Sources of Welfare Disparities Within and Between Regions in Latin America and Caribbean Countries

Transcription

1 Sources of Welfare Disparities Within and Between Regions in Latin America and Caribbean Countries Vol. 1: Synthesis Draft: December 1, 2008 Emmanuel Skoufias Gladys Lopez-Acevedo THE WORLD BANK Washington, D.C.

2 CURRENCY EQUIVALENTS Currency Unit = Real (R$) US$1 = R$2.15 (March 2006) Fiscal Year December 31 January 1 ACCRONYMS AND ABBREVIATIONS CCT Conditional Cash Transfers ECV ENAHO ENCASEH ENCOVI ENIGH ENNVIH Survey of Living Conditions (Encuesta de Condiciones de Vida) Encuesta Nacional de Hogares Encuesta de Características Socioeconómicas de los Hogares Encuesta sobre las Condiciones de Vida Encuesta de Ingresos y Gastos de los Hogares Encuesta Nacional sobre Niveles de Vida de loa Hogares IBGE ICBF Instituto Brasileiro de Geografia e Estatística Colombian Institute for Family Welfare (Instituto Colombiano de Bienestar Familiar) INE INEP INESPRE INP INSS ISS National Institute of Statistics Instituto Nacional de Estudos e Pesquisas Educacionais Instituto de Estabilización de Precios Institute of Social Security Normalization (Instituto de Normalização Previsional) National Social Security Institute (Instituto Nacional de Seguridade Social) Social Security Institute (Instituto de Seguridade Social)

3 PNAD POF PPP PROGRESA Pesquisa Nacional Por Amostra de Domicílios Household Budget Survey (Pesquisa de Orçamentos Familiares) Purchasing Price Parity Program for Education, Health and Food (Programa de Educación, Salud y Alimentación) Vice President Chief Economist Sector Directors Sector Managers Task Managers Pamela Cox Jaime Saavedra, Louise Cord Emmanuel Skoufias and Gladys Lopez Acevedo 3

4 Executive Summary In spite of the economic growth experienced by Latin American countries, poverty is still high and tends to be geographically concentrated. Two competing perspectives may be associated with the underlying determinants of spatial differences in welfare: concentration vs. geography. The concentration hypothesis posits that poor areas arise from the persistent concentration in these areas of individuals with personal attributes that inhibit growth in their living standards (e.g. low education). The policy prescription that emerges from the concentration hypothesis is about enhancing individual production-related characteristics such as education and health. The geography hypothesis, on the other hand, asserts that the residential location of households is the primary cause of the high level of poverty and weak growth of living standards over time. According to this view, given two identical individuals, the one living in an area with lower access to infrastructure and other basic services (electricity, water and sanitation) may be condemned to stagnation and poverty over time due to the lower returns earned on their assets. Basic intuition, ex-ante, would suggest that the concentration effect will be a prominent explanation for the differences in welfare across regions as migration is likely to equalize of returns to a given set of (observable) characteristics across regions within a country. However, the role of migration in equalizing returns to characteristics may be limited by a number of factors such as the monetary and psychic costs of migration (transportation, strong social ties and cultural values, such as the attachment of the indigenous peoples to land), and uncertainty about the benefits from migration (low probability of getting a high paying job, risk of unemployment, etc.). Another important factor is the presence of agglomeration effects brought to prominence by the new economic geography. In these models, as labor migrates in response to a real wage differential, the size of the market grows in the destination region, and, through a variety of mechanisms, related to scale economies, the real wage in the destination region increases rather than decreases. In light of these considerations, the answer to the question of whether it is the concentration or the geography effect that is the dominant explanation for disparities in welfare across space becomes an empirical issue. Yet, most of the policies and programs for poverty alleviation targeted either at poor people or at poor areas have been based more on ideological and political considerations and less on empirical evidence. A major part of the problem to date has been the absence of any empirical evidence on these issues that could directly inform and facilitate decision making. In this study we begin to provide some empirical evidence on these critical issues. We use household survey data from eight countries (Bolivia, Brazil, Colombia, Ecuador, Guatemala, Mexico, Nicaragua, and Peru) to provide a diagnostic analysis (x-ray view) of the main determinants of spatial differences in welfare within and across regions in LAC countries. Using Oaxaca s (1973) decomposition method, we examine whether the primary explanation for the differences in living standards is differences in the characteristics of households living in the poorer or richer areas, or differences in the returns to these characteristics between poorer and richer areas. Our study reveals a rich picture. Within regions, differences in welfare levels between urban and rural areas are associated primarily with differences in characteristics. This is consistent with notion that the migration of labor from rural to urban areas within regions manages to equalize returns to individual attributes within regions. Therefore, welfare differences between urban and rural regions

