POVERTY THE WORLD BANK REDUCTION AND ECONOMIC MANAGEMENT NETWORK (PREM) Economic Premise JUN 2012 NOVEMBER 010 • Numbe 97 • Number 18 73663 When Job Earnings Are behind Poverty Reduction Gabriela Inchauste, João Pedro Azevedo, Sergio Olivieri, Jaime Saavedra, and Hernan Winkler Improvement in labor market conditions has been the main explanation behind many of the poverty success stories observed in the last decade—that is the primary conclusion of an analysis of changes in poverty by income source. Chang- es in labor earnings were the largest contributor to poverty reduction for a sample of 16 countries where poverty increased substantially. In 10 of these countries, labor income explained more than half of the change in poverty, and in another 4 countries, it accounted for more than 40 percent of the reduction in poverty. A declining dependency rate accounts for over a fifth of the reduction in poverty in 10 out of 16 countries, while transfers and other nonearned incomes account for more than a quarter of the reduction in poverty in 9 of these countries. A further decomposition of the contribution of labor income to poverty reduction in Bangladesh, Peru, and Thailand found that changes in individual characteristics (education, work experience, and region of residence) were important, but that overall, increases in real earnings among the poor matter the most. For the first time since the World Bank began monitoring that led to lower dependency ratios? Was poverty reduction progress against poverty in the late 1980s, 2008 data indicat- the result of higher employment or higher labor force partici- ed a decline in both the poverty rate and the number of poor pation? Was it higher labor productivity that led to higher real in all regions of the developing world. Moreover, within each earnings, or did higher transfers from public or private sourc- region, poverty has fallen over the past decade in the vast ma- es make the difference? Was it the result of changes in the sec- jority of countries, whether one uses national or international toral composition of employment? Were these changes the poverty lines.1 This progress represents a substantial step in result of improved human capital (better trained or more ex- reaching the Millennium Development Goals. According to perienced workers), or were they the result of changes in the the latest projections, the goal of halving absolute poverty by returns to those characteristics? 2015 was reached in 2010. Despite this progress, many peo- To answer these questions, this note focuses on a subsam- ple remain desperately poor across the world: 1.2 billion peo- ple of countries where there was a substantial decline in pov- ple live below the very frugal line of US$1.25 per day, and 2.4 erty as a way to contribute to the evidence base for policy go- billion live below the low line of US$2.50 per day. ing forward. For example, in some countries in Latin America, In this context, the last decade affords us an opportunity there is debate around the relative roles of better job opportu- to better understand the most significant factors at work and nities and expansion and effectiveness of transfer policies in contributing to the progress in the poverty front. Was the ob- explaining the observed reductions in poverty and inequality. served reduction in poverty a result of demographic changes For instance, Brazil was successful in reducing poverty and 1 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise inequality despite modest growth. Was this success due to im- An obvious question is how growth led to poverty reduc- proved labor market conditions or to more effective social tion and whether changes in redistribution in Latin America policies? In some South Asian and Eastern European coun- were associated with the introduction of public transfers of or tries, some question whether the reduction in poverty was on the result of market forces. To answer this question, analysis account of better job opportunities at home, or due to higher was conducted using a method that generates entire counter- remittances. In East Asia, several countries have seen strong factual distributions, allowing a decomposition of the ob- growth, job creation and poverty reduction, but are lately served poverty changes on account of changes in different questioning whether social policy should have a stronger fo- sources of income and in individual and household character- cus on redistribution. istics.2 In particular, two approaches were used; the first one, adapted from Barros et al. (2006), is a simple accounting ap- Growth and Poverty Reduction proach that quantifies the contributions to household welfare The links between economic growth and poverty reduction on account of changes in demographics, changes in employ- have long been of interest to economists. As detailed in Fer- ment and earnings and changes in nonlabor income, includ- reira (2010), cross-country literature has found considerable ing changes in remittances, public transfers, and other private evidence that