8411
8411
8411
8411
2.87
0.60
0.43
0.56
0.55
1.84
0.49
0.49
0.23
0.26
1
0
0
0
0
13
1
1
1
1
4375
8411
0.39
0.10
0.49
0.30
0
0
1
1
8411
8411
8411
8411
8411
0.25
0.01
0.00
602.16
43.18
0.43
0.08
0.03
2115.52
828.00
0
0
0
0
0
1
1
1
43176
33660
8411
8411
8411
8411
8411
0.37
0.07
0.19
0.10
0.26
0.48
0.25
0.39
0.30
0.29
0
0
0
0
0
1
1
1
1
1
7916
8411
8411
17.77
863.57
1.04
29.03
2813.15
2.02
0
0
0
530
57795.7
20
Appendix 2: Descriptive statistics (Balance panel)
Variable
Human capital
Age of the hh
Education of the hh: equal or lower than complete primary (yes=1)
HH is literate (yes=1)
At least one children don't go to school (yes-=1)
Low qualitity of dwelling's materials (yes=1)
Owner of house (yes=1)
Number of rooms use to sleep
Water: access to public network (yes=1)
Sewerage connected to public network (yes=1)
Electricity as lightning source (yes=1)
Telephone (fixed) (yes=1)
Welfare indicators
Poor [consumption] (yes=1)
Poor [assets] (yes=1)
Monetary expenses (as proportion of total expenses)
Monetary income (as proportion of total income)
Received credit from any source (year 2004-2006) (yes=1)
Remittances
International Remittances (yes=1)
Local Remittances (yes=1)
Remittances (both sources)
Food assistance (at least one member, yes=1))
Glass of milk
Schoolar breakfast
Other programm
Assets
Vector of assets
Number of plots (2004-2006)
58
Obs
Mean
Sd
Min
Max
4150
4147
4150
4150
4150
4150
2025
4150
48.99
0.29
4.57
20.72
4.65
0.13
0.58
0.03
15.47
0.46
3.06
16.14
4.65
0.34
0.49
0.17
15
0
0
0
0
0
0
0
94
1
17
131
18
1
1
1
4150
4150
2265
4150
4150
4150
4150
0.18
0.87
1.67
0.39
0.59
0.40
0.00
0.39
0.34
1.07
0.49
0.49
0.49
0.07
0
0
0
0
0
0
0
1
1
7
1
1
1
1
4144
4150
4150
4150
4150
2.86
0.61
0.39
0.57
0.57
1.78
0.49
0.49
0.22
0.25
1
0
0
0
0
13
1
1
1
1
4375
4375
0.39
0.10
0.49
0.30
0
0
1
1
4375
4375
4375
4375
4375
0.25
0.00
0.25
609.97
20.05
0.43
0.07
0.44
2190.52
491.70
0
0
0
0
0
1
1
1
43176
33660
4375
4375
4375
4375
4375
0.43
0.08
0.22
0.12
0.28
0.50
0.27
0.42
0.32
0.29
0
0
0
0
0
1
1
1
1
1
4375
4375
4375
18.59
943.48
1.01
30.89
2740.07
1.74
0
0
0
366
39777.7
20
Appendix 3: Quantile regression, Dependent variable: (log) Monthly per capita
consumption (2006). Controls: 2002
Variables
Total
Constant
3.932***
(0.181)
-0.272**
(0.087)
0.01
(0.064)
-0.082
(0.058)
-0.069
(0.059)
-0.189**
(0.062)
0.009***
(0.001)
0.069
(0.064)
-0.044***
(0.011)
25th
Percentile
3.591***
(0.160)
-0.179*
(0.077)
0.052
(0.055)
-0.017
(0.050)
-0.138*
(0.054)
-0.163**
(0.050)
0.009***
(0.001)
0.041
(0.057)
-0.031***
(0.008)
0.057
(0.084)
-0.112
(0.116)
-0.086
(0.074)
0.002**
(0.001)
0
0.000
0.05
(0.040)
-0.343
(0.191)
0.246***
(0.044)
771
-0.093
(0.075)
-0.115
(0.105)
-0.102
(0.062)
0.001*
0.000
0.000*
0.000
0.076*
(0.036)
0.085
(0.108)
0.170***
(0.038)
771
Total years of education (all members) [2002]
N
Standard errors in parenthesis
* p<0.05, ** p<0.01, *** p<0.001
59
50th
Percentile
3.925***
(0.234)
-0.265*
(0.113)
0.004
(0.083)
-0.07
(0.074)
-0.086
(0.076)
-0.188*
(0.080)
0.009***
(0.002)
0.036
(0.083)
-0.042**
(0.016)
-0.008
(0.016)
0.085
(0.109)
-0.126
(0.150)
-0.086
(0.096)
0.002*
(0.001)
0
0.000
0.058
(0.052)
-0.368
(0.248)
0.252***
(0.057)
771
75th
Percentile
4.235***
(0.276)
-0.176
(0.113)
0.002
(0.082)
-0.053
(0.079)
-0.186*
(0.078)
-0.079
(0.078)
0.007***
(0.002)
0.178*
(0.081)
-0.019
(0.017)
-0.004
(0.017)
0.053
(0.131)
-0.072
(0.188)
-0.186
(0.111)
0.002*
(0.001)
0.000***
0.000
0.147**
(0.057)
0.05
(0.237)
0.169**
(0.061)
771

Desastres naturales

Transcription

Similar documents

Useful advises to help you protect your Child From domestic

Life changes, the household changes Useful advises to help you

Diptico de la Comarca de Guadix

St. Therese

Ad, 8.625x11.125, Laverson, RanchandCoast

Enero 3, 2016 Sagrada Familia - St. Anthony of Padua Catholic

After Party Flyer

Oportunidades Negocios RD Feb 2012 \(2\) \(3era Parte\)\(1\)

June 19, 2016 - St. Bartholomew Catholic Church