5 seem to be primarily due to the sorting or concentration of people with higher attributes in the urban areas of these regions. However, when it comes to explaining differences in welfare levels between urban (or rural) areas of two different regions in any country it is differences in returns that tend to become the primary explanation. In terms of policy prescription, the results of the present analysis suggest that governments can help poor people the most by investing in their human capital and in assets that people can carry with them wherever they decide to go. This includes investing in the quantity and quality of such assets. However, the results also revealed that returns across regions may be substantially different and that migration may not be enough to equalize returns across regions. In those cases, governments can follow two paths: One is to try to eliminate the impediments to labor mobility and facilitate the movement of people towards the areas of economic opportunity. This means ensuring that all markets land, labor, credit, insurance, goods, and services operate efficiently everywhere. It also means providing every person with basic services and human capital so they have equal opportunity to succeed economically regardless of their place of birth. To the extent that markets operate differently across space and that basic services and human capital are also being provided differently across space, eliminating the impediments to labor mobility will imply spatially targeted interventions. However, it is important to highlight that the aim of these interventions is to level the institutional playground throughout a country s territory. In cases where leveling the playing field will not be enough to raise the livelihoods of its inhabitants e.g. because due to cultural attachment the people living there will never migrate to more dynamic areas governments can also resort to investing directly in the less favored areas and create special incentives there so as to raise the levels of welfare and economic potential of those places. While evidence on the effectiveness of programs focused on the individual is extensive, there is very little evidence on the effectiveness of territorial development programs. This study carried out a thorough comparative assessment of four territorial programs and found little evidence on their effectiveness. This is partly due to the paucity of adequate data with which to carry out a rigorous evaluation. However, there is evidence suggesting that these programs have not been very effective either because of insufficient funds or because their ambitious scope has made them extremely complicated to administer successfully JEL Classifications: Key words: Inequality, 5

6 Acknowledgements The report is the product of a research program led by Emmanuel Skoufias and Gladys Lopez Acevedo with analysis and contributions from Roy Katayama, Hector Valdes Conroy, Francisco Haimovich, Penelope Brown, Erika Strand, Monica Tinajero, Guillermo Rivas, Javier Escobal, and Jorge Mario Soto. The team is grateful for funding and support from the Regional Studies Program of the Office of the Chief Economist, with support from the Poverty Reduction and Economic Management (PREM) Poverty Unit in the anchor and the PREM Department of the Latin America and Caribbean Regional Office of the World Bank. We would like to thank our peer reviewers, Martin Ravallion, Paul Dorosh, John Nash, Somik Lall, Gabriel Demombynes, and Alain De Janvry, for their comments and suggestions on this research program. The team also gratefully acknowledges the following people for the support and comments on various drafts: Tito Cordella, Augusto de la Torre, Ana Revenga, Louise Cord, Jaime Saavedra, Pierella Paci, Farhad Shilpi, Kenneth Simler and Ambar Narayan. We are also grateful to the individuals who provided comments and suggestions at the annual meeting of NIP and LACEA in Rio de Janeiro, Brazil. The findings, interpretations and conclusions expressed in this paper are entirely those of the authors, and do not necessarily represent the opinions of the World Bank, its Board of Directors or the countries it represents. Contacts: eskoufias@worldbank.org, gacevedo@worldbank.org vi

7 Chapter 1: Introduction and Overview Middle Income Countries (MICs) in Latin America (LAC) continue to struggle to achieve growth that is also shared by the poor segments of their population despite significant progress in key development indicators. A striking fact has been the resilience of inequality in the region over the longer term. For example, between 1996 and 2000, GDP per capita in Mexico increased by 9.7 percent in real terms but the incomes of the poorest 30 percent of the population contracted during this period. The increase, which was the highest the country had experienced in 16 years, was characterized entirely by higher incomes for the richest 30% of the population. Similarly, Chile, who s real GDP per capita expanded by more than 30 percent between 1992 and 1996, experienced an increase in its Gini index of 7 percent in that same period. (World Bank 2005, 2004, Attanasio and Szekely 2001, Mision de Pobreza, 2006). The progress towards achieving the first Millennium Development Goal of eradicating extreme poverty by 2015 is still far from being achieved in most Latin American countries. The share of the population living on less than 2 dollars a day in Latin America has been around 25% since the mid 90s with little reduction in MIC countries such as Peru, Colombia and Mexico (World Bank, 2005). Another feature common in many countries in LAC is geographic disparities in living standards that continue to persist over time in spite of overall economic growth. These differences in the standard of living in different areas or regions within LAC countries have always been a cause of concern to policy makers. In response to these concerns governments continue to support and develop programs that are aimed to poor people and to poor areas. Figure 1 Figure 1 shows just how much poverty rates vary within countries in Latin America and the Caribbean. 1 For each country included in the graph, the left end of the horizontal line represents the lowest poverty rate 1 Note that this graph is derived using an absolute poverty line (US $2 / day PPP). This report will later discuss regional disparities and provide information and graphs that use consumption as a measure of welfare as well as country-specific national poverty lines. These indicators are better measures of welfare within a given country though it will inhibit our capacity to make comparisons across countries. 1