economic growth is strongly and negatively cor- transfers. The second approach adapts the Bourguignon, Fer- related with changes in poverty (Ravallion and Chen 2007). reira, and Lustig (2005) methodology and further distin- One common way to look at these correlations is the Datt and guishes distributional changes on account of changes in en- Ravallion (1992) decomposition, which splits the change in dowments or/and returns to those endowments, changes in poverty into growth and redistribution components. Using occupation, location, age and gender structure of the popula- this method, Azevedo et al. (forthcoming) focus on 16 coun- tion, along with the nonlabor dimensions mentioned above.3 tries with substantial declines in poverty, defined as an aver- Although these decompositions do not allow for the age decline in moderate poverty of 1 percentage point per year identification of causal effects, they are useful to focus atten- or more over the last decade. Moderate poverty lines refer to tion on the elements that are quantitatively more important the international poverty line that is closest to the national in describing changes in poverty.4 moderate poverty rate. The conclusion is that growth ac- The Forces behind Poverty Reduction counts for most of the reduction in moderate poverty in 14 of the 16 countries over the past decade (figure 1). Redistribu- Behind the changes in growth and redistribution, there are tion was found to be more important only in the case of Ar- other structural factors that might explain changes in poverty. gentina and Paraguay. First, demographics. Declining dependency rates5 can lead to increases in income and consumption per person. Among the countries considered here, this Figure 1. Growth and Redistribution Decomposition of Moderate Poverty Changes decline has already started to bear fruit, while at annual % point the same time the youth bulge observed in earlier growth redistribution reduction in periods has now reached a working age. For exam- poverty Moldova, 2001–10 3.9 ple, dependency ratios have declined by about 20 Romania, 2001–9 5.3 percent in Costa Rica, Honduras, Moldova, and Bangladesh over the last decade. Costa Rica, 2000–2008 -1.3 Honduras, 1999–2009 -1.4 Second, growth in labor income. Labor is the Thailand, 2000–2009 -1.6 main asset of the poor, as such, and through varied Peru, 2004–10 -2.6 Colombia, 2002–10 -2.8 mechanisms, labor incomes are potentially the key Ecuador, 2003–10 -2.6 factor for moving out of poverty. For most of the Panama, 2001–9 -1.7 countries in the sample used here, both the share of Chile, 2000–2009 -1.3 Brazil, 2001–9 -1.9 occupied adults per household and household la- Paraguay, 1999–2010 -1.0 bor income increased. One exception is Romania, Argentina, 2000–2010 1.3 which was among the hardest hit countries of the Nepal, 1996–2003 -2.9 2008 economic crisis. However, for the other coun- Ghana, 1998–2005 -2.1 tries, growth in labor income was typically accom- Bangladesh, 2000–2010 -1.7 panied by employment growth, which in some cas- -40 -20 0 20 40 60 80 100 120 140 percent es was related to increases in female employment Source: Azevedo et al. forthcoming. across countries in the sample.6 Note: Consumption-based measures of poverty are used in the case of Bangladesh, Ghana, Nepal, Peru, Thailand, Moldova, and Romania. Income-based measures of poverty are used in the case of Third, growth in nonlabor income. For exam- Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, Honduras, Panama, and Paraguay. ple, government spending for subsidies and trans- 2 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise fers as a share of gross domestic product (GDP) increased Although changes in labor income are the main contrib- more than sevenfold in the case of Ghana, and they increased utors to changes in poverty in most countries, demographics more than sixfold in the case of Bangladesh. In addition to also matter. In particular, a higher share of working age adults public sources of transfers, private transfers, in the form of in the household made the largest contribution to poverty re- remittances, have also grown strongly. For example, in Nepal, duction in Paraguay and Costa Rica. Changes in the share of remittances grew by 36 percent per year on average over the adults per household were also relatively important in ex- decade, while in Honduras they grew by 11 percent per year plaining declines in moderate poverty in Bangladesh, Chile, on average. The question is how important these changes in Ecuador, and Honduras (figure 2). public and private transfers have been for poverty reduction While public and private transfers were important, they relative to labor market–related sources.7 played a relatively smaller role in explaining declines in mod- erate poverty for most countries in the sample. The excep- Decomposing Poverty Reduction tions were Romania and Moldova, where transfers contribut- Demographics, labor incomes, public transfers, and remit- ed relatively more to changes in poverty. In Romania, this was tances—which