A Critical Analysis of New Estimates by International

Declining Poverty in Latin America? A Critical Analysis of New Estimates by International Institutions Declining Poverty in Latin America? A Critical Analysis of New Estimates by International Institutions Abstract Indicators of progress in overcoming poverty in Latin America have been heralded recently by international institutions. Yet a closer look at data from the World Bank and the United Nations reveals contradictions that are not easily resolved by reference to the underlying methodologies. This paper provides an introduction to how poverty is measured, what the data indicate about trends in poverty, and reasons to tread cautiously in interpreting it as evidence of progress or stagnation. While significant progress has been achieved in a few large countries, the poorest countries are still very poor, and some countries have even seen increases in their poverty rates despite economic growth. Executive Summary Recent data from the World Bank and the United Nations suggest that poverty in Latin America is going down, and down significantly among the ‘extremely poor.’ With the renewed attention to poverty reduction created by the Millennium Development Goal to halve world poverty by 2015, the reported progress is welcome news. But how well do we understand these trends? Lost in the statements of progress are serious discrepancies in the available data, questions about the correct methodology for measuring poverty, and overgeneralizations about regional progress. In this paper, we look behind the optimistic assessments to examine the data and the methodologies used to interpret them. What we find is an alarming number of inconsistencies in the current poverty data. These should make scholars and policymakers pause before drawing conclusions about the quality of life experienced by Latin America’s poorest citizens. Widely Varied Measures Poverty data is collected and analyzed by national governments and international agencies. Most rely on the same set of household surveys, done periodically in most developing countries. The two most important international poverty measures come from the World Bank, with its World Development Indicators (WDI), and the United Nations, from its Economic Commission for Latin America and the Caribbean (ECLAC, or 1 CEPAL, to use its Spanish acronym). The two agencies analyze the same data quite differently. A third research organization, CEDLAS, offers its own interpretation of this data in collaboration with a different branch of the World Bank, its LAC Poverty Group. The World Bank’s data are perhaps the most widely cited for international comparisons. They are based on poverty lines of $1/day for extreme poverty and $2/day for moderate poverty. For Latin America as a whole, the World Bank suggests that moderate poverty has fallen from 26% in 1990 to 22% in 2004, while extreme poverty fell from 10% to 9% in the same period (World Bank 2007). CEPAL has developed country-specific poverty lines based on the estimated costs of a basic food basket. These are intended to capture the widely varying costs of living among the poor in different parts of the region. CEPAL’s poverty rates are generally higher than the World Bank’s but also tend to show greater improvement. According to CEPAL, moderate poverty in the region fell from 48% in 1990 to 40% in 2005, while extreme poverty dropped from 23% to 15% (Economic Commission for Latin America and the Caribbean 2007). Different methodologies naturally generate different numbers. But they also define the problem we refer to as ‘poverty,’ which implicitly calls for a policy response. Do we consider the extent of poverty 40%, as CEPAL states, or 22%, as the World Bank argues? Clearly, discrepancies of such magnitudes matter for policy formulation. And while both agencies document declines in both moderate and extreme poverty over time, the extent of decline is quite different, raising doubts about claims of progress achieved under particular policy regimes. Unless the distribution of income deteriorates, poverty ought to decline with economic growth. Reasoned debate about progress should focus on the pace of decline. Unfortunately, the erratic nature of the data precludes any such analysis, and even the direction of trend is in doubt in some cases. While we offer tentative interpretations of recent data, a central message of this paper is that meaningful work on poverty requires cautious use of these data. No Shared Trends Across the Region Are most individual Latin American countries experiencing similar reductions in poverty rates? Not surprisingly, there is great variation from one country to the next, and regional trends are dominated by the two largest countries, Mexico and Brazil. In fact, a closer examination suggests that generalizations about Latin American poverty reduction are driven largely by improvements in Mexico, at least as they are measured by international institutions. 