8 observed among the country s regions; the right end of the line corresponds to the highest regional poverty rate; while the dot represents the country s overall poverty rate. In Bolivia, Honduras, Mexico, Paraguay, and Peru, the difference in poverty rates across regions is more than 40 percentage points. At the most extreme, some regions in Peru have poverty rates under 10 percent while others hover above 70 percent. In Brazil, poverty rates are higher in rural areas than in urban areas with very high rates in the North and North East. In the poorest region in Brazil, income per capital was barely 10 percent of the richest. Governments in the region have responded proactively to these regional inequalities countries in their countries introducing diverse poverty alleviation programs. Many of these programs are targeted towards poor regions, while other programs are targeted towards poor people. Brazil s effort to deal with the longstanding income disparities between the Southeast and the North and Northeast provide an illustrative example of the issues so far raised. Since the 1960s, the policy emphasis in Brazil was targeted to poor areas rather than people. It consisted of tax incentives and public credits to subsidize private initiatives, direct governments investment in infrastructure and related territorial development programs in the Northeast. This managed to hasten industrialization but failed to reduce poverty or redistribute income in the North and Northeast. In the last decade the Brazilian government shifted its emphasis towards a program targeted to poor people instead of poor areas. The Bolsa Familia conditional cash transfer program aims to invest in the health and education of the members of poorer households in the country as a means for poverty alleviation in the country as well as a means of reducing regional disparities in welfare. There is clearly a lack of consensus about the most effective approach to reducing regional inequalities. Surely, this is a very complicated issue and it probably has a nuanced answer. The lack of consensus and the shift in perspectives within a span of two years is even reflected in the two most recent World Development Reports (WDR) that have emphasized different aspects on the issue of regional disparities in welfare. The WDR 2009 Reshaping Economic Geography (World Bank 2008) argues that spatial disparities increase in the early stages of development, and then diminish as countries reach high income status. Drawing from the development experiences of the OECD countries and upper-middle income countries, it also argues that different dimensions of well-being converge at different speeds. Essential household consumption is hypothesized to converge first, although that is expected to occur only after a majority of a country s population is in urban areas. Next to converge is access to basic public services such as basic education, health services, drinking water, and sanitation. Ruralurban disparities in these basic services can be expected to persist until a country reaches uppermiddle income levels. Last to converge are wages and incomes. A central message of the WDR 2009 is that spatial disparities in production and income are inevitable, and that efforts to spread [growth] prematurely will jeopardize progress. Specific policies should therefore focus more on facilitating the drivers of growth and less on spatial inequality, for example, providing minimal basic services in lagging regions so that the motives for migration are productivity related. In contrast, the WDR 2008 Agriculture for Development (World Bank 2007) outlines the role of agriculture as an essential contributor to economic growth and poverty reduction, particularly in the world s poorest countries. It argues that growth in agriculture reduces poverty more rapidly than 2

9 growth in other sectors, and makes the case for investments in agriculture as a means of reducing regional (rural-urban) disparities in well being. This lack of consensus may be attributed to two alternative perspectives about the determinants of spatial differences in welfare: concentration vs. geography. The concentration hypothesis posits that poor areas arise from the persistent concentration in these areas of individuals with personal attributes that inhibit growth in their living standards (e.g. low education). According to this view, otherwise identical individuals will have the same growth prospects independently of where they live. A critical element of this hypothesis is migration. Free migration induced by income differentials between regions will, holding all else equal, serve to reduce any wage premium that may be associated with living in a specific geographic region where labor might have been relatively scarce initially. The policy focus that then emerges centers on enhancing individual production-related characteristics (such as education and health). Classical examples of such programs include the conditional cash transfer programs such as the Oportunidades program in Mexico, the Bolsa Familia program in Brazil, the Familias en Accion program in Colombia, the PRAF program in Honduras, Red de Proteccion Social in Nicaragua, and the JUNTOS program in Peru (among others). The key characteristic of these programs is that they typically involve targeting at the household level and aim to alleviate poverty through monetary and in-kind benefits that encourage investments in education, health and nutrition. Beneficiaries of these programs, are free to migrate to the areas (typically urban) where their welfare may be higher, thus ultimately contributing to the migration of people out of the poorer areas and the ultimate equalization of welfare across space. The alternative hypothesis is that the geographic location is the primary cause of the high level of poverty and weak growth of living standards over time. In areas better endowed with local public goods, such as better access to infrastructure and other basic services (electricity, water and sanitation) there may be geographic externalities that facilitate the exit of poor households from poverty. According to this view, given two identical individuals, the one living in an area with lower endowment of these public goods may be condemned to stagnation and poverty over time. 2 Naturally, the policy focus that emerges from this perspective is programs targeted to the poor regions, aimed at increasing access to infrastructure and other basic services, frequently accompanied by institutions for local governance, and social capital. An example, among many of such programs is the recently launched program of the Mexican government aimed at reducing extreme poverty in the 100 poorest municipalities of the country. The emphasis on the municipality level is based on the assumption that the main determinant of extreme poverty is geographically differential access to public infrastructure. Basic intuition, ex ante, would suggest that the concentration effect will be a prominent explanation for the differences in welfare across regions because we expect that migration will affect the equalization of returns to a given set of (observable) characteristics across regions within a country. However, the role of migration in equalizing returns to characteristics may be limited by a number of factors such as monetary and psychic costs of migration (transportation, strong social ties and 2 For example, Escobal and Torero (2005) find in Peru that there are strong complementarities between human capital assets and public geographic assets (such as transport, telephones, sewerage). Poor, remote areas may perpetuate their situation by struggling to attract workers, producing for only a small local market and hence finding few economies of scale or protection against adverse shocks. On the other hand, once an area has started to grow, then the spatial externalities arising from the growing specialized labor force, available technology and cheaper transport costs serve to reinforce this geographic concentration of economic activity creating an increasingly dynamic area. 3