factor contributed the most to observed short- related to changes in transfers and capital income, while in run reductions in poverty? The key result that emerges is that Moldova it was mostly related to the increase in international the most important contributor to the reduction in poverty remittances. has been the growth in labor income. In particular, changes in When looking at changes in extreme poverty, measured labor income per adult had the largest contributions to chang- by a US$2.5-dollar-a-day poverty line, nonlabor incomes are es in moderate poverty (figure 2). In 10 out of 16 countries relatively more important in accounting for changes in pov- with substantial poverty declines, labor income explained erty.8 In particular, for some of the middle-income coun- more than half of the change in moderate poverty, and in an- tries in the sample, transfers were especially important in other 4 countries, it accounts for more than 40 percent of the reducing extreme poverty in the cases of Argentina, Brazil, reduction in poverty. Chile, Costa Rica, Ecuador, Romania, and Thailand. This Figure 2. Decomposition of Changes in Poverty (percent contribution to the change in moderate poverty) 140 120 100 80 60 40 20 0 -20 -40 Ghana, 1998–2005 Nepal, 1996–2003 Bangladesh, 2000–2010 Costa Rica, 2000–2008 Paraguay, 1999–2010 Chile, 2000–2009 Ecuador, 2003–10 Honduras, 1999–2009 Argentina, 2000–2010 Brazil, 2001–9 Colombia, 2002–10 Panama, 2001–9 Thailand, 2000–2009 Peru, 2004–10 Romania, 2001–9 Moldova, 2001–10 US$1.25 per day US$4.00 per day US$5.00 per day share of working-age family members employment + earnings nonlabor income consumption-to-income ratio Source: Azevedo et al. forthcoming; Inchauste et al. (2012). Note: “Labor income� refers to the change in employment and earnings per adult; “nonlabor income� refers to transfers, pensions, capital, and other nonlabor income. Consumption-based measures of poverty are used in the case of Bangladesh, Ghana, Nepal, Peru, Thailand, Moldova, and Romania. Income-based measures of poverty are used in the case of Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, Honduras, Panama, and Paraguay. 3 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise finding is consistent with the social protection systems in in the returns to characteristics, as well as nonlabor income those countries, which are typically targeted to the bottom and other factors. The main result that emerges, consistent of the distribution. with the results in the previous page, is that the largest contri- butions to poverty reduction in all three countries were labor Why Has Labor Income Increased? market–related factors (table 1). Indeed, these contributions The next logical question is what accounts for the increase in to moderate poverty reduction amount to 61 percent in Ban- labor income for workers in the different parts of the distribu- gladesh, 75 percent in Peru, and 65 percent in Thailand. tion? Was it the result of more people working, or higher earn- Within this, it was the increase in the returns to endowments, ings per worker? Was it the result of changes in the occupa- rather than changes in these endowments, that explain pov- tional structure, or changes in the sectoral composition of erty reduction: returns to land and experience in the case of employment? Were these changes the result of improved hu- Bangladesh, returns to land in Peru, and returns to education man capital characteristics (education or experience), or and experience in Thailand, pointing to an increase in real higher returns to those characteristics? To answer these ques- earnings and higher productivity as the main contributors to tions in some detail, Inchauste et al. (2012) focus on Bangla- poverty reduction in each case. While increases in farm in- desh, Peru, and Thailand. These three economies experienced come were mostly responsible for poverty reduction in Ban- fast poverty reduction during the last decade, while moderate gladesh and Thailand, nonfarm income was mostly responsi- national poverty headcount rates in each fell by over 12 per- ble in Peru. Finally, a separate analysis focusing on changes in centage points. In each case, growth was very high during the extreme poverty shows that while nonlabor income played an decade, well above 4 percent per year during 2002–8. In Peru important role in reducing extreme poverty in Thailand and and Thailand, there was a sharp deceleration due to the finan- Peru, labor income, either from farm or nonfarm sources, was cial crisis in 2009 that rebounded very quickly the following the main contributor to poverty reduction.9 Within this, it year. In contrast, Bangladesh went through the great crisis un- was increases in the returns to endowments, which can be scathed. In all countries, employment and public social trans- thought of as increases in the marginal value of work, that fers increased, as did remittances. There were changes in the made the difference. occupational structure, with workers moving away from farm Concluding Remarks and daily