2 A closer look at the country-level poverty trends suggests the following tentative conclusions: • Mexico – Data from the World Bank’s PovcalNet suggest that from 1990-2004 moderate poverty went down from 24% to 12%, and extreme poverty dropped from 6% to 2%. Did Mexico more than halve poverty levels since 1990? CEPAL data show both higher poverty levels and less dramatic reductions since 1989: from 48% to 37% for moderate poverty and from 19% to 12% for extreme poverty. These data are the subject of some controversy in Mexico, both because of doubts about the empirical evidence and normative reactions to the pace of progress in the context of an oil boom. Still, in a world driven by World Bank and CEPAL poverty estimates, the poverty cuts are large and drive the regional data, as such a large portion of the region’s poor reside in Mexico. • Brazil – The other large economy that drives regional averages has also shown progress in reducing poverty since 1990. Moderate poverty fell between 1990 and 2004 from 32% to 20% (WB), or from 48% to 38% (CEPAL). During that same time period, extreme poverty fell from 14% to 8% (WB), or from 23% to 12% (CEPAL). However, according to both sources, much of this progress occurred before 1996, with little taking place since then. • Everyone else – Take the two largest economies – which have both shown progress according these sources – out of the regional picture, and things look much worse for Latin America. Using the World Bank’s data, moderate poverty increased from 21% to 27% and extreme poverty rose from 7% to 11%. • Worst are in the middle – Four of the region’s more developed countries – Argentina, Colombia, Peru, and Venezuela – have shown the worst performance. As a group, they’ve seen moderate poverty balloon from 11% to 25%, while extreme poverty jumped from 2% to 10% (WB). In all four cases, the data are incomplete and sometimes inconsistent, but the extent of problems in these relatively large countries raises doubts about prospects for halving poverty by 2015. • Small countries with signs of progress – Chile, Costa Rica, Dominican Republic, El Salvador, Guatemala, and Panama have reduced poverty according to international data sources, and in some cases, the improvement is dramatic. As a group, these countries saw moderate poverty drop from 30% to 19% and extreme poverty from 13% to 7% (WB). Unfortunately, together they account for only 9.5% of the region’s population. • Poorest countries struggle – Among the region’s poorest countries – Bolivia, Ecuador, Haiti, Honduras, Nicaragua, and Paraguay – there has been varied success in reducing poverty. As a group, moderate and extreme poverty each fell by only 1%, from 49% to 48% and from 26% to 25%, respectively (WB). While there is evidence that some of these poor countries have seen notable 3 improvements, it tends to falter when cross-checked against other estimates. For example, the near halving of poverty in Honduras reported by the World Bank is inconsistent with the very slow progress reported by CEPAL. Bolivia, Ecuador, and Paraguay all seem to have experienced rising poverty, based on the limited data available. Conclusions Generalizations about declining poverty in Latin America are misleading. The data on which such claims are made do not agree with one another on the extent of poverty or about the pace of progress. Regional averages are driven largely by measured progress in its largest economies, but these averages mask wide variation among countries. Significant progress has been concentrated in a few countries, the poorest countries are still very poor, and some countries have even seen increases in their poverty rates despite economic growth. Policy analysts may reasonably differ in their opinions about an acceptable pace of progress in reducing poverty, but inconsistent data only distract us from finding solutions that improve people’s lives. If we are to replicate successes and rectify policy errors, we need more reliable measures of progress. The international community that established the Millennium Development Goals would create a stronger basis for assessing their achievement by creating common, transparent approaches to data collection. 4 “Over the last twenty-five years, following the prescriptions of international organizations, the governments of Latin American countries have implemented economic and social policies based in ‘free market’ principles, amounting to a ‘race to the bottom’ strategy. As a result, poverty has grown and labor and social policy standards have been gradually reduced to the minimum …” Damián and Boltvinik 2006 “The last four-year period (2003–2006) has thus seen Latin America’s best performance, in terms of social indicators, for 25 years. For the first time the poverty rate has come below the figure for 1980, when 40.5% of the population was classified as poor, while the indigence rate is now more than three percentage points below the 18.6% figure for that year. Moreover, the new figures show a reduction for the third consecutive year in the absolute numbers of poor and indigent, which is unprecedented in the region.” United Nations/CEPAL, Social Panorama of Latin America 2006 Latin America is a relatively rich region. Its per capita GDP of $4044 is three times that of East Asia, and seven times that of South Asia and Sub-Saharan Africa.1 Data showing that 120 million (out of 550 million) Latin Americans live on less than $2 per day, and 47 million live on less than $1 per day rightly appalls even many rich Latin Americans. But are these numbers accurate? There is little agreement about the extent of poverty, much less about how to solve it. Data on poverty has taken on greater significance since 2001, when the United Nations, the World Bank, and the IDB joined national governments in setting the Millennium Development Goals.2 Key to these MDGs is a commitment to reduce the 1990 poverty rate by half before 2015. With so much attention devoted to poverty reduction, one would think it easy to find information about trends in poverty. Instead, any foray into this field is bound to reveal contradictory data and controversial interpretations of the same data. The two quotes cited above reflect how easily interpretations are driven by an author’s choice of time frames. Yet even for the same year, estimates of poverty differ markedly. The United Nations, for example, estimated that 17% of Latin Americans lived in extreme poverty in 2004, while the World Bank estimated that 9% did.3 Policy advocates and scholars have understandably questioned whether the data provide a reliable reflection of reality. For users baffled by inconsistencies in this data, and for those unaware of its contradictions, this paper provides an introduction to how poverty is measured, what the data indicate about trends in poverty, and reasons to tread cautiously in interpreting it as evidence of progress or stagnation. The first section frames questions that can be 1 World Bank (2007) World Development Indicators Online (WDI Online), World Bank. 2 ECLAC (2005) The Millennium Development Goals: A Latin American Perspective. Santiago, Chile, ECLAC. 3 ECLAC (2007) Social Indicators and Statistics Database (BADEINSO); ECLAC; S. Chen and M. Ravallion (2007) PovcalNet, World Bank; World Bank (2007), WDI Online. 5 addressed by these data, and distinguishes between issues of evidence and interpretation. An important concern is that common definitions of poverty are too narrow and fail to direct policies toward the multiple dimensions of hardship experienced by the poor. The second section describes how poverty is defined and measured by three major international sources. Some contradictions in poverty data reflect methodological choices that can be clarified with a careful reading of source notes. Beyond these explanations, though, lie many unresolved discrepancies. Ironically, the institutions that generate poverty data are well aware of how methodological choices affect poverty estimates. They simply have not established standardized approaches to measuring poverty. The third section turns to the data itself. For the two largest countries, Mexico and Brazil, it seems possible to make sense of overall trends. International estimates indicate that Mexico has reduced poverty in the past two decades, perhaps thanks to oil revenue, while Brazil’s progress has been considerably slower. Several small countries can also claim unambiguous gains. In three large countries, Argentina, Peru and Venezuela, poverty has risen over the past decade and a half, but whether it has tripled or merely increased by 10% differs by data source. For Colombia, the region’s third largest country, the data sources do not even agree about the direction of trend. In several other cases, the evidence is also too erratic to support any conclusions about trends. Of course, even where we can find consistent data, interpretations cannot escape normative debate. Poverty ought to decline with growth. Unless inequality rises sharply, a higher GDP per capita should pull more people over a fixed poverty line. Thus it is troubling that users will find it difficult to figure