10 cultural values, such as the attachment of the indigenous peoples to land), and uncertainty about the benefits from migration (low probability of getting a high paying job, risk of unemployment, etc.). Another important factor is the presence of agglomeration effects brought to prominence by the new economic geography (e.g. Krugman, 1991). In these models, as labor migrates in response to a real wage differential, the size of the market grows in the destination region, and, through a variety of mechanisms, related to scale economies, the real wage in the destination region increases rather than decreases (Kanbur and Rapoport, 2004). In light of these considerations, the answer to the question of whether it is the concentration or the geography effect that is the dominant factor in explaining welfare disparities across space becomes an empirical issue. Yet, most of the policies and programs for poverty alleviation targeted either at poor people or at poor areas have been developed based more on ideological and political considerations and not on empirical evidence. A major part of part of the problem to date has been the absence of any empirical evidence on these issues that could directly inform and facilitate decision making. This is precisely the gap that this regional study attempts to fill. With this background in mind, we use household survey data from eight countries (Bolivia, Brazil, Colombia, Ecuador, Guatemala, Mexico, Nicaragua, and Peru) to address the broad question of whether the primary explanation for the differences in living standards is differences in the characteristics of households living in the poorer or richer areas or differences in the returns to these characteristics between poor and richer areas. Specifically, our study addresses a number of interrelated questions. These include: How does the standard of living vary within (e.g. urban vs. rural) and across (e.g. leading vs. lagging) regions in a country? Are the differences in returns to characteristics or the difference in the characteristics themselves the main determinant of the welfare differential within and across regions? How do returns and characteristics vary across the income distribution (quantile), over time? What is the role of migration in equalizing the returns to characteristics across regions? What are the main lessons learnt from some major regional development programs in LAC? Chapter 2 sets the stage by presenting the poverty profiles within and between regions in each of the eight countries in our study. The details of the methodology and the Oaxaca (1973) decomposition method that is used to asses the relative size of the concentration and geography effect in welfare differences across regions/areas are contained in Chapter 3. In essence, we classify the variety of factors associated with spatial differences in the standard of living into two broad groups: a set of covariates that summarize the portable or non-geographic attributes of the household, such as age, level of education, type of occupation etc., and a set of parameters that summarize the marginal effects or returns of these characteristics (either at the mean or at different points of the welfare distribution). Based on this framework, we then address the question of whether the spatial disparities in welfare are better explained by the sorting of people with low portable characteristics in some areas (e.g. less educated people being concentrated in the rural areas of any given region) or by persistent spatial differences in the returns to portable characteristics such as human capital. Our findings from the 4

11 various comparisons we conduct between urban and rural areas within regions and the similar areas between regions are contained in Chapter 4. In Chapter 5 we investigate further the extent to which our main inferences drawn from the simple welfare decompositions performed at the mean of the welfare difference between within regions change when we conduct similar decompositions at different points of the distribution of welfare. We also examine whether the spatial differentials in welfare have evolved over time. In consideration of the critical role of migration in the policy debate about spatial welfare disparities, Chapter 6 focuses on the role of internal migration within LAC countries. Given the relative scarcity of studies on internal migration within LAC countries, the analysis in this chapter is conducted using household data not just from the eight countries examined in our study but from all LAC countries. Chapter 7 summarizes the empirical evidence that is available regarding the poverty and welfare impacts of the two most distinct interventions that can be associated with the concentration and the geography view: Conditional Cash Transfers and Territorial Policies. Finally Chapter 8 summarizes our findings and discusses the policy implications based on our findings. 5

12 Chapter 2: Poverty profiles within and between regions In this section we aim to build a more thorough picture of regional differences in welfare throughout Latin America. We look at the standard of living both within regions and across regions, disaggregate by urban and rural areas and examine which variables appear to influence poverty. We first look at eight countries in total: Brazil, Mexico, Colombia, Ecuador, Peru, Bolivia, Guatemala and Nicaragua. A poverty profile is provided for each country followed by a brief discussion of how the variables under study contributed to regional differences. We then explore two views of the sources of poverty; poverty that is driven by weak endowments or poverty driven by low returns. The following table shows how the countries were divided by region and the source of data used for the study. Table 2.1 Regions of Countries and Sources of Data Country Regions Urban Rural Metro Data Source Brazil South Household Budget Survey (Pequisa de Southeast Orcamentos Familiares). Representative at the national and at Center West the regional level for metropolitan, urban and rural areas. North Mexico Colombia Ecuador Northeast North Center Mexico City South Pacific Gulf & Caribbean Atlántica Oriental y Bogotá Central Pacífica Antioquia Valle Sierra Costa Oriente Encuesta Nacional de Ingreso y Gasto de los Hogares Produced by Mexico s statistics institute (INEGI). Representative at national, urban and rural areas. Survey of Life Quality (La Encuesta de Calidad de Vida) The survey is representative at the national and regional level. Living Conditions Survey (Encuesta de Condiciones de Vida) Micro level data set collected by Instituto Nacional de Estadística y Censos. Representative at national level, urban and rural areas. Peru Living Standards Measurement Survey 2006 (ENAHO). Representative for regional, urban and rural areas. Peru Costa Selva Sierra Bolivia Pando The Bolivian Household Survey of Living Conditions 2002 Beni (MECOVI) is representative at national, urban, and rural levels Santa Cruz Tarija Oruro La Paz Cocha Chuquisaca Potosi Guatemala Metro Encuesta Nacional de Condiciones de Vida Produced by 6

13 Nicaragua Central Southeast Northeast Petén Southwest Northwest North Manuagua Pacifico Central Atlantico Instituto Nacional de Estadistica. Data representative at national, urban, rural, regional and departmental level. Living Standards Measurement Survey 2005 (Encuesta Nacional de Hogares Sobre Medición del Nivel de Vida). Produced by the Nicaraguan Program for Improving living Standards Measurement Surveys (MECOVI). Representative at national, regional, urban, and rural levels. A brief profile of poverty both within and across regions for every country follows. The headcount poverty rates reported here are derived using household consumption per capita as the measure of household welfare and the region-specific poverty lines estimated and reported in World Bank Poverty Assessments. 3 Brazil Poverty rates vary considerably both within and across regions, but it is clear that poverty is highest in rural areas and lowest in metropolitan areas (see Figure 2.1). In the aggregate, rural poverty rates (41%) are more than double urban poverty rates (18%). This is complicated somewhat, however, by issues of density. In terms of population, more poor people reside in urban areas. That is, 67 % of the poor reside in urban areas and 33% in rural areas. Looking across regions, the poverty pattern is similar for metropolitan, urban, and rural areas: the North and the Northeast consistently have the highest poverty rates in each of the areas. Headcount poverty rates in rural Northeast are estimated at 55%, while the South has the lowest (see Figure 2.2). 3 At this point we present poverty profiles based on headcount rates mainly due to their intuitive appeal. It is important to note that in some countries (e. g. Mexico, Ecuador), by construction the sampling of household surveys is designed for estimates that are representative separately for rural and urban areas for the whole country rather than within each of the regions of the country. Thus, the precision of the poverty estimates for rural and urban areas within regions may be quite low. For consistency with the analysis that follows we could have presented similar profiles using the log of the welfare ratio (discussed in more detail in chapter 3). 7

14 Figure 2.1: Poverty Within Regions Headcount poverty % Northeast North Center West Southeast South Metro Urban Rural Figure 2.2: Poverty Across Regions Headcount poverty % Metro Urban Rural Northeast North Center West Southeast South 8

15 Mexico Looking at poverty within regions in Mexico, it is clear that the Southeast of the country (the South Pacific and Gulf and Caribbean regions) has substantially lower standards of living than the North and Mexico City. 4 In 2005, Mexico City had a poverty rate of 31.8 percent while the three southern states of Guerrero, Oaxaca, and Chiapas were 71.6 percent. As in Brazil, the regional gaps are striking. It is notable also that Mexico City s income per capita is more than five times that of the Southern Pacific states yet Mexico City s poverty rate (31.8 percent) is slightly less than half of the poverty rate in the Southern Pacific states (71.6 percent). This suggests that income distribution is much more unequal in Mexico City. It is possible that Mexico s North benefits from its geographic proximity to the U.S. and Mexico City s economic activity is heavily concentrated on the service sector and secondarily on manufacturing, whereas the South and Southeast are in more remote locations where most agricultural production is small-scale and inefficient. Thus the market accessibility problem in the South and Southeastern regions of the country may be exacerbated by insufficient basic infrastructure, institutions and geography. Alternatively, the characteristics of the inhabitants could be driving Mexico s welfare disparities. Figure 2.3 Mexico Poverty Within Regions Headcount Poverty % North Center Mexico City South Pacific Gulf & Caribbean Urban Rural 4 Note that there is no poverty rate for rural Mexico City (DF). This is because there are no rural areas in this region. 9

16 Figure 2.4 Mexico Poverty Across Regions Headcount Poverty % Urban Rural Mexico City Center North Gulf & Caribbean South Pacific Again as in Brazil, rural areas are appreciably poorer than urban ones. The differences between urban and rural areas are significant in almost the entirety of the variables used and welfare ratios are also consistently substantially lower in rural areas. Ecuador Ecuador presents dramatic differences in poverty rates between urban and rural areas within each of its three regions Sierra, Costa, and Oriente. The smallest of the differences in poverty rates between urban and rural areas is in the Costa region, where the poverty rate in rural areas is around 51% while in urban areas it is slightly above 24%. In the Oriente region, the poverty rate in rural areas is more than four times the rate observed in urban areas (68% versus 16%). The Oriente region of the country is indeed the poorest. Although, the Oriente region urban poverty rate is lower compared to the urban poverty rate of the Costa region, most of the scarce population in Oriente is rural. Therefore, Oriente s overall poverty rate is much closer to the rates observed among its rural areas. As shown in Figures 2.5 and 2.6 below, it is the urban-rural differences across regions, more so than the within region differences, that are the most striking feature in Ecuador. Across regions, the largest differential in poverty rates occurs between rural areas of the Oriente and Costa regions (67.6% versus 51%), which is dwarfed by the staggering differential described above between urban and rural areas of the Oriente region. 10

17 Figure 2.5: Ecuador Poverty Within Regions headcount poverty % Sierra Costa Oriente Urban Rural Figure 2.6: Ecuador Poverty Across Regions headcount poverty % Urban Rural Sierra Costa Oriente Peru Figure 2.7 Peru Poverty Within Regions 11

18 80 70 Headcount poverty % Costa Selva Sierra Urban Rural Peru s regional disparities reflect the urban-rural divide we have seen in Brazil and Mexico and will continue to see in all of the countries under study. The Sierra and Selva regions continue to lag behind Costa and this is particularly exacerbated when we contrast rural Sierra and Selva s poverty rates of 77% and 62% with urban Costa s under 30% poverty rate. Further, extreme poverty affects almost half the population of rural Sierra versus less than the 5% of urban Costa s population. Figure 2.8 Peru Poverty Across Regions Headcount poverty % Urban Rural Costa Selva Sierra Colombia 12

19 Poverty rates across Colombia s six regions vary greatly, from a 37% in urban areas of the Valle region to a 72% among rural areas of the Pacific region. These figures roughly correspond to the country s wealthiest and poorest regions, however it is important to note that this classification does not hold when one looks at urban and rural areas separately. Indeed, although the Pacific region has the highest overall and rural poverty rates, the highest urban poverty rates are found in the Antioquia region, where they increase as high as 60%. As in the other countries included in this study, rural areas in Colombia are substantially poorer compared to urban areas within the same region (see Figure 2.9). However, Figure 2.10 shows that when looking across regions, some urban areas (in the Central and Antioquia regions) have higher poverty rates than some rural areas (in the Valle and Bogota and Oriente regions). Also, the figure indicates the ranking of regions in terms of poverty rates changes when one looks at urban and rural areas separately, and shows that the greatest poverty rates differentials across regions occur when considering urban areas only. Figure 2.9 Colombia Poverty Within Regions Headcount poverty % Atlántica Oriental y Bogotá Central Pacífica Antioquia Valle Urban Rural Figure 2.10 Colombia Poverty Across Regions 13

20 80 70 Headcount poverty % Valle Oriental y Bogotá Central Antioquia Atlántica Pacífica Guatemala Figure Guatemala Poverty Within Regions Headcount poverty % Metro Central Southeast Northeast Petén Southwest Northwest North Urban Rural Figure 2.12: Guatemala Poverty Across Regions 14

21 Headcount poverty % Urban Rural Metro Central Southeast Northeast Petén Southwest Northwest North Guatemala s poverty rates range from a staggering 86% headcount poverty rate rural North region to 14% in the urban Metro region. The above graphs show that while there is regional variation in poverty from the wealthier Metro and Central regions to the poorer Northeast and North of the country, it is the urban rural disparities within a region that are most striking. Nevertheless, with 23% of the population living in the Metro region and given a geographically small land area, Guatemala s urban poverty density is high. In looking at urban and rural regional issues it emerges that living in the rural North or Northeast regions of Guatemala lowers the mean per capita welfare ratio by about 30% relative to rural Metro. Bolivia In looking at the nine regions of Bolivia separated by urban and rural areas, it is in rural Potosi where households face structural regional disadvantages relative to all other areas. The mean welfare ratio is lowered by 28 percent for rural Potosi relative to other rural regions. Rural Tarija, Santa Cruz and Beni have some regional advantages over La Paz as does rural Pando where the greatest regional advantage is experienced (mean welfare is higher by 51% when other factors are controlled). Urban Santa Cruz has the greatest regional advantage. 15

22 Figure 2.13 Bolivia Poverty Within Regions Headcount poverty % Pando Beni Santa Cruz Tarija Oruro La Paz Cocha Chuquisaca Potosi Urban Rural 16

23 Figure 2.14 Bolivia Poverty Across Regions Headcount poverty % Urban Rural Pando Beni Santa Cruz Tarija Oruro La Paz Cocha Chuquisaca Potosi Notably, in Bolivia while there are clear differences between regions and between urban and wealthier areas that follow the same patters as the other Latin America countries under study, the disparities are not as striking. This should not obscure the fact however, that poverty in all areas is very high. Note that even the most advantageous area, Urban Pando, headcount poverty sits at 38% and the most extreme (of all the countries??) is rural Potosi with the staggering poverty rate of 92%. Nicaragua Figure 2.15 Nicaragua Poverty Within Regions 17

24 80 70 Headcount poverty % Manuagua Pacifico Central Atlantico Urban Rural Figure 2.16: Nicaragua Poverty Across Regions Headcount poverty % Urban Rural Manuagua Pacifico Central Atlantico Poverty rates across and within regions vary considerably. The highest poverty rates (70.8%) are experienced in the Nueva Segovia department of the Central region. The lowest headcount poverty rate is 19.1% in the Managua region though notably this region holds 24.4% of the total Nicaraguan population and so has high poverty density. Looking across regions, households in the Pacifico, Central and Atlántico regions face structural regional disadvantages relative to their counterparts in Managua. People living the Managua region, as measured by per capital annual consumption, have 50-90% higher levels of welfare than other regions. 18

25 Urban - rural disparities in Nicaragua are pronounced as Figure X above demonstrates. When we compare urban and rural areas within any given region, in the Managua region the urban mean welfare ratio is about 13% greater than the rural mean. The Pacifico region urban mean is 8% greater and this increases to a 21% differential in the Central and Atlántico regions. Intra-regional Urban-Rural Differential Nicaragua (2005) urban w.r. / rural w.r Managua Pacifico Central Atlantico The eight countries outlined above all experience significant regional inequality. Each country has a clear leading region and one or two lagging regions. People in who live in these lagging regions experience far greater levels of poverty as measured through poverty rates (graphed above), welfare levels and consumption rates. Rural areas in each country consistently show sharply increased poverty when compared to urban areas. While we are cognizant that due to population density more people live in urban regions, this does not mitigate the social inequality that exists between urban and rural settings. 19

26 Chapter 3 Methodology In this section, we outline the methodology used for our investigation of the factors behind the spatial disparities in the standard of living within select LAC countries. We being with a brief discussion of the measure used for the standard of living of households and their members and then summarize the Oaxaca-Blinder methodology, first used by Ravallion and Wodon (1999) to explore differences in mean welfare between urban and rural areas, within and between regions. The Oaxaca-Blinder decomposition allows us to estimate the relative contributions of differences in household characteristics and returns in accounting for differences in living standards. Comparing Living Standards Within and Between Regions The measure of standard of living that we use in our analysis is the welfare ratio 5. The welfare ratio is constructed by the ratio of nominal consumption expenditures per capita deflated by the region- specific poverty line. The region-specific poverty lines are assumed to incorporate all the cost of living differences faced by the poor in different regions of the countries examined. 6 There are both conceptual and pragmatic reasons why consumption expenditures available from household surveys are preferable for the purpose of poverty and inequality analysis to an indicator such as household income. It is argued, for example, that consumption expenditures reflect not only what a household is able to command based on its current income, but also whether that household can access credit markets or household savings at times when current incomes are low or even negative (due perhaps to seasonal variation or a harvest failure). In this way, consumption is thought to provide a better picture of a household s longer run standard of living than a measure of current income. Further, consumption expenditures for the poor are often better captured than household incomes. While poor households are probably purchasing and consuming only a relatively narrow range of goods and services, their total income may derive from multiple different activities with strong seasonal variation and with associated costs that are not always easily assigned. Given that in many LAC countries consumption expenditures are not usually collected, in the cases where consumption expenditures were not available we used household nominal income. Based on the arguments outlined in the previous chapter about the two extreme views regarding the sources of welfare disparities across regions we classify the variety of determinants of the welfare ratio into tow broad groups: a set of covariates that summarize the portable or non-geographic attributes of the household, such as age, level of education, demographic composition, and type of occupation, denoted by the vector X, and a set of structural parameters, denoted by the vector that summarize the marginal effects or returns of these household attributes. 5 The welfare ratio and its theoretical properties is discussed by Blackorby and Donaldson (1987). More practical applications of the welfare ratio in the measurement of poverty can be found in Ravallion (1998) and Deaton and Zaidi (2002). 6 Given that most poverty lines are constructed based on the cost of basic needs (CBN) approach which in essence is a Lapaayer s price index with fixed weights. As such the welfare ratio is also analogous to real expenditures. 20

27 Specifically, given any two regions or areas within a region, A and B, we assume that logarithm of the welfare ratio each region, denoted by lnc can be summarized by the linear regression ln C ln C A B A X A A, and (1) B X B B, (2) where is a random disturbance term with the usual properties, for summarizing the influence of all other factors on the standard of living. 7. In this specification, the returns to characteristics summarize the influence of a variety of factors on the standard of living in different regions. Basic infrastructure, and ease of access to markets and other basic services are some of the most important of these factors. In addition, returns to characteristics are also affected by the role of institutions, social customs and other cultural factors that are typically too difficult to quantify. Oaxaca-Blinder Type Decomposition Based on the specification above, and given that estimated regression lines always cross through the mean values of the sample, the mean difference in the standard of living between regions A and B, can then be expressed as lnc A lnc B X X A A B B (3) where the bar over the relevant variables denotes the sample mean values of the respective variables E 0 and we have used the assumption that j, for j A, B. After adding and subtracting the term difference above as B X A to the above differences we can express the ln C ln C A A ln C ln C B B A X X X X A B X A X B B A B X A B B A B A, or (4) (ln C) X B X A (4a) Alternatively, if one were to add and subtract the term expressed as A X B, the difference in (6) could be 7 Agglomeration effects are likely to have an influence on both the X s and the coefficients. 21

28 ln C A ln C B X A X B A A B X B, or (5) (ln C) X A X B (5a) Both expressions (4) and (5) imply that the differential in the mean welfare ratios between regions A and B, can be decomposed into two components: a component that consists of the differences in average characteristics summarized by the term X and another component that is due to the differences in the coefficients or returns to characteristics in different regions of a country summarized by the term. This is the decomposition method first proposed by Oaxaca (1973) and Blinder (1973). The decompositions given by expressions (4) and (5) are equally valid. The only difference between them lies in how the differences in the characteristics X and the differences in coefficients are weighted. In expression (7) the differences in the characteristics X are weighted by the returns of the characteristics in region B, whereas the differences in the returns are weighted by the average characteristics of households in region A. In contrast, in expression (8) the differences in the characteristics X are weighted by the returns of the characteristics in region A, whereas the differences in the returns are weighted by the average characteristics of households in region B. Since the original decomposition by Oaxaca, there have been numerous papers extending the method by proposing alternative weights for the differences in the characteristics X and the differences in returns, (e.g., Reimers (1983), Cotton, 1988; and Neumark, 1988). 8 We have followed Reimers (1983) and used as weights the average of the coefficients and the average of the characteristics, i.e. lnc A B X X X X lnc 5 A B A B A B A B, (7) At this point, it is important to point out that the use and interpretation of the decomposition method discussed above involves a number of caveats. For a start, these decompositions are simple descriptive tools that provide a useful way of summarizing the role of endowments and returns in explaining existing welfare differential. For this reason, we refrain from attributing causality to either endowments or returns in the welfare differences between or within regions. The variables that we use in place of X above, are composed only of portable nongeographic household characteristics. Specifically, the household characteristics used each 8 In fact, Oaxaca s decomposition can be shown to be as a special case of another decomposition (for a comprehensive summary see O Donnell, et al. 2008). In our study, the decompositions employed are done using the Stata command oaxaca written by Ben Jann (2008) using the weight (0.5) option. Thus we weighted the differences in the X s by the mean of the coefficient vectors. For additional details see the Brazil case study in volume 2 of this report. 22

29 country study consist of household size and structure, such as the number of household members less than 2 years of age, 3-11 years of age etc, binary variables identifying whether the head of the household is female, his/her race (black/indigenous), whether there is a spouse present, the age (and age squared) of the household head, the education of the head and its spouse, the differential in educations between the head and its spouse, and binary variables for major occupational categories. Our specification, intentionally excludes infrastructure and access to basic services. 9 The influence of the infrastructure as well as other omitted variables, is captured by default by the estimated coefficients of the portable characteristics of the household. As the formula for omitted variable bias suggests, the estimated coefficients of the household characteristics can be considered as including the direct effect of the omitted variables (such as infrastructure, local institutions and other household variables possible correlated with the location of the household) on welfare and their correlation with the included household characteristics. The decomposition formula in equation (7) holds only at the mean of the two regions being compared. It is likely that the findings obtained from the decompositions at the mean may not hold at other deciles of the distribution of welfare. Chapter 5 looks into this issue in more detail. The decompositions are performed for only one point in time. The extent to which the results of these decompositions change substantially over time is a question that is worth looking into. This question is also examined in Chapter 5. The decomposition results may be biased because of the presence of selection bias. To the extent there is free internal migration within and between different regions, then the current place of residence may not be exogenous. Most of the country studies did not control for the role of selection bias in the decomposition results. The only exception is the case of Brazil where, as in Ravallion and Wodon (1999), the decomposition results did not change significantly after correcting for selection bias (see the Brazil case study in Vol. 2). 9 The Peru country study were able to check the sensitivity of their findings by including and excluding three variables summarize access to basic infrastructure. The decomposition results they obtained were practically identical which suggests that many of the findings reported here are not sensitive to the specification. 23