work toward salaried employment, to jobs likely to be of higher productivity. There was also a sharp shift in em- The last decade saw an unprecedented reduction in poverty ployment away from agriculture and toward the higher pro- across the globe, which provides a fantastic opportunity to ductivity manufacturing and service sectors. Moreover, there study the most significant factors that were at work in favor of was an improvement in the educational composition of the the poor. In a sample of 16 countries that showed significant workforce over the last 10 years in each of these countries, a declines in poverty, analysis shows that the most important result of higher investment in education in previous decades, contributor to changes in moderate poverty has been the with a smaller share of the population illiterate by the end of growth in labor income. While a declining dependency rate 2010, a higher share of the workforce having completed pri- has been a significant contribution to poverty reduction, the mary and lower secondary school in Bangladesh and Thai- main source of changes in poverty have been labor market–re- land, and a higher share of the population having completed lated factors: the number of occupied adults per household secondary and tertiary school in Peru and Thailand. increased, but overall, increases in real earnings among the However, specific patterns across the income distribu- poor mattered the most. tion vary across countries, and the potential role of the differ- In Bangladesh, Peru and Thailand, improvement in labor ent factors in reducing poverty is clearly different. Moreover, market conditions are the main factors behind the sharp re- the starting points are very different; Bangladesh, despite ductions in poverty. The observed growth in incomes of the strong growth, is still a low-income country, with a GDP per poor was mainly due to higher returns to endowments signal- capita of US$1,710, while Peru and Thailand are firmly in ing an increase in the marginal value of work, either due to the middle-income country ranks, with GDP per capita of increases in productivity and/or higher relative price of labor. US$10,439 and US$9,630, respectively (all figures in pur- In both Bangladesh and Peru, labor incomes of the poor (in chasing power parity [PPP] terms). Peru is already highly ur- agriculture or the less educated) increased faster than for the banized, as opposed to Thailand and Bangladesh, where the rest of the population. In contrast, in Thailand, greater educa- share of urban population is still below 30 percent. tional attainment and higher returns to human capital seem To determine which of these changes has been most im- to have boosted the marginal value of work, potentially portant in reducing poverty, Inchauste et al. (2012) follow through productivity increases. Bourguignon, Ferreira, and Lustig (2005) to account for oc- However, when accounting for changes in the extreme cupational changes, sectoral and educational shifts, changes poverty headcount (measured by a lower poverty line), the 4 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Table 1. Contributions to the Change in Poverty Head Count Ratioa Contribution to national Bangladesh, 2000–2010 Peru, 2004–10 Thailand, 2000–2009 moderate poverty reduction on account of Percent point Share of total Percent point Share of total Percent point Share of total changes in: change change (%) change change (%) change change (%) Nonfarm labor income -4.56 26 -9.35 58 -3.46 27 Returns to -3.52 20 -4.93 31 -1.25 10 endowments Occupational -1.61 9 -3.44 21 0.08 -1 choice Economic sector -0.48 3 -0.08 1 -1.01 8 Education -0.55 3 -0.25 2 -1.34 10 Unobservable 1.59 -9 -0.65 4 0.06 0 factors Farm income -6.02 35 -2.74 17 -4.91 38 Returns to -6.98 40 -2.04 13 -4.83 38 endowments Occupational 0.56 -3 -0.25 2 1.31 -10 choiceb Economic sectorb -0.14 1 -1.11 9 Education 0.13 -1 -0.08 1 -0.56 4 Unobservable 0.26 -2 -0.23 1 0.28 -2 factors Nonlabor income 1.05 -6 -2.28 14 -5.8 45 of which: International -1.94 11 0.19 -1 -2.19 17 transfers Other transfers 1.68 -9 0.25 -1 -1.12 9 Capital 1.31 -8 -0.58 4 0.02 0 Public transfers 0 0 -1.38 9 -2.51 20 Other -7.5 43 -0.78 5 2.5 -20 Age/gender -3.48 20 -1.17 7 -1.18 9 Consumption-to- 0.93 -5 -1.73 11 3.43 -27 income ratio Unexplained -4.95 29 2.12 -13 0.26 -2 Total -17.34 100 -16.13 100 -12.84 100 Source: Authors’ compilation. a. Changes in poverty are defined using the national moderate poverty line in each case. In Bangladesh, this line is equivalent to US$1.08 a day; in Peru, it is equivalent to US$4.45 on average; and in Thailand, it is equivalent to US$3.53 on average, all in PPP terms. b. Refers to the secondary occupation of individuals who work as self-employed agricultural workers. 