out whether poverty has fallen, and not by how much, in several countries. To move forward in policy development, we need a better understanding of actual trends. I. What Do We Mean by Poverty? A short visit to poor communities in Latin America reveals a striking vulnerability to health crises, earthquakes, hurricanes, violent crime, unemployment and agricultural droughts. Ideally, we want to measure the extent to which this vulnerability is declining, whether through access to hurricane shelters or the employment of nurses in rural health clinics. Traditional measures of poverty focus more narrowly on household income or expenditure, with modest adjustments. Certainly higher incomes reduce such vulnerability, but much progress can be achieved without changing the financial status of the poor, and conversely, improvements in the incomes of the poor will not always resolve problems of vulnerability.4 4 More needs to be done to understand how our measures of poverty affect policy choices. Excessive focus on financial income can divert attention – and perhaps fiscal expenditures– from policies to improve lives in other respects. In fact, one concern is that the ‘success’ of cash transfers in reducing measured poverty undermines support for the provision of public services. 6 Poverty measurement can be categorized by the degree to which it includes aspects of well-being among the poor. Income poverty, typically measured against a poverty line, indirectly reflects access to goods and services that can affect well-being. Basic quality of life indicators are considered “direct” measures.5 Throughout Latin America, indicators such as infant mortality, malnutrition and illiteracy have improved markedly in nearly every country over the past two decades (see Figure 1). On this point, there is little controversy, although there is debate about whether progress has been fast enough.6 Figure 1 Quality of Life Indicators in Latin America 1990-2004 100 80 60 40 Infant mortality 20 Undernourishment 0 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 The broadest conceptualization of poverty includes such aspects as relative deprivation, political freedoms, gender equity and environmental sustainability, some of which are encompassed in Sen’s ideas of capabilities and substantive freedoms. This conceptualization is the most difficult to measure, due to its complex and often contextdependent nature.7 It also raises important questions about agency among the poor ─ that 5 Individual indicators have long been used by national and international institutions to monitor progress on specific aspects of wellbeing. More recently, composite indicators have been developed to measure poverty in a more comprehensive way. One well-known such indicator is the UNDP’s Human Poverty Index (available at http://hdr.undp.org/hdr2006/statistics/indicators/18.html). Others include the NBI (Necesidades Básicas Insatisfechas), a version of which covers urban areas in Latin America and is available from CEDLAS, as well as the MMIP (Medición Integrada de la Pobreza), covering Mexico only. See CEDLAS and the World Bank’s LAC Poverty Group (LCSPP) and MECOVI Program (2006) SocioEconomic Database for Latin America and the Caribbean (SEDLAC); Boltvinik, J. (2003) "Tipología de los métodos de medición de la pobreza. Los métodos combinados." Comercio Exterior 53(5): 453-465. 6 See A. Fiszbein, ed. (2005), Citizens, Politicians and Providers: The Latin American Experience with Service Delivery Reform, World Bank. 7 Indicators for this broadest conceptualization of poverty and development tend not to be aggregated into composite indicators of “poverty.” A few (among many) examples of relevant indicators include the Gender Empowerment Measure (http://hdr.undp.org/hdr2006/statistics/indicators/) and the various indicators in the UNDP’s Democracy in Latin America: Towards a Citizens’ Democracy, particularly the 7 is, apart from the hardship of surviving on too little food, in cramped quarters, with grueling workhours, do the poor get a voice in the meta-system that condemns them to poverty? The focus of this paper is on quantitative measures of absolute poverty. Absolute poverty refers to the ability to consume a minimal bundle of goods. The emphasis is on measures of consumption or income, and the ability to afford a fixed standard of living. While the choice of such a bundle is obviously driven by cultural standards, it is a different concept than relative poverty, which measures income distribution.8 We focus on income- or consumption-based measures because, despite their limitations, such measures are pervasively used in the assessment of success in anti-poverty programs and of progress in general. The choice of a poverty line is frequently controversial. Multilateral institutions have settled into a common range of roughly $2 per day for moderate poverty and $1 per day for extreme poverty. Damián and Boltvinik ask whether the UN’s definition of extreme poverty, the cost of a basic food basket, adequately reflects the resources needed to sustain even the crudest human existence: To purchase the Normative Food Basket (NFB), households with income equal to the extreme poverty line (EPL) would have to spend 100 percent of their earnings on raw food, but they would find it impossible to consume it since the NFB does not include the requirements for its preparation and consumption (gas, pans, utensils, plates, a table, detergents, etc.). This definition of extreme poverty reduces human beings to an animal state, eating raw food on the floor with 9 their hands. To complicate matters, even the poorest Latin Americans do not necessarily devote their resources to food. In “Economic Lives of the Poor,” Banerjee and Duflo suggest that the poor spend considerable money on – and presumably value highly – religious festivals, televisions, alcohol and tobacco, even when they cannot afford enough socio-economic breakdowns in Second Part of the Statistical Compendium. Various empowerment indicators for specific ethnic groups can be found in the Minorities at Risk database (http://www.cidcm.umd.edu/mar/). 8 The relative poverty line is usually defined as the percent of the population that receives less than half the median income in a country. SEDLAC data indicates that 26% of Argentines fell below this relative poverty line in 2004, while only 22% of Guatemalans were ‘relatively poor.’ Yet a far greater proportion of Guatemalans lacked adequate housing, food and health care. Because the median income in Argentina is higher than in Guatemala, the relative poverty line is, as well. Even within countries, the concept of relative poverty can be misleading for trend analyses, since median income changes over time. Chen and Ravallion (2001) create a hybrid concept of ‘relative poverty’ which excludes individuals with income above $1 a day and one third the mean national income. 9 A. Damián and J. Boltvinik, “A Table to Eat on,” in E. Hershberg and F. Rosen, Latin America after Neoliberalism, New Press, 2006. 8 to eat. Patterns of such expenditures vary widely among people living on less than $1 a day in Latin America. 10 New, higher poverty lines are beginning to displace the $1 and $2 per day measures used here. For example, the media reports that half of Mexicans reportedly live on less than $4 per day.11 While this higher standard makes sense from a normative perspective, shifting standards makes it more difficult to assess the pace of progress against poverty. Since a substantial portion of Latin Americans still cannot afford a basic food basket, much less twice that, the existing poverty lines are not yet irrelevant. Given this construction of ‘poverty’ as the inability to afford a minimal bundle of goods, what can we expect to learn from the data? Users might reasonably hope to find: 1) information on whether poverty has been declining, and if so, how fast; 2) whether some countries experience higher rates of poverty than other countries at comparable levels of GDP per capita; and 3) whether trends in some countries warrant a closer examination of policies that can be replicated (or avoided) elsewhere. What the basic data will not reveal, even at its best, is whether the same families are trapped in poverty (chronic vs. temporary poverty); how far the income of poor families falls short of the poverty line (the poverty gap); and whether some groups (e.g., indigenous people) suffer disproportionately from poverty. Nor will the data provide guidance in setting normative benchmarks for evaluating the pace of progress. II. Measures of Absolute Poverty In addition to work done by individual countries, there are three major sources of data on poverty in Latin America. The World Bank provides data in its World Development Indicators (WDI), as well as in an online software tool called PovcalNet, which uses the same underlying data as the WDI.12 The United Nations’ CEPAL tracks poverty in its BADEINSO database.13 An Argentine research institute, CEDLAS, compiles its own estimates in its SEDLAC database.14 Conceptually, all follow a 10 Banerjee, A., and E. Duflo, "The Economic Lives of the Poor," Journal of Economic Perspectives, 21(1): 141-167, Winter 2007. To our knowledge, the metrics of poverty have not been tested to see how well they measure what the poor seek most urgently. 11 “Cost of Corn Soars, Forcing Mexico to Set Price Limits,” New York Times, 1/19/07, p. A12 12 PovcalNet’s calculations are based on the same underlying data and calculation methodologies as the World Development Indicators. However, estimates sometimes vary slightly as WDI uses individual household records, while PovcalNet uses grouped distributions. 13 Acronyms: BADEINSO ─ Base de Estadísticas e Indicadores Sociales; CEDLAS- Centro de Estudios Distributivos Laborales y Sociales; CEPAL ─ Comisión Económica para América Latina y el Caribe (also, Economic Commission for Latin America and the Caribbean, ECLAC); SEDLAC─ Socio-Economic Database for Latin America and the Caribbean; WDI─ World Development Indicators. 14 CEDLAS works in collaboration with the World Bank’s LAC Poverty Group. 9 common approach in defining poverty and use similar lines of absolute poverty. In addition, they all rely on the household surveys carried out by national statistical agencies. In this section, we will examine similarities and differences among these three sets of estimates, as well as some of the controversies associated with them, with a focus on the poverty line chosen, the methodology used in calculating the poverty estimates, and the household surveys on which they are based. II.A. Poverty Line Selection The World Bank and CEPAL each set two poverty lines, the higher of which is referred to as moderate poverty, or simply poverty, and a lower line that defines extreme poverty, or indigence. The World Bank sets poverty lines that are common across all countries: the poverty line is set at $2.15 per day and the indigence line at $1.08 per day.15 In contrast, CEPAL has developed country-specific poverty lines. Its extreme poverty line reflects the income necessary to buy a minimally nutritious basket of food, as defined by World Health Organization standards and local dietary customs. CEPAL’s moderate poverty line is set at twice this indigence line for urban areas and 1.75 times the indigence line for rural areas. CEDLAS calculates poverty rates for several different poverty lines, including $2.15 and $1.08 per day.16 The World Bank’s approach has been credited with consistency across countries, particularly since the dollar-a-day standard is adjusted for exchange rate biases using purchasing power parity (PPP) indexes. CEPAL takes issue with this, arguing that its own approach better reflects the bundle of goods actually consumed by the poor. Because patterns of consumption vary across countries, CEPAL argues, so does the cost of living.17 How different are the measures? In fact, the two lines are not very different except in a few countries, notably Bolivia, Brazil and Nicaragua on the low end, and Mexico and Venezuela on the high end. Table 1 shows the CEPAL poverty line for several Latin American countries. The significance of these differences depends on how many people live just above or below the poverty line. Wade argues that the World Bank’s poverty rates are very sensitive even to small differences in the poverty line.18 Even where CEPAL and the 15 These numbers originally started out at an even $2 and $1 per day, but purchasing power adjustments have since been made. 16 Gasparini, L., F. Gutiérrez, et al. (2005) Growth and Income Poverty in Latin America and the Caribbean: Evidence from Household Surveys. La Plata, CEDLAS, Universidad Nacional de La Plata. Chen and Ravallion (2007) PovcalNet; ECLAC (2006) BADEINSO; World Bank (2007) WDI Online. 17 ECLAC (2006) Social Panorama of Latin America: 2006. Santiago, United Nations Publications. 18 Wade, R. H. (2004). "Is Globalization Reducing Poverty and Inequality?" World Development 32(4): 567-589. 10 Table 1 CEPAL Extreme Poverty Lines in US$ 2000-2004, monthly, using current (IMF) exchange rates Argentina Bolivia Brazil Chile Colombia Costa Rica Dominican Republic Ecuador El Salvador $37.40 Guatemala $22.60 Honduras $22.10 Mexico $34.90 Nicaragua $34.60 Panama $39.20 Paraguay $42.20 Peru $34.60 Uruguay $34.90 Venezuela Source: CEPAL 2005, p. 319 $43.60 $41.60 $70.60 $27.60 $40.70 $41.80 $34.50 $37.30 $69.10 World Bank use closely comparable poverty lines, and apparently the same household survey data, it is surprising how significantly their estimates differ. For example, according to CEPAL, 51% of Colombians lived below the moderate poverty line (roughly $2 per day) in 2004, while World Bank and CEDLAS estimates were 18% and 26%, respectively. II.B. Methodology for Handling Survey Data Some differences in estimates reflect how the institutions handle survey data, as well as their choice of data sources (see Table 2). CEPAL and CEDLAS make adjustments for the different costs of living in rural and urban areas: CEPAL adjusts the rural poverty line downward by 15-25%, while CEDLAS scales up rural income a similar percent. The World Bank uses the same poverty line for rural and urban areas. Similarly, on the assumption that children require fewer resources than adults, CEDLAS makes ‘adult equivalence’ adjustments for household demographics; neither the World Bank nor CEPAL do so. A controversial methodological issue is the treatment of non-monetary income, particularly the imputation of rent to owner-occupied housing. During housing booms, imputed rent from owner-occupied housing will rise, even though the occupants of a simple shack may lack enough money to buy food. These owners are arguably better off than renters or squatters with the same cash income, but rising imputed rent may move some households above the poverty line despite their penury. Each source also differs in terms of how it handles missing data and resolves inconsistencies with national accounts. CEPAL assigns income to households that fail to report income (‘omitted income’); it also scales the whole database to reconcile changes with macroeconomic data. Neither CEDLAS or the World Bank make such changes. The importance of these seemingly subtle methodological differences has been systematically examined by Szekely, Lustig, Cumpa and Mejia.19 Their estimates show 19 M. Székely, N. Lustig, M. Cumpa, J.A. Mejía (2000) “Do We Know How Much Poverty There Is?” Inter-American Development Bank Working Paper 437, Washington, D.C., December 2000. 11 Table 2: Summary of Methodological Differences among Databases Poverty line? Government transfers included? SEDLAC CEPAL PovcalNet $1 & $2 per day Based on cost of food basket Some monetary transfers are included, especially for large programs. Yes No Usually Yes Yes $1 & $2 per day Yes Some transfers are included, depending on the country and the survey. No No Usually No No Yes Rural/urban adjustments? Yes Adult equivalence scale? Yes Imputed rent from own-housing? No Adjustments for omitted income? No Adjustments based on national accounts? Sources: Chen and Ravallion 2000, 2004, and 2007; Gasparini, Gutiérrez et al. 2005; ECLAC 2006; Chen 2007; Mancero 2007 that typical adjustments for nonmonetary income, adult equivalence and treatment of missing or under-reported income significantly change the level of poverty, often by a factor of three or four. Methodological differences also affect the ranking of Latin American countries by poverty rates. Even if one cannot compare CEPAL, World Bank and CEDLAS data directly, trends in the three series should be fairly similar. For most countries, the direction of trends is generally consistent across the three sources, but the slope often differs. According to Gasparini, Gutiérrez et al. (2005), the CEDLAS estimates are most highly correlated with the CEPAL estimates (a linear correlation of 0.931), and less well correlated with the estimates from the World Development Indicators database (correlations of 0.878 and 0.745 for moderate and extreme poverty rates, respectively). Since our primary concern is about how World Bank and CEPAL data is interpreted by the public, we refer to CEDLAS data only where it indicates a different perspective. The Appendix provides data from all three sources. II.C. Underlying Household Survey Data How good are the data themselves? Do they measure what they purport to measure? Survey frequency has increased dramatically over the past twenty years, leading to less need for interpolation between surveys. In addition, survey quality and standardization has increased, in part due to the Program for the Improvement of Surveys and the Measurement of Living Conditions in Latin America and the Caribbean (MECOVI).20 MECOVI has been administered by the World Bank, CEPAL, and Interamerican Development Bank, and applied in Argentina, Bolivia, Colombia, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, Paraguay, and the Dominican Republic.21 Unfortunately, this standardization is far from complete. 20 MECOVI is the Spanish acronym for “Mejoramiento de las Encuestas de Hogares y la Medición de Condiciones de Vida.” 21 CEDLAS and LAC Poverty Group (2006) A Guide to the SEDLAC. World Bank (2007) “Poverty and 12 Despite the improvements (and, in some cases, because of the improvements), there remain time series comparability problems that are sometimes quite substantial. Similarly, underlying survey data differences affect comparisons across countries and result in problems of regional aggregation.22 Even where survey coverage and methodology have remained similar enough over a long time to reasonably support conclusions about progress on poverty, changes in relative prices and exchange rates must be addressed. Recent research suggests that this ought to be more complicated than simply scaling up a database by the new CPI, because relative prices of some goods disproportionately affect the poor.23 Finally, forecasts are often made and distributed between survey years based on macroeconomic growth rates and the assumption of constant distribution. It is not unusual for these forecasts to be significantly revised in later years.24 III. What Do the Data Tell Us? III.A. Cross-country Comparisons One expects poverty rates to be higher in poor countries, as they generally are. Table 3 provides data on average income and poverty rates in individual countries. We tend to look to economically similar countries to set standards for their peers. Unfortunately, absolute cross-country comparisons are extremely problematic, whether one uses World Bank or CEPAL data. Differences in methodology and household survey design yield results that are not comparable. For example, the World Bank data indicate that poverty is twice as prevalent in Nicaragua as in Honduras, a result that is implausible, given their similar income levels and Honduras’ higher inequality. According to both the World Bank and CEPAL, Chile and Costa Rica stand out as countries with lower poverty rates than one would expect based on national income. On the other hand, Argentina and Venezuela have higher than expected poverty rates. But Social Indicator Monitoring (MECOVI).” 22 CEDLAS and LAC Poverty Group (2006) A Guide to the SEDLAC. Wade, R. H. (2004) “Is Globalization Reducing Poverty and Inequality?” 23 All three sources use homogeneous price deflators that fail to account for differences in consumption patterns of the rich and poor. S. Suarez Dillon Soares and R. Guerreiro Osório construct price indices for each hundredth of the Brazilian population in “The Impact of Relative Prices on Welfare and Inequality in Brazil, 1995-2005,” International Poverty Centre, Working Paper 37, May 2007. Their results show that distributional measures like the Gini coefficient are sensitive to class-specific price indices because prices of food and other basic goods do not move in close relationship to prices of luxuries like telecommunications and transportation. They suggest that the poor may be acutely sensitive to exchange rate movements during financial crises. 24 See for example, ECLAC (2006) Social Panorama of Latin America: 2006. 13 Table 3: Moderate and Extreme Poverty Ranked by Average GDP per Capita, 2004 Percent of population GDP Moderate Poverty Extreme Poverty per World CEPAL World CEPAL capita Bank Bank $437 75.8 n.a. 50.7 n.a. Haiti $859 79.8 69.3 44.9 42.3 Nicaragua $968 36.0 74.8 14.1 53.9 Honduras $1,039 42.8 63.9 23.9 34.7 Bolivia $1,348 29.8 65.9 13.6 36.9 Paraguay $1,501 35.2 51.2 14.7 22.3 Ecuador $1,721 31.0 60.2 13.1 30.9 Guatemala $2,099 19.4 51.1 7.6 24.2 Colombia $2,106 40.4 47.5 20.4 19.0 El Salvador $2,227 30.6 51.1 10.5 18.6 Peru $2,442 16.2 54.4 2.8 29.0 Dominican Republic $3,564 19.8 37.7 7.6 12.1 Brazil LATIN AMERICA $4,044 22.3 42.0 8.7 16.9 $4,217 16.8 31.8 6.0 14.8 Panama $4,321 9.6 20.5 1.8 8.0 Costa Rica $4,596 40.2 45.4 18.7 19.0 Venezuela $5,436 5.6 18.7 0.5 4.7 Chile $5,902 9.2 20.9 0.0 4.7 Uruguay (urban) $6,056 12.5 37.0 1.9 11.7 Mexico $7,486 17.4 29.4 6.6 11.1 Argentina (urban) Notes: CEPAL estimates - Bolivia, Chile, & Honduras: 2003; Guatemala: 2002; Nicaragua: 2001; Argentina & Uruguay: urban only. Sources: ECLAC 2007; Chen and Ravallion 2007; World Bank 2007. rankings of several countries are surprisingly different between the data sources: Mexico has far too much poverty for its income level according to CEPAL, but fits right into line if one ranks countries by income and poverty rates according to the World Bank. Even where rankings are similar, the relative estimates of poverty can vary widely between sources. The World Bank average poverty level for the region is exactly half that reported by CEPAL, but this ratio varies considerably across countries. It is unusually high in Bolivia, Ecuador, El Salvador, Nicaragua and Venezuela.25 The World Bank estimates of poverty in Colombia, Chile, the Dominican Republic and Mexico are markedly lower than half the CEPAL rate.26 25 These differences are only partly explained by CEPAL’s country-specific poverty lines, as CEPAL sets very high poverty lines in both Mexico and Venezuela, but estimates comparatively higher poverty in Mexico and lower poverty in Venezuela than the World Bank. 26 The deviance of World Bank poverty rankings from GDP rankings ironically suggests that distribution matters greatly, although the Bank has tended to shy away from advocating strong redistributional policies, notwithstanding its recent publication of World Development Report 2006: Equity and Development. 14