5 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise contribution of transfers was relatively large. It is critical to by the availability of public of private transfers. At the same note that this exercise does not take into account the payoff of time, transfers might be conditioned on the level of income increased access to many public services that are not part of and employment status of the family members. household income. It also does not account for the poverty 8. Results using extreme poverty lines are available upon impact of improvements in the quality of public services. request. About the Authors 9. Results using extreme poverty lines are available upon request. Gabriela Inchauste is a Senior Economist in the Poverty Reduc- tion and Equity Group for the Poverty Reduction and Economic References Management (PREM) Network. João Pedro Azevedo is a Senior Azevedo, J., G. Inchauste, S. Olivieri, J. Saavedra, and H. Winkler. Forthcoming. “Is Labor Income Responsible for Poverty Reduc- Economist in Poverty, Gender, and Equity for the Latin America tion? A Decomposition Approach.� World Bank Policy Research and Caribbean Region. Sergio Olivier is an Economist in the Pov- Working Paper. erty Reduction and Equity Group in the PREM Network. Jaime Barros, Ricardo Paes de, Mirela de Carvalho, Samuel Franco, and Saavedra is the Sector Director of the Poverty Reduction and Eq- Rosane Mendoça. 2006. “Uma Análise das Principais Causas da uity Group in the PREM Network. Hernan Winkler is a Consul- Queda Recente na Desigualdade de Renda Brasileira.� Revista tant in the Poverty, Gender and Equity Group for the Latin Econômica 8 (1): 117–47. Universidade Federal Fluminense. Rio de Janeiro, http://www.uff.br/revistaeconomica/V8N1/ America and Caribbean Region. RICARDO.PDF. Notes Bourguignon, François, Francisco Ferreira, and Nora Lustig, eds. 2005. The Microeconomics of Income Distribution Dynamics in East 1. For more details on global changes in poverty see Chen and Asia and Latin America, 17–46. Washington, DC: World Bank. Ravallion (2008) and World Bank (2012). Chen, S., and Martin Ravallion. 2008. “The Developing World is 2. For a recent review, see Essama-Nssah (2012). Poorer Than We Thought, But No Less Successful in the Fight Against Poverty.� World Bank Policy Research Working Paper 3. Both approaches are documented in Inchauste et al. (2012) 4703, Washington, DC. 4. Panel data that can track the life and labor histories of Datt, Gaurav, and Martin Ravallion. 1992. “Growth and Redistribu- households over time can be used to answer questions about tion Components of Changes in Poverty Measures: A Decom- economic mobility and poverty dynamics. However, panels position with Applications to Brazil and India in the 1980s.� are often not available with the frequency required. Moreover, Journal of Dev Economics 38: 275–96. panel data are often not representative of the population as a Essama-Nssah, B. 2012. “Identification of Sources of Variation in whole; and if they initially are, it is unlikely that over the Poverty Outcomes.� World Bank Policy Research Working Paper No. 5954, Washington, DC. course of a decade the panel would remain representative of Ferreira, Francisco. 2010. “Distributions in Motion: Economic the population. Alternative methods using repeated cross- Growth, Inequality and Poverty Dynamics.� World Bank Policy sections have been used. One approach is to construct pseudo Research Working Paper No. 5424, Washington, DC. panels, which can delve into some issues of economic mobili- Inchauste, G., S. Olivieri, J. Saavedra, and H. Winkler. 2012. “What ty (Lanjouw and McKenzie 2011). However, these models are Is behind the Decline in Poverty Since 2000? Evidence from often troubled by their lack of precision and the fact that they Bangladesh, Peru and Thailand.� World Bank Policy Research Working Paper No. 6199, Washington, DC. often do not measure the contributions of different factors to Lanjouw, Luoto, and David McKenzie. 2011. “Using Repeated poverty reduction. Cross-Sections to Explore Movements into and out of Poverty.� 5. That is, an increase in the number of employable adults in World Bank Policy Research Working Paper No. 5550, Wash- the household with respect to the household size. ington, DC. 6. This was true for all Latin American countries in the Ravallion, Martin, and Shaohua Chen. 2007. “China’s (Uneven) Progress Against Poverty.� Journal of Dev Economics 82 (1): 1–42. sample and for Bangladesh (based on Azevedo et al [forth- World Bank. 2012. “An Update to the World Bank’s Estimates of coming]). Consumption Poverty in the Developing World.� Briefing Note 7. Note that both factors could be interdependent, as deci- prepared by Shaohua Chen and Martin Ravallion, Development sions regarding labor force participation could be influenced Research Group, http://econ.worldbank.org/povcalnet. The Economic Premise note series is intended to summarize good practices and key policy findings on topics related to economic policy. They are produced by the Poverty Reduction and Economic Management (PREM) Network Vice-Presidency of the World Bank. The views expressed here are those of the authors and do not necessarily reflect those of the World Bank. The notes are available at: www.worldbank.org/economicpremise. 6 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise