WPS6384 Policy Research Working Paper 6384 Fifteen Years of Inequality in Latin America How Have Labor Markets Helped? João Pedro Azevedo María Eugenia Dávalos Carolina Diaz-Bonilla Bernardo Atuesta Raul Andres Castañeda The World Bank Poverty Reduction and Economic Management Network Poverty, Gender and Equity Unit March 2013 Policy Research Working Paper 6384 Abstract Household income inequality has declined in Latin effect (other components, within skill groups, affecting America in the past decades, contributing significantly labor income). The results show that falling returns to to poverty reduction in the region. Although available skills for both education and experience is, on average, evidence shows that changes in the labor income are driving the decline in labor income inequality in Latin among the main factors behind these inequality trends, America. The quantity effect, in turn, has contributed few studies have analyzed more closely the labor market little to inequality reduction, mostly attributable to a dynamics that have led to a decline in total income larger dispersion in years of experience, possibly linked inequality in some countries, but also to an increase to the region’s demographic transition and to significant in others. Using household survey data for a sample of increases in female labor force participation. Additional 15 countries in Latin America from 1995 to 2010, this findings show that wage inequality, still high in the paper uses an extension of the Juhn-Murphy-Pierce region, is coupled with inequality in terms of hours methodology to decompose changes in labor income worked. The paper complements the existing literature by inequality (hourly wages) into a quantity effect (capturing presenting separate results for males and females, as well changes in the distribution of workers’ skills), price as formal and informal sector workers as an attempt to effect (reflecting returns to skills), and unobservables control for secular shifts in these characteristics. This paper is a product of the Poverty, Gender and Equity Unit, Poverty Reduction and Economic Management Network. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank. org. The author may be contacted at jazevedo@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Fifteen Years of Inequality in Latin America How Have Labor Markets Helped? 1 João Pedro Azevedo** María Eugenia Dávalos** Carolina Diaz-Bonilla* Bernardo Atuesta+ Raul Andres Castañeda* Keywords: Inequality; Decomposition; Labor Income; Latin-America JEL Codes: Q15; I24; J30 Sector Board: Poverty Reduction (POV) 1 The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development, World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. The authors’ benefitted from comments from Louise J. Cord, Augusto de la Torre, Leonardo Gasparini, Luis Felipe López Calva, Samuel Freije, Julian Messina, Amparo Ballivian, Maurice D. Kugler, José Antonio Cuesta Leiva, Nora Luistig, Jamele Rigolini, Francisco Ferreira, Sergei Soares, Miguel Foguel, and Leopoldo Tornarolli. The author’s would also like to thank the comments received from Catherine Porter and other participants of the 32nd IARIW General Conference held in Boston, 2012. The usual disclaimer applies. * author’s from the Poverty, Gender and Equity Unit from the Poverty Reduction and Economic Management Team (LCSPP) in the Latin America and Caribbean Region from the World Bank; (**) author from Poverty Reduction and Economic Management Team from the Europe and Central Asia Region of the World Bank; (+) Paris School of Economics. Corresponding author: jazevedo@worldbank.org I. Introduction In the past decades, Latin America has seen reductions in both poverty and inequality. From 1995 to 2010, the region achieved a decline in poverty of around 18 percentage points, with the moderate poverty rate going from 46 to 28 in this period. Most of the reduction took place in the last decade when the rate of decline significantly accelerated (Figure 1). Similarly, the Gini coefficient of total household income per capita declined by 9 percent from 1995 and 2010, going from 0.57 to 0.52. 2 Consistent with the sharper decline in poverty in the last decade, inequality also declined more rapidly from 2000 to 2010 than in previous periods (Table 1 and Figure A1). The fall in inequality has indeed played an important role in reducing poverty in the region. Decomposing changes in poverty into a growth and a redistributive component (Datt and Ravallion, 1992) shows that for the past decade inequality had a significantly larger contribution to poverty reduction than that from economic growth alone. In fact, between 2000-2005 and 2005-2010, the decline in inequality accounted for 58 and 37 percent, respectively, of the total poverty reduction in Latin America, and close to 44 and 40 percent, for extreme poverty (Table 2). This is true for many countries in the region. For example, in Brazil, 60 percent of the total change in poverty between 2000 and 2005 is due to redistributive effects. In the same period, growth contributed to poverty increases in Argentina, which were offset by strong redistributive effects. With a break from a historically high and persistent inequality in Latin America, it is key to better understand what has driven the declining trend. A number of author’s have shown most of the income inequality in the Latin American region is generated in the labor market. The analysis of household survey of 15 countries from 1995 to 2010 suggests that although, on average, labor income inequality has reduced its contribution to total household income inequality in Latin America, reducing from 77 to 74 percent of total per capita household income inequality from 1995 to 2010 (Figure 2). Labor income still accounts for the highest share of total per capita household income in the region (Figure A2) and remains the main contributor to inequality. Given the importance of earnings in driving the overall inequality trends in the region, this paper aims at disentangling the factors behind the decline in labor income inequality (hourly wages) in the past fifteen years in Latin America. Using an extension of the methodology by Juhn-Murphy- Pierce (1993) which decomposes labor income inequality into a quantity, price and unobservable 2 Gini calculated pooling data for all countries in the sample and excluding zero values. Other specifications are presented in Table 1. 2 (residuals) effects, we explain the trends in and drivers of labor income inequality in the region and highlight the differences in patterns across Latin American countries. Using four measures of inequality, including the commonly used Gini coefficient and the Theil-T index, findings show that the price effect, which captures returns to skills (education and experience), has been, on average, the main driver of the inequality decline. Understanding the factors behind the declining labor income inequality has important policy implications. First, it helps determine, at least partially, how the region broke with its persistent inequality. Second, the analysis can be useful for tackling inequality in countries that have not yet joined the declining trend, both in Latin American and possibly in other regions as well. Finally, it can better inform policymakers on (i) whether the decline is likely to be sustainable over time, (ii) the possible threats to the path towards further reducing inequality and (iii) policy options that could contribute to further falls in inequality. The next section briefly reviews some of the recent literature exploring the declining trend in inequality in the region, including labor income inequality. Section III details the JMP methodology and the adaptations employed in this paper. Section IV describes the data and the empirical strategy, while Section V and VI provide detailed results for the region and for each country in the sample. Conclusions are presented in Section VII. II. Literature Review Latin America has been singled out as the world's most unequal region. As such, a growing literature has tried to understand the historic reasons behind its persistent and high inequality, as well as the determinants behind the recent declining trends. This section provides a brief overview of the most recent work on inequality in the region. Putting the recent decline in inequality into a historical perspective, Lustig and Gasparini (2011) note that inequality trends in Latin American countries have undergone two distinct periods in the past three decades. During the crisis of the 1980s and 1990s, and the period of structural reforms of the 1990s, most of the countries in their analysis experienced an increase in inequality. This trend seems to be related to the macroeconomic crises that took place in those two decades, coupled with inexistent or inefficient social safety nets and regressive effects of structural adjustment programs. 3 A number of authors have been recently analyzing the main factors behind the recent inequality decline in the region. López-Calva and Lustig (2010) compile a detailed analysis of the inequality trends in four countries in the region: Argentina, Brazil, Mexico, and Peru. Results show that the decline in inequality in these countries can indeed be attributed to two main factors: first, a shrinking earnings gap between skilled and low-skilled workers, from an expansion in education in the last decades. This effect was not compensated, as in the 1980s and part of the 1990s, by a higher demand for skilled labor. Second, from an equalizing effect of government transfers, related to larger and better targeted conditional cash transfer programs in these countries. Evidence in Figure A3 shows that transfers (public and private) have the highest inequality-reducing marginal effect of the various household income sources, at -2.2 percent in 2010. Azevedo, Inchausete and Sanfelice (2012) decompose the recent observed changes in inequality in 16 Latin American countries over the last decade in order to find the main contributors to the observed reduction in inequality. In contrast to methods that focus on aggregate summary statistics, their method generates entire counterfactual distributions, allowing them to account for changes due to demographics, labor income, transfers, pensions and other non-labor income sources. The results shows that for most countries in the sample, the most important contributor to the observed decline in inequality has been the relatively strong growth in labor income at the bottom of the income distribution (over 43 percent of the total change in the Gini for the period). Other factors are also linked to the falling inequality. For instance, recent studies refer to the role of social-democratic political regimes in the region during the past decade, and how the policies put in place by them had a more pronounced redistributive effect (Cornia, 2010; Birdsall, Lustig and McLeod, 2011). Moreover, the shrinking wage gap between skilled and unskilled workers in Argentina, for example, seems to be related to factors such as the commodity boom of the last decade, the exchange rate devaluation, and the role of labor unions, all of which pushed up the demand for unskilled labor relative to skilled labor (López-Calva and Lustig, 2010). These and other studies point to labor income as one key factor of inequality changes. However, most of the existing literature analyzing income inequality in Latin America focuses on total income inequality, and more in-depth labor markets analyses are only available for a limited number of countries. A few important exception in this literature is the work of Battistón et al. (2011) and Gasparini et al (2011) which also addresses the issue of earnings inequality. Battistón et al. (2011) focus on the determinants of the level of inequality and not on the change of inequality over the recent years, their work does confirm the presence of the so called “paradox of 4 progress�: changes in education induced higher income inequality levels (through the highly convex structure of returns) in most countries in the region, and that the changes were more unequalizing in the 1990s than in the 2000s. Gasparini et al (2011) apply the Murphy-Katz decomposition to establish which factor related to the demand and suply of labor by skill level can account can account for this differential evolution of inequality in 16 Latin American countries over the decades of the 1990s and the 2000s, concentrating on the change in the distribution in terms of return to skills (as proxied by educational attainment). Their paper disentangled the relative contribution of supply and demand factors in explaining wage premia evolution. In a context of constant increase in the relative supply of skilled and semi-skilled workers, Tinbergen’s framework suggests that differential evolution indicates a strong shift in demand towards skilled labor in the 1990s and a deceleration of this relative demand in the second period. Hence, this paper contributes to the literature by providing a regional perspective on how labor income inequality has shifted in the region. The analytical framework chosen in this paper allow us to more clearly disentangle how the educational and demographic (here define of years of experience), components have contributed to the changes of inequality over time, as well as the residual wage inequality—that is, the wage dispersion within demographic and skill groups increased simultaneously. Moreover, this approach can also help us understand if the observed changes were driven by changes in returns or the composition. One last, important contribution of this analysis, are the separate results by males and females; as well as, formal and informal sector works as an attempt to control for secular shifts in these characteristics. III. Methodology This paper uses the Juhn-Murphy-Pierce (JMP, 1993) methodology to decompose labor income inequality, with an extension proposed by Foguel and Azevedo (2007) that allows for a counterfactual interpretation of inequality changes over time. One advantage of this methodology vis-à-vis alternative methods, such as the Oxaca-Binder decompositon, is the possibility to account for within-group inequality, captured by the inclusion of the residual term in the counterfactual distribution. This is particularly important for the decomposition of distributional sensitive measures such as the ones analyzed in this paper. The Juhn-Murphy-Pierce methodology 5 The JMP approach is based on Mincer-type Ordinary Least Squares (OLS) regressions that allow decomposing labor income inequality, using any measure of inequality, in three parts. First, a quantity effect which refers to the distribution of observable workers’ characteristics, such as education and labor market experience, and are included as regressors in the equation. Second, a price effect which captures changes in returns to observed characteristics through the regression’s coefficients. Third, the regression residual reflects changes in inequality within education and experience groups driven by unobserved factors. The starting point is a Mincerian equation: y= it X it βt + uit , (1) where i represents a worker observed in time t , yit is the log of labor income, X it represents the vector of the worker’s observable characteristics, β t the vector of coefficients for time t, and uit the error term assumed to have zero mean (i.e. E[uit | X it ] = 0 ). Let Ft (. | X it ) be the conditional cumulative distribution of the residuals for period t. Denoting θ it as the percentile of individual i at time t in the residuals distribution, equation (1) can be expressed as: yit = X it βt + Ft −1 (θit | X it ). (2) Changes in earnings over time can occur from (i) changes in the distribution of workers’ observable characteristics, X it , known as the quantity effect; (ii) changes in returns to these observed characteristics, β t , or the price effect,; and, finally, (iii) changes in the distribution of unobservables ( F −1 (. | X ) ). This framework allows us to simulate the distribution of earnings for each period t by keeping some components fixed, i.e., by substituting one or more of the right-hand side components with their mean over time. Particularly, let β be the vector of observable characteristics for a regression including all years; similarly; F (. | X it ) is the conditional distribution of the residuals of that regression. By rewriting equation (2) with these components as 1 yit = X it β + F −1(θit | X it ). , (3) 6 it can now be interpreted as the distribution of labor income in period t when keeping prices and residuals constant, so that only the observable characteristics, Xs, change over time. Following a similar approach, we can once more rewrite equation (2) to simulate the distribution of earnings by letting both quantities and prices vary over time, while keeping the distribution of residuals fixed. This equation will be 2 yit = X it βt + F −1(θit | X it ). (4) A third and final simulation allows for all components to change over time, reflecting the original distribution of earnings, so that y= 3 it X it βt + Ft −1 (θit | X it ) ≡ yit , . (5) With all three simulated labor income distributions in place, the concept of inequality is introduced. Let D (.) be any measure of inequality, such as the Gini coefficient or the Theil index. If Yitk = exp ( yit k ) , k=1,2,3, the contribution of quantities, prices and unobservables to total inequality in period t (i.e., Tt = D (Yit ) )can be expressed as Qt D(Yit1 ), = (6) =Pt D(Yit2 ) − D(Yit1 ) (7) and Rt = D(Yit3 ) − D(Yit2 ). (8) The sum of each of these components in period t equals total inequality namely Rt D(Y= Qt + Pt + = 3 it ) D(Y= it ) Tt , so that total inequality is decomposed into contributions of the quantity, price and unobservables effects. The JMP methodolgy just described has been widely used and allows for the decomposition to be interpreted as the contribution of each component to inequality in a particular year. However, a 7 limitation of this approach is that the overall methodology is not suited for comparisons of how each effect contributes to inequality over time. More specifically, let’s consider two time periods, −1 −1 τ ′ and τ ′′ , and simplify the notation of F (θit | X it ) = F and Ft (θit | X it ) = −1 Ft −1 . Taking time differences for Qt , Pt and Rt we arrive at the following: = Qτ ′′ ( ) ( − Qτ ′ D exp ( X iτ ′′ β + F −1) − D exp ( X iτ ′ β + F −1) , ) (9) = Pτ ′′ −Pτ′ ( )  D exp ( X ′′ β ′′ + F −1) − D exp ( X ′′ β + F −1)   iτ τ iτ (  ) ( ) ( ) −  D exp ( X iτ ′ βτ ′ + F −1) − D exp ( X iτ ′ β + F −1)    (10) and Rτ ′′ − Rτ ′  D ( exp ( X iτ ′′ βτ ′′ + Fτ− =  1 ( ′′ ) ) − D exp ( X ′′ β ′′ + F iτ τ −1 )   ) −  D ( exp ( X iτ ′ βτ ′ + Fτ−  ( ′ ) ) − D exp ( X ′ β ′ + F 1 iτ τ −1 ) ) .  (11) As mentioned before, JMP is limited in providing information on changes over time in the contributions to inequality of each component. The exception is the first component, the quantity effect, expressed in (9). More specifically, the time differences in (9) show that the only component −1 that changes between τ ′ and τ ′′ is the observable characteristics, while the β and F remain fixed. Therefore, this difference in fact reflects the effect of changes in quantities between the two time periods. Conversely, expressions (10) and (11) fail to provide a temporal interpretation. In (10), for instance, the time difference in the price component cannot be interpreted as the contribution of the price effect to changes in inequality. This is because it is not only the prices (i.e. the β s) that change in Pτ ′′ −Pτ′ , but also the Xs . Unless the distribution of quantities remains fixed over time, JMP is limited in providing a counterfactual interpretation of the price effect. A similar analysis 8 leads us to the conclusion that a counterfactual analysis cannot be derived from Rτ ′′ − Rτ ′ , given −1 −1 that changes over time cannot be only attributed to changes between Fτ ′ and Fτ ′′ . Adapting JMP for a counterfactual interpretation This study presents a modification to the original JMP method by Foguel and Azevedo (2007), so that it allows for a counterfactual interpretation over time. By letting s be a fixed time period (e.g., 2000) we can rewrite equations (3), (4) and (5) as follows: ∗1 yit = X it β s + Fs−1 (θit | X it ), (12) ∗2 yit =X it βt + Fs−1 (θit | X it ) (13) and = X it βt + Ft −1 (θit | X it ) ≡ yit , ∗3 yit (14) where Fs (θit | X −1 = it ) Fs−1 ( F (uit | X it )) , denoted as Fs−1 for simplicity. Equation (12) simulates labor income allowing quantities to change over time, but keeping prices and residuals fixed at a reference period s. The difference with (3) is therefore straightforward: while (12) leaves prices and residuals fixed at a specific period, (3) uses the mean of prices and residuals for all periods under consideration. Similarly, equation (13) simulates a distribution of labor income where quantities and prices vary over time (as in equation (4) in the JMP methodology), but in which the distribution fo residuals is that from s. As equation (14) allows all components to vary it is identical to (5). ∗k ∗k Following the same steps of JMP of (6), (7) and (8), and with Yit = exp ( yit ) , k = 1, 2, 3 , the quantity, price and unobservable components for period t are as defined follows: Qt∗ D(Yit∗1 ), = (15) 9 =Pt ∗ D(Yit∗2 ) − D(Yit∗1 ) (16) and Rt∗ = D(Yit∗3 ) − D(Yit∗2 ). (17) As before, the sum of the three components equals total labor income inequality, i.e. ∗ Qt∗ + Pt ∗ + = Rt∗ D(Yit 3 ) ≡ D(Y =it ) Tt . Note also that for t = s , Qs∗ = Ts and P =s ∗ = Rs ∗ 0. This modification of the original JMP provides a counterfactual interpretation of changes in labor income inequality over time between any time period t and time period s. This is derived from the following expressions: Qt∗ − Qs∗ D ( exp ( X it β s + Fs−1 ) ) − D ( exp ( X is β s + Fs−1 ) ) , = (18)  D ( exp ( X it βt + Fs ) ) − D ( exp ( X it β s + Fs ) )  Pt ∗ − Ps∗  = −1 −1   D ( exp ( X is β s + Fs ) ) − D ( exp ( X is β s + Fs ) )  − −1 −1  = D ( exp ( X it βt + Fs−1 ) ) − D ( exp ( X it β s + Fs−1 ) ) (19) and  D ( exp ( X it βt + Ft ) ) − D ( exp ( X it βt + Fs ) )  Rt∗ − Rs∗  = −1 −1   D ( exp ( X is β s + Fs ) ) − D ( exp ( X is β s + Fs ) )  − −1 −1  = D ( exp ( X it βt + Ft −1 ) ) − D ( exp ( X it βt + Fs−1 ) ) . (20) 10 The difference in (18) shows that only the Xs change between t and s, so that it can be interpreted as the effect of changes in quantities on inequality between this two periods. This interpretation can also be derived from the original JMP, with the difference that in JMP the reference period in which prices and unobservables are kept fixed for evaluating changes in quantities over time, is the mean of all periods instead of s. The main difference comes when evaluating (19) and (20). In (19), for example, the difference ∗ ∗ between Pt and Ps can now only be attributed to changes in prices between t and s, as the second term in brackets will equal zero. In sum, (19) provides a counterfactual interpretation of changes in total labor income inequality between t and s from price changes between those two time periods. A similar interpretation is derived from (20) for the case of unobservables, capturing only the effect of changes in the unobservable component in changes of total labor income inequality between t and the reference period s. As described above, this adaptation of the JMP methodology allows for a counteractual interpretation of the quantity, price and unobservables effects between s and t. It is important to keep in mind, however, that it does not allow for the evaluation of these effects’ contributions to total inequality between any two periods τ ′ and τ ′′ , where none of them are the reference s. IV. Data and Empirical Strategy The data used in this paper are from a harmonized database of household surveys from 15 Latin American countries compiled in the Socio-Economic Database for Latin America and the Caribbean (SEDLAC), a joint effort of the Centro de Estudios Distributivos Laborales y Sociales of the Universidad Nacional de La Plata and the World Bank’s poverty group for Latin America and the Caribbean. The countries included in this analysis are Argentina, Bolivia (urban), Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Honduras, Mexico, Panama, Paraguay, Peru, and Uruguay (Montevideo-urban). So as to make time periods comparable across time we use the circa criteria for the years 1995, 2000, 2005 and 2010. Tables A1 and A2 provide more detail of the countries, years and surveys included in this study. Labor income is calculated as individual hourly wages (only for individuals with positive wages in the dataset), and all regional estimations are computed as working population-weighted averages. For the variables of observed workers’ characteristics ( X it ) we use two measures of skills: years of education and potential experience in the labor market. The latter is measured as age minus years 11 of education minus six. All regressions are estimated using OLS, and workers’ characteristics are included in the regression as dummies, with years of education covering a range from 0 to 17 years; experience including the following categories of potential experience (as in JMP, 1993): 0-10, 11-20, 21-30 and 31 or more; and finally, an interaction term of the years of education and potential experience dummies. The JMP methodology does not require for the same individual to be followed over time, i.e., panel data, so we use four years of cross-sectional data (1995, 2000, 2005 and 2010) to run the decompositions. The reference period chosen for this study is the year 2000, so that all results are interpreted as deviations of inequality (or each of the components) from that year. Following Foguel and Azevedo (2007) we estimate equation (1) for each available year. The regression residuals are ranked in ascending order for each year and divided in percentiles. For each percentile in each year we estimate the mean to create a discrete empirical approximation of Ft −1 (θit | X it ) . We employ this discrete distribution to construct the earnings of each individual (i) 𝑗 in year (t) as 𝑋𝑖𝑡 𝛽𝑡 + 𝐹𝑡−1 (𝜃𝑖𝑡 |𝑋𝑖𝑡 ), where 𝛽𝑡 is the vector of estimated coefficients in year t and θ j is the mean value of residuals in percentile θ j in which the individual was located. For Fs−1 ( F (uit | X it )) , we use the mean residuals for each percentile in the reference year (2000). For more details see Foguel and Azevedo (2007). Finally, the simulations of labor income of (12), (13) and (14) are calculated using the estimated coefficients and the discrete distribution of residuals, which in turn are used to estimate the final decomposition in (15), (16) and (17). We conduct our methods using four measures of inequality, namely the Gini coefficient, the Theil-T (GE(1)) index of inequality and the ratios of mean labor income between the 90th and 10th deciles (90/10), and the 80th and 20th deciles (80/20). One issue that is often raised in analyses of this nature is the importance of selection into the labor market, which can potentially bias the estimated coefficients of the earnings equation. Attempting to methodologically address this issue is beyond the scope of the current work. However, in order to test the robustness of the findings and explore the extent to which selection might be biasing our results, this paper also estimates separate models for Males-Females as well as Formal-Informal workers. This choice was based on two relevant trends in LAC: the increase in female labor force participation in the region between 1995-2010 (World Bank, 2012), as well as the expansion of 12 formal employment 3. As further discussed below, the findings of this paper support a qualitatively consistent story across these groups. V. Results 5.1 Decomposing Labor Income Inequality As a starting point to assessing changes in labor income inequality in Latin America, Table 3 presents the evolution of labor income inequality measures in the past fifteen years. All four measures suggest a monotonic decline, on average, in labor income inequality for the region. More specifically, the regional Gini (working-population-weighted averages of countries in the sample) declined at an average rate of -0.6 percent per year and the Theil by -1.3 percent from 1995 to 2010. However, not all countries joined the declining trend in labor income inequality in this period. The labor income Gini increased for Costa Rica, Honduras, Panama and Uruguay, while the Theil also increased in those four countries in addition to El Salvador and Bolivia. The fastest fall in the Gini took place in Brazil, with a -0.75 percent annualized rate from 1995 to 2010 (Figure 3). To explore the factors behind the regional fall in labor income inequality, Figure 4 and 5 present the adapted JMP decomposition results for Latin America. The first panel of each figure (panel A) shows the observed total changes in inequality from 1995 to 2010, with 2000 as the reference year. The rest of the panels (B-D) decompose the total changes into quantity, price and unobservable (residuals) effects. Moreover, Table 4 through 7 present detailed decomposition results for each country and each inequality measure for the period of the fastest inequality decline (2000-2010). A negative sign denotes a contribution to inequality decline, while a positive sign indicates that the component was inequality-increasing over this period. Quantity effect: focusing on the quantity effect (i.e., the contribution of changes in the composition of skills to labor income inequality, ceteris paribus), shows that in most measures (the exception being the 90/10 ratio) the quantity effect further reduced its already low contribution to inequality decline in 1995 (panel B of Figures 4 and 5), resulting in a very small share of inequality falls attributable to this factor by 2010. Results by country presented in Tables 4-7 for the last decade (2000-2010) show that in 5 out of 15 countries, the quantity component contributed to increasing the labor income Gini and the Theil index. The decompositions of the 90/10 and 80/20 measures 3Data on informality from SEDLAC (CEDLAS and The World Bank) show a decline in informality in most countries in the region. 13 show, however, a positive contribution of the quantity effect to labor income inequality in 11 and 9 countries, respectively. Price effect: the driving factor behind labor income inequality declines in the past fifteen years, independent of the measure of inequality used, was the falling returns to skills (Panel C), also known as the price effect. Between 2000 and 2010, for example, around 64 percent of the total change in the Gini coefficient can be attributed to declining returns to skills. This result is consistent in most countries in the sample, with an inequality-reducing effect of the skill premia in 12 out of 15 countries for the Gini, Theil index and 90/10 ratio (13 countries for the 80/20 ratio). In fact, one of the highest achiever in terms of declines in the labor income Gini coefficient and Theil Index from 2000 to 2010, Brazil, can attribute around 61 percent and 72 percent of the changes, respectively, to falling returns to years of education and experience. Conversely, in Costa Rica, for example, were the Gini coefficient of labor income increased in the past decade (2000 to 2010), both the quantity and the price effects were inequality-increasing. Other factors effect: The role of unobservables (within skill-group inequality, measured by the residual) is very heterogeneous across countries. This effect could be capturing a wide range of things not accounted for in our empirical strategy, such as quality of education, changes across sectors or occupations (including changes in demand for workers in specific sectors), among others. On average, inequality within groups decreased over time, although its contribution to total labor income inequality changes was relatively small by 2010. Nonetheless, this effect was particularly strong in some countries in terms of enhancing inequality, fully offsetting the role of the price and quantity effects. This is the case, for example, of Paraguay, where the Gini coefficient would have fallen between 2000 and 2010 driven by the reductions in returns, if it had not been more than compensated by the larger and positive contribution to inequality of within-group changes. To summarize the patterns across countries of the quantity, price and unobservables effects, Table 8 presents a typology of countries based on whether the various components were inequality- increasing or inequality-decreasing in the last decade in terms of the Gini coefficient. Only in four countries in the sample, i.e., Argentina, Brazil, Mexico and Peru, all three components moved in the same direction, thus enhancing the overall change in labor income inequality. 5.2 Decomposing Labor Income Inequality by Gender and Sector 14 Overall regional and country trends in labor income inequality could be masking differences across subgroups. We therefore apply the adapted JMP methodology to subsamples of the working population by gender and by formal/informal sector workers. To simplify the analysis, we focus on the Gini coefficient of labor income inequality for all four subgroups (Annex tables A3-A6) in the period 2000 to 2010. At a regional level, results show a larger decline in inequality for male workers compared to female workers from 2000 to 2010. From a country perspective, while labor income inequality declined in 12 out of the 15 countries for males, it declined in only seven countries for females. The decomposition results show that the price and unobservables effects were inequality-reducing for both groups, but much more powerful for males. The quantity effect, on the contrary, contributed to increasing inequality only for females, as the new women joining the labor market had on average more experience (age) and education (its effect on pushing down male labor income inequality, nonetheless, is very small). Labor income inequality in Latin America declined more in the informal than in the formal sector from 2000 to 2010. Results are very heterogeneous across countries; while inequality declined relatively more in the formal sector in Argentina, Bolivia and Peru, it fell relatively sharper in the informal sector for Brazil and Mexico. The price effect across sectors is very similar. The unobserved effects have a four times larger contribution to reducing inequality in the informal sector than in the formal sector, even if the quantity effect is inequality-increasing for the informal only. Looking at returns to skills and unobservables, returns to experience have declined relatively more for formal workers. 5.3 Price Effect: Unbundling returns to education and experience Given that falling returns to skills seem to be dominating, on average, inequality changes over time, we try to unbundle this price effect to better understand its dynamics. Figure 6 presents the mean returns to education, experience and unobservables over time (captured by the residual). 4. Results 4 Returns to education and experience are calculated from the coefficients of the Mincer equation (1) for each characteristic. The mean return to education for year t, for example, is calculated as a weighted average (by population share with each level of education) of the return to each level of education, divided by the weighted average (by population share with each level of education) of each level of education. A similar approach is taken for experience level, and in both cases the interaction terms are also included in the estimations. For the unobservables, equation (1) is rewritten as yit = X it βt + σ t ε it , where, = X it βt + uit assuming that ε it is a random independently and identically distributed variable (iid) following a normal 15 in Figure 6 show that returns to all three factors declined during the period of analysis and that the pace of reduction accelerated after 2005. Overall, mean returns to years of education and experience declined a total of 30 and 20 percent, between 1995 and 2010. Detailed results for males and females (Figures A4- A6) show that trends for both groups in returns to education are similar. In experience, we observe a decline for both groups, but slightly larger for males. It is important to notice that part of the price story captured in this decomposition is the effect of intertemporal changes on the quantities (supply of educated people, who can change the skill premium). An important implication to this, is that the after-mentioned quantity effect can be interpreted as a composition effect, as it is net of any quantity effect that might have picked up by the price effect (i.e. changes in the returns). The decline in returns to skills has been driven by a larger supply of experienced and educated workers in the region. Both mean years of education and experience have increased in the region for the working population (Figure 7), more sharply for education. Mean years of education and experience have increased for both sexes (Figures A5 and A6). For education, for example, investments in the past decades have resulted in a significant average increase in educational attainment of the population (1.7 additional years on average in Latin America). For experience, changes in the mean could be driven by an increase in female labor force participation and by the aging of the population (further discussed in section 5.4). In all countries in the sample, except for the Dominican Republic and Ecuador, average education levels of workers increased in the period (Table 10). The largest expansion took place in Brazil and Mexico, where years of education of workers increased a total of 35 percent and 26 percent, respectively, from 1995 to 2010. Similarly, mean years of experience also rose in all countries (except for a slight decrease in Bolivia), although at slower rates. Changes in returns to education show a very mixed picture across countries (Table 9). While returns to years of education declined a total of 43 percent in Brazil from 1995 to 2010, they increased in Argentina (40 percent) and Chile (83 percent). Similarly, while returns to experience fell by 34 percent in Mexico and 28 percent in Chile in the period under study, they increased by 38 percent in Honduras. distribution, N (0,1) , and σt is a factor (standard-deviation) that alters the dispersion of the distribution of errors, σt can be interpreted as capturing the “price� of unobservables. For more details see Foguel and Azevedo (2007). 16 5.4 Quantity Effect: Unbundling inequality of education and experience Previous results showed that on average, the quantity effect contributed very little to the reduction of inequality in the region. It is important to keep in mind that quantity, in particular education, still play 5 an important role in explaining the high level of inequality in the region (Battistón, et al 2011). This subsection explores the factors behind the quantity effect by looking at mean levels of education and experience and the dispersion in these characteristics over time. In other words, we further explore the composition of skills among workers, all else equal. The abovementioned expansion in years of education and experience has not been uniform across the population, resulting in changes in the distribution of these skills among the working population. On the one hand, the evolution of the standard deviation of years of education suggests that inequality in education slightly decreased (by around 2 percent) in the 2000s. This seems to be primarily driven by falling educational inequality of women (Figure A5). Overall, the reduction in educational inequality reflects a catch up from those at the bottom of the education distribution. For instance, the bottom income quintile in Latin America achieved an additional 1.8 years of education from 1995-2009, while the top quintile increased by 1.3 years. 6 On the other hand, the changing composition in years of experience has led to higher inequality of experience in workers in the past fifteen years (total increase of 1.8 percent from 1995-2010). In fact, in 11 out of 15 countries in the sample there was an increase in the standard deviation of years of experience among workers (Table 11). Both the mean and the standard deviation of experience have increased for men and women over time. By 2010, mean years of experience for women had increased more for than for men. Looking at the workforce by sector, informal workers have increased their mean education significantly more than formal workers; the changing composition of education has resulted, however, in a growing dispersion in education for informal workers and a decline in educational inequality for the formal sector. For the formal sector, the decline in educational inequality of the formal sector is likely offset by a sharp increase in inequality of experience, not observed in the informal sector (Figures A5 and A6). 5 Since the seminal work of Langoni (1973) several authors have found the effect of educational expansion was to increase inequality, including Bourguignon et al (2005), Reis and Barros (1991), Knight and Sabot (1983), Reyes (1988), and Lam (1999). 6 World Bank (2011). 17 As the education and experience effect are working in opposite directions, the overall quantity effects (net effect) is, on average, small. This suggests that the experience component is, on average, dominating in the overall JMP quantity effect. The question then arises of what is driving the dispersion in experience levels? Given that our experience variable (reflecting potential experience) is a construct including age, years of education and a constant, and given that years of education have increased and are less dispersed among the working population, it seems likely that the explanation behind a higher experience inequality lies in the aging of the working population. The age profile of people in the labor markets is likely related to two factors: (i) the demographic transition in the region, which has resulted in a bulge of newcomers into the labor market since 2000 and (ii) the increase of female labor force participation in the region. From 1995 to 2000 alone, occupied workers between 19-24 years old increased by around 14 percent in LAC (much higher than increases in subsequent periods at 11 percent from 2000 to 2005 and 5 percent from 2005 to 2010), as the largest birth cohort of the region enters the labor market (Cotlear 2010). As this cohort aged through 2010 and gained more experience in the labor market: (i) overall mean experience increased in the region, pushing down returns to experience and (ii) the dispersion of experience also increased, more so as they joined employment with very little experience in the first five years of our sample. The increase in experience inequality has persisted over time, but at lower rates. The demographic transition story is complemented by a generalized increase in female labor force participation in the region. The ratio of male/female ratio of workers rate went from 1.9 in 1995 to 1.5 in 2010. Figure A7 presents the growth rate of male and female workers from 2000 to 2010 in the region and by age. As shown, the increase in women workers is significantly higher than that of men, particularly for women in their late forties and early fifties. This is possible linked to, first, the higher increase in mean experience for women compared to men (Figure A6) (given that experience is an age construct). Second, it could also result in the lower dispersion in experience observed than for men, as the bulge of young workers entering the workforce with little experience (from the demographic transition) is partially offset by a relatively older group of women (estimated to have more experience) joining the labor markets. VI. Earnings Inequality 18 So far the analysis has concentrated in the evolution and factors behind inequality of hourly wages. This section aims at more explicitly linking inequality in hourly wages to total household income inequality. Following Juhn, Murphy and Pierce (1993), we use annual earnings as a proxy for income under the assumption that hourly wages hold a stronger link to annual earnings than to family income per-capita. Using annual earnings as a proxy for income is reasonable in this context, given that this source of income represents around three-quarters of total household income for Latin American households. To assess the contribution of inequality in hourly wages to that of annual earnings, we calculate the annual earnings as the product of the hourly wage and the number of hours worked per year. Departing from 𝑦 = ℎ + 𝑤, where 𝑦 is the log of annual earnings, 𝑤 is the log of hourly wages and ℎ is the log of hours worked per year 7, the variance of log annual earnings, σ2 y , is 2 2 σ2 2 y = σh + σw + 2σhw , 2 2 where σ2 w is the variance of log hourly wages, σh is the variance of log hours worked, and σhw is the covariance of log hourly wages and log hours worked. Figure 8 shows the variance of log annual earnings and its components for the population- weighted average of countries in our sample. Over the period as a whole, the movements of the annual earnings variance depended mostly on the hourly wage variance. The variance of the weekly hours worked also contributed to the annual earnings variance, especially in the increase of both 2005 and 2010. An interesting finding is that the covariance of hourly wages and weeks worked is negative for all countries (contrary to JMP, 1993, results for the United States). This means that the higher the hourly wage, the smaller the amount of hours worked per week. Although this negative covariance is slowly approaching zero for almost every country in the period under study, this result is the reflection of the high inequality not only in terms of hourly wages but also in terms of hours worked in Latin American countries. In other words, people that earn less per hour also work more hours per week. Finally, although wages are a key component of changes in total earnings, only 50% of the increase in the variance of annual earnings from 1995 to 2010 is due to the increase in the variance of hourly wages. This fact highlights the difference between the concepts of earnings and wage inequality, something that should be kept in mind when analyzing inequality trends. 7 Due to data constraints, we assume that all individuals worked 52 weeks per year. 19 VII. Conclusions Latin America is finally on a path towards reducing income inequality. To better understand the factors behind this trend and given that labor income contributes the most to total household income and to total income inequality, this paper explores the drivers of labor income inequality changes. The results show that a more equitable distribution of labor market income has been the main force behind falling inequality. The decline in labor income inequality, in turn, has been mainly driven by falling returns to education and experience. As inequality in the region remains high, two things should be considered in the path towards further inequality reduction. First, improved access to education, which has been a key driver falling inequality, needs to be coupled with improvements in skills. The educational system needs to strengthen its ability to generate the skills that are valued in the labor market (Aedo and Walker, 2012); in other words, quality of education and skills are the new margin for inequality. A recent study tests for the intergenerational persistence of inequality using PISA scores 8 and finds that Latin American countries have relatively higher rates of intergenerational persistence of inequality in educational achievements than, for example, countries in Asia (Ferreira and Gignoux, 2010). Also employing PISA data, the Human Opportunity Index for quality of education is consistently lower for science, mathematics and reading for Latin American countries than countries in Europe and North America (Molinas et al., 2010). Second, Latin America is currently undergoing a demographic transition with a larger proportion of working-age adults. As a result, the region is likely to generate a demographic dividend that can provide resources to be geared towards inequality and poverty-reducing investments. This favorable scenario is projected to continue until around 2020, when this ratio of workers/retirees should reach its maximum, before starting to decline again, this time due to the growing proportion of older persons and a relatively smaller workforce. It is important to notice that while such demographic transition lasted for over a century in developed countries, similar changes are occurring much more quickly in Latin America and other developing countries today. France had 115 years to accommodate the doubling of its elderly population from 7 percent to 14 percent; in Latin America this process is happening much more quickly and the adjustment will likewise need 8 PISA refers to the OECD Programme for International Student Assessment, which is an internationally comparable dataset that assess competencies in math, reading and science for 15 year old students in many countries. 20 to be quicker. Chile is projected to face this change in 26 years, Brazil in 21, and Colombia in 19 years. 9 Going forward it is important to continue to invest in country specific analysis, to deepen the understanding of the channels and mechanisms under which many of the stylized facts presented in this paper operate. In particular, it might be important to better understand the roles of some of the factors within particular occupational choices, and greater attention should be devoted to changes in the distribution of skills and characteristics over time. 9 Cotlear, Daniel (Editor) “Population Aging: Is Latin America Ready?� The World Bank: Washington, DC. 21 References Aedo, C. and I. Walker (2012). Skills for the 21st Century in Latin America and the Caribbean. Directions in Development Series. World Bank. Alvaredo, Facundo and Thomas Piketty (2010), “The Dynamics of Income Concentration in Developed and Developing Countries: A View from the Top�, in Luis F Lopez-Calva and Nora Lustig (eds.), Declining Inequality in Latin America: a Decade of Progress?, Brookings Institution Press and UNDP. Azevedo, Joao Pedro, Gabriela Inchauste and Viviane Sanfelice (2012) Decomposing the Recent Inequality Decline in Latin America. World Bank. Battistón, Diego, Carolina García Domench and Leonardo Gasparini (2011) Could an increase in education raise income inequality? Evidence from Latin America. CEDLAS Working Paper. Birdsall, Nancy & Nora Lustig & Darryl McLeod (2011). "Declining Inequality in Latin America: Some Economics, Some Politics," Working Papers 1120, Tulane University, Department of Economics. Bourguignon, F., Ferreira, F. and Lustig, N. (2005). “The microeconomics of income distribution dynamics in East Asia and Latin America�, Washington DC: Oxford University Press and The World Bank. CEPAL (2010). Hora de la igualdad: brechas por cerrar, caminos por abrir. Santiago, Chile. Chinhui Juhn, Kevin M. Murphy and Brooks Pierce (1993), "Wage Inequality and the Rise in Returns to Skill."; Journal of Political Economy, 101(3), pp. 410-42. Cornia, Giovanni Andrea (2010) “Income Distribution under Latin America’s New Left Regimes�. Journal of Human Development and Capabilities, 11(1). Cotlear, Daniel (Editor) (2010) “Population Aging: Is Latin America Ready?� The World Bank: Washington, DC. Ferreira, Francisco H. G. and Jérémie Gignoux (2010) “Educational Inequality and its Intergenerational Persistence: International Comparisons�. Foguel, Miguel Nathan and João Pedro Azevedo (2007), “Decomposição da desigualdade de rendimentos do trabalho no Brasil: 1995 -2005�, in Desigualdade de Renda no Brasil: uma análise da queda recente, Ricardo Paes de Barros, Miguel Nathan Foguel and Gabriel Ulyssea eds., vol. 2, ch. 27, Brasília: IPEA. Gasparini, L., S. Galiani, G. Cruces and P. Acosta. (2011) “Educational Upgrading and Returns to Skills in Latin America Evidence from a Supply-Demand Framework, 1990- 2010.� CEDLAS Working Paper. 22 Gasparini, Leonardo, Guillermo Cruces and Leopoldo Tornarolli (2008). “Is income inequality in Latin America falling?�, CEDLAS, Universidad Nacional de La Plata, August version. Gaurav Datt and Martin Ravallion (1992) "Growth and Redistribution Components of Changes in Poverty: A Decomposition with applications to Brazil and China in 1980s", Journal of Development Economics 38: 275-295. Helwege, Ann and Melissa B.L. Birch (2007). “Declining Poverty in Latin America? A Critical Analysis of New Estimates by International Institutions�, September, Tufts University, Tufts University Medford MA 02155, USA http://ase.tufts.edu/gdae. Knight, J. B. and Sabot, R. H. (1983). “Educational Expansion and the Kuznets Effect�. American Economic Review 73(5): 1132–36. Lam, D. (1999). “Generating Extreme Inequality: Schooling, Earnings, and Intergenerational Transmission of Human Capital in South Africa and Brazil.� Research Report 99-439. Population Studies Center, University of Michigan, Ann Arbor. Langoni, C. G. (1973). “Distribuição da Renda e Desenvolvimento Econômico do Brasil�. Rio de Janeiro, Brazil: Expressão e Cultura López-Calva, Luis Felipe and Nora Claudia Lustig (eds) (2010) “Declining Inequality in Latin America, A Decade of Progress?� Brookings Institution Press and United Nations Development Programme, c. 253pp. López-Calva, Luis Felipe and Nora Lustig, eds. (2010). Declining Inequality in Latin America: A Decade of Progress?, Brookings Institution Press and UNDP, Washington D.C. Lustig, Nora & Leonardo Gasparini (2011). "The Rise and Fall of Income Inequality in Latin America," Working Papers 1110, Tulane University, Department of Economics. Lustig, Nora & Luis F. Lopez Calva & Eduardo Ortiz-Juarez (2011). "The Decline in Inequality in Latin America: How Much, Since When and Why," Working Papers 1118, Tulane University, Department of Economics. Lustig, Nora (2009). Poverty, Inequality and the New Left in Latin America, Washington, D.C.: Woodrow Wilson International Center for Scholars, Latin American Program, October. Lustig, Nora (2009). Poverty, Inequality and the New Left in Latin America, Washington, D.C.: Woodrow Wilson International Center for Scholars, Latin American Program, October. McLeod, Darryl & Nora Lustig (2011). "Inequality and Poverty under Latin America's New Left Regimes," Working Papers 1117, Tulane University, Department of Economics. Molinas-Vega J., Barros, R., Saavedra, J. and Giugale, M. (2010) “Do Our Children Have a Chance? The 2010 Human Opportunity Report for Latin America and the Caribbean�. The World Bank. Pinkovskiy, Maxin and Xavier Sala-i-Martin (2009). “Parametric Estimations Of The World Distribution of Income�, NBER Working Paper 15433, Boston, Ma, www.nber.org/papers/w15433. 23 Pritchett, L. (2001). “Where has all the education gone?�. World Bank Economic Review, 15, 367- 391. Reis, J. Almeida dos and Barros, R. Paes de (1991). “Wage Inequality and the Distribution of Education: A Study of the Evolution of Regional Differences in Inequality in Metropolitan Brazil.� Journal of Development Economics 36: 117–43. Reyes, A. (1988). “Evolución de la Distribución del Ingreso en Colombia.� Desarrollo y Sociedad 21: 39–51. Székely, M. and Hilgert, M. 1999. “What’s Behind the Inequality We Measure?: An Investigation Using Latin American Data.� Research Department Working Paper Series No. 409. Washington, DC, United States: Inter-American Development Bank, Research Department. UNDP (2010). Actuar sobre el futuro: Romper con la transmision intergeneracional de la desigualdad, First Regional Human Development Report, Regional Bureau for Latin American and the Caribbean, http://www.idhalc-actuarsobreelfuturo.org/site/index.php. World Bank (2010) Did Latin America learn to shield its poor from economic shocks? Poverty and Labor Brief. Washington, DC: October, USA. World Bank (2011) A Break with History: Fifteen Years of Inequality Reduction in Latin America Poverty and Labor Brief. Washington, DC: April, USA. World Bank (2012) The Effect Of Women's Economic Power in Latin America and the Caribbean, Poverty and Labor Brief. Washington, DC: April, USA. 24 Figure 1. Headcount Poverty Ratio in Latin America, US$ 2.5/day and US$ 4/day (2005 PPP) and GDP per Capita PPP (constant 2005 international $), 1995-2010 50.0 11,000 GDP per capita, PPP (constant 2005 international $) 45.0 10,500 46.0 44.9 43.6 44.0 42.3 43.0 42.7 40.0 41.5 41.4 41.0 10,000 Poverty headcount (%) 38.1 35.0 9,500 34.1 30.0 31.6 30.6 30.4 9,000 27.9 27.7 28.0 25.027.5 26.8 26.7 26.7 25.1 25.7 24.5 24.1 8,500 20.0 21.9 18.6 15.0 17.2 8,000 16.4 16.1 14.0 10.0 7,500 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Extreme Poverty Moderate Poverty GDP per capita PPP (constant 2005 international $) Source: “On the Edge of Uncertainty, Poverty Reduction in LAC during the Great Recession and Beyond� by the Poverty and Gender Unit in LAC, The World Bank (2012). 25 Table 1. Gini and Theil-T index of Total Household Income Per Capita, circa 1995-2010 Gini Country 1995 2000 2005 2010 Argentina 0.48 0.50 0.49 0.44 Brazil 0.59 0.59 0.56 0.54 Bolivia 0.53 0.54 0.52 0.51 Chile 0.55 0.55 0.52 0.52 Colombia 0.55 0.59 0.55 0.55 Costa Rica 0.45 0.46 0.47 0.50 Dominican Republic 0.47 0.52 0.50 0.47 Ecuador 0.50 0.55 0.53 0.49 El Salvador 0.50 0.52 0.50 0.48 Honduras 0.55 0.54 0.59 0.55 Mexico 0.54 0.54 0.51 0.47 Panama 0.55 0.56 0.54 0.52 Paraguay 0.58 0.57 0.53 0.52 Peru 0.54 0.56 0.52 0.48 Uruguay 0.41 0.44 0.46 0.47 Data without zeros LAC (pooled data) 0.57 0.57 0.54 0.52 LAC (population weighted average) 0.56 0.56 0.54 0.51 Data with zeros LAC (pooled data) 0.58 0.58 0.55 0.53 LAC (population weighted average) 0.57 0.57 0.54 0.52 Theil Country 1995 2000 2005 2010 Argentina 0.43 0.46 0.45 0.35 Brazil 0.71 0.71 0.65 0.59 Bolivia 0.57 0.60 0.56 0.54 Chile 0.62 0.65 0.56 0.58 Colombia 0.69 0.80 0.65 0.65 Costa Rica 0.37 0.38 0.41 0.49 Dominican Republic 0.42 0.55 0.49 0.41 Ecuador 0.50 0.65 0.75 0.49 El Salvador 0.49 0.53 0.47 0.44 Honduras 0.64 0.57 0.70 0.61 Mexico 0.61 0.59 0.55 0.45 Panama 0.58 0.61 0.54 0.52 Paraguay 0.69 0.69 0.60 0.65 Peru 0.58 0.66 0.52 0.44 Uruguay 0.29 0.34 0.39 0.40 Data without zeros LAC (pooled data) 0.67 0.66 0.61 0.55 LAC (population weigthed average) 0.65 0.66 0.60 0.54 Data with zeros LAC (pooled data) 0.70 0.70 0.63 0.57 LAC (population weigthed average) 0.68 0.70 0.63 0.56 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank). 26 Table 2. Decomposing poverty changes: % of total poverty changes from growth and redistribution a. Poverty at $4/day Poverty Headcount Poverty Gap Poverty Gap Squared 1995- 2000- 2005- 1995- 2000- 2005- 1995- 2000- 2005- Countries 2000 2005 2010 2000 2005 2010 2000 2005 2010 Argentina Growth 0.43 0.30 -0.56 0.31 0.23 -0.49 0.30 0.16 -0.44 Distribution 0.57 -1.30 -0.44 0.69 -1.23 -0.51 0.70 -1.16 -0.56 Bolivia Growth 0.60 -0.90 0.23 0.47 -0.61 2.00 0.33 -0.42 0.32 Distribution 0.40 -0.10 -1.23 0.53 -0.39 -1.00 0.67 -0.58 0.68 Brazil Growth 2.00 -0.41 -0.68 0.97 -0.23 -0.63 0.53 -0.19 -0.63 Distribution -1.00 -0.59 -0.32 0.03 -0.77 -0.37 0.47 -0.81 -0.37 Chile Growth -1.00 -0.43 -0.80 -0.96 -0.38 -1.00 -1.10 -0.34 -1.66 Distribution 0.00 -0.57 -0.20 -0.04 -0.62 0.00 0.10 -0.66 0.66 Colombia Growth 0.68 -0.64 -1.04 0.59 -0.47 -1.02 0.53 -0.39 -0.94 Distribution 0.32 -0.36 0.04 0.41 -0.53 0.02 0.47 -0.61 -0.06 Costa Rica Growth -1.59 -1.35 -1.54 -1.38 -0.96 -1.32 -0.97 -0.72 -1.35 Distribution 0.59 0.35 0.54 0.38 -0.04 0.32 -0.03 -0.28 0.35 Dominican Rep. Growth -2.00 1.33 -0.73 -2.00 1.60 -0.49 -2.00 1.97 -0.40 Distribution 1.00 -0.33 -0.27 1.00 -0.60 -0.51 1.00 -0.97 -0.60 Ecuador Growth -2.00 -0.87 -0.46 -1.00 -0.88 -0.36 -1.00 -0.87 -0.30 Distribution 1.00 -0.13 -0.54 2.00 -0.12 -0.64 2.00 -0.13 -0.70 El Salvador Growth -1.77 1.00 -0.35 -1.00 0.63 -0.17 -0.51 0.38 -0.12 Distribution 0.77 -2.00 -0.65 2.00 -1.63 -0.83 1.51 -1.38 -0.88 Honduras Growth -0.55 -2.00 -0.79 -0.94 -1.00 -0.78 -2.00 -1.00 -0.78 Distribution -0.45 1.00 -0.21 -0.06 2.00 -0.22 1.00 2.00 -0.22 Mexico Growth -0.97 -0.44 1.00 -0.87 -0.40 0.56 -0.81 -0.38 0.46 Distribution -0.03 -0.56 -2.00 -0.13 -0.60 -1.56 -0.19 -0.62 -1.46 Panama Growth 0.90 -0.60 -0.60 2.00 -0.38 -0.47 0.93 -0.28 -0.39 Distribution 0.10 -0.40 -0.40 -1.00 -0.62 -0.53 -1.93 -0.72 -0.61 Paraguay Growth 1.75 0.31 -0.82 1.31 0.14 -0.92 0.87 0.10 -1.00 Distribution -0.75 -1.31 -0.18 -0.31 -1.14 -0.08 0.13 -1.10 0.00 Peru Growth -2.00 0.60 -0.69 -2.00 0.29 -0.74 -1.44 0.23 -0.78 Distribution 1.00 -1.60 -0.31 1.00 -1.29 -0.26 0.44 -1.23 -0.22 Uruguay Growth 0.34 0.72 -1.02 0.45 0.70 -0.89 0.57 0.70 -0.84 Distribution 0.66 0.28 0.02 0.55 0.30 -0.11 0.43 0.30 -0.16 LAC Growth -1.22 -0.42 -0.63 -1.36 -0.30 -0.59 -1.59 -0.26 -0.58 Distribution 0.22 -0.58 -0.37 0.36 -0.70 -0.41 0.59 -0.74 -0.42 27 b. Poverty at $2.5/day Poverty Headcount Poverty Gap Poverty Gap Squared 1995- 2000- 2005- 1995- 2000- 2005- 1995- 2000- 2005- Countries 2000 2005 2010 2000 2005 2010 2000 2005 2010 Argentina Growth 0.27 0.24 -0.46 0.28 0.16 -0.42 0.28 0.08 -0.37 Distribution 0.73 -1.24 -0.54 0.72 -1.16 -0.58 0.72 -1.08 -0.63 Bolivia Growth 0.56 -0.72 1.00 0.30 -0.36 0.22 0.19 -0.24 0.09 Distribution 0.44 -0.28 -2.00 0.70 -0.64 0.78 0.81 -0.76 0.91 Brazil Growth 1.00 -0.25 -0.66 0.49 -0.18 -0.62 0.27 -0.15 -0.62 Distribution -2.00 -0.75 -0.34 0.51 -0.82 -0.38 0.73 -0.85 -0.38 Chile Growth -0.78 -0.37 -1.05 -1.39 -0.31 -2.00 -2.00 -0.27 -1.00 Distribution -0.22 -0.63 0.05 0.39 -0.69 1.00 1.00 -0.73 2.00 Colombia Growth 0.62 -0.54 -1.10 0.51 -0.37 -0.93 0.44 -0.28 -0.79 Distribution 0.38 -0.46 0.10 0.49 -0.63 -0.07 0.56 -0.72 -0.21 Costa Rica Growth -2.00 -1.05 -1.18 -0.86 -0.63 -1.33 -0.51 -0.45 -1.72 Distribution 1.00 0.05 0.18 -0.14 -0.37 0.33 -0.49 -0.55 0.72 Dominican Rep. Growth -2.00 1.36 -0.51 -2.00 2.00 -0.37 -2.00 2.00 -0.30 Distribution 1.00 -0.36 -0.49 1.00 -1.00 -0.63 1.00 -1.00 -0.70 Ecuador Growth -1.00 -0.93 -0.35 -1.00 -0.89 -0.28 -1.00 -0.83 -0.24 Distribution 2.00 -0.07 -0.65 2.00 -0.11 -0.72 2.00 -0.17 -0.76 El Salvador Growth -1.00 0.90 -0.19 -0.41 0.34 -0.10 -0.23 0.23 -0.07 Distribution 2.00 -1.90 -0.81 1.41 -1.34 -0.90 1.23 -1.23 -0.93 Honduras Growth -0.59 -2.00 -0.79 -2.00 -1.00 -0.78 -1.00 -1.00 -0.79 Distribution -0.41 1.00 -0.21 1.00 2.00 -0.22 2.00 2.00 -0.21 Mexico Growth -0.89 -0.44 0.58 -0.79 -0.36 0.39 -0.71 -0.33 0.41 Distribution -0.11 -0.56 -1.58 -0.21 -0.64 -1.39 -0.29 -0.67 -1.41 Panama Growth 1.11 -0.43 -0.49 0.91 -0.26 -0.36 0.28 -0.19 -0.31 Distribution -0.11 -0.57 -0.51 -1.91 -0.74 -0.64 -1.28 -0.81 -0.69 Paraguay Growth 2.00 0.13 -1.12 0.77 0.08 -1.03 0.52 0.07 -1.12 Distribution -1.00 -1.13 0.12 0.23 -1.08 0.03 0.48 -1.07 0.12 Peru Growth -2.00 0.47 -0.77 -1.12 0.21 -0.81 -0.76 0.17 -0.80 Distribution 1.00 -1.47 -0.23 0.12 -1.21 -0.19 -0.24 -1.17 -0.20 Uruguay Growth 0.55 0.67 -0.85 0.83 0.69 -0.79 1.75 0.73 -0.77 Distribution 0.45 0.33 -0.15 0.17 0.31 -0.21 -0.75 0.27 -0.23 LAC Growth -1.13 -0.34 -0.60 -1.69 -0.24 -0.57 -2.00 -0.20 -0.56 Distribution 0.13 -0.66 -0.40 0.69 -0.76 -0.43 1.00 -0.80 -0.44 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank). Notes: Decomposition follows Datt and Ravallion (1992). A negative sign indicates a contribution to increasing poverty, a positive sign indicates a contribution to poverty reduction. 28 Figure 2. Decomposition of Latin American household income inequality, by share attributable to each source of income, 1995 and 2010 1.000 0.140 0.124 0.120 0.119 0.900 0.013 0.020 0.020 0.014 0.800 0.075 0.115 0.124 0.131 0.700 0.600 0.500 0.400 0.771 0.742 0.736 0.736 0.300 0.200 0.100 0.000 1995 2000 2005 2010 labor pensions transfers other Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank). Note: Calculated using total household income per capita. 29 Table 3. Labor Income Inequality Indices in Latin America, 1995-2010 Gini coefficient Theil Index Country 1995 2000 2005 2010 1995 2000 2005 2010 Argentina 0.414 0.432 0.435 0.400 0.322 0.341 0.373 0.300 Bolivia 0.535 0.567 0.548 0.525 0.570 0.670 0.606 0.611 Brazil 0.581 0.571 0.548 0.519 0.709 0.712 0.670 0.613 Chile 0.597 0.558 0.548 0.547 0.873 0.727 0.713 0.690 Colombia 0.512 0.547 0.514 0.508 0.556 0.671 0.568 0.562 Costa Rica 0.418 0.425 0.445 0.453 0.338 0.354 0.431 0.407 Dom. Rep. 0.474 0.488 0.479 0.469 0.449 0.467 0.475 0.415 Ecuador 0.461 0.517 0.472 0.449 0.411 0.572 0.447 0.409 El Salvador 0.467 0.469 0.469 0.442 0.451 0.430 0.485 0.382 Honduras 0.539 0.529 0.609 0.576 0.658 0.573 0.889 0.821 Mexico 0.538 0.534 0.507 0.484 0.629 0.606 0.560 0.485 Panama 0.470 0.491 0.492 0.472 0.414 0.469 0.467 0.451 Paraguay 0.545 0.506 0.521 0.507 0.623 0.503 0.536 0.558 Peru 0.524 0.576 0.529 0.510 0.562 0.778 0.564 0.525 Uruguay 0.438 0.434 0.469 0.459 0.374 0.371 0.421 0.430 LAC 0.547 0.546 0.524 0.500 0.638 0.649 0.601 0.546 90/10 ratio 80/20 ratio Country 1995 2000 2005 2010 1995 2000 2005 2010 Argentina 14.8 17.2 20.3 15.3 8.1 9.2 10.1 8.2 Bolivia 33.4 47.1 40.7 34.0 16.5 20.0 17.6 15.4 Brazil 39.5 39.4 34.5 28.8 19.4 17.8 15.2 12.7 Chile 36.0 29.2 26.4 24.9 17.8 14.4 13.3 12.8 Colombia 35.7 39.8 35.4 32.2 14.5 17.3 14.9 14.1 Costa Rica 16.0 16.9 17.9 17.5 8.1 8.4 9.1 9.2 Dom. Rep. 21.1 19.3 20.0 19.6 10.7 10.7 10.7 10.5 Ecuador 21.8 33.8 24.5 22.6 10.8 14.9 11.5 10.4 El Salvador 22.0 22.0 20.7 17.6 11.0 11.0 10.4 9.1 Honduras 35.3 40.4 82.4 54.9 16.5 17.6 29.3 22.1 Mexico 43.3 42.6 37.9 31.4 17.5 17.0 15.1 13.0 Panama 23.4 35.8 37.4 28.4 11.5 15.1 15.8 12.6 Paraguay 48.6 35.8 41.8 36.4 19.6 15.4 17.5 15.1 Peru 45.8 62.9 43.6 42.2 17.8 24.0 18.4 16.7 Uruguay 16.9 16.2 21.0 20.0 9.0 8.8 11.0 10.3 LAC 38.3 39.3 34.7 29.7 17.3 17.0 14.9 13.0 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. 30 Figure 3. Annualized changes in the Labor Income Gini, 1995-2010 0.6% 0.53% 0.45% 0.4% 0.32% 0.2% 0.03% 0.0% -0.2% -0.06% -0.05% -0.12% -0.17% -0.22% -0.18% -0.4% -0.37% -0.6% -0.49% -0.59% -0.58% -0.8% -0.70% -0.75% -1.0% LAC El Salvador Argentina Dominican Rep. Chile Ecuador Panama Mexico Costa Rica Brazil Paraguay Bolivia Colombia Uruguay Honduras Peru Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. 31 Figure 4. Decomposition of Labor Income (individual hourly wages) inequality changes, 1995-2010: Gini coefficient and Theil Index 0.050 0.050 Observed Quantities 0.030 0.030 Changes in the Gini from 2000 0.010 0.010 -0.010 -0.010 -0.030 -0.030 -0.050 -0.050 -0.070 -0.070 -0.090 -0.090 -0.110 -0.110 1995 2000 2005 2010 1995 2000 2005 2010 Gini Theil Gini Theil 0.050 0.050 Unobservables Prices 0.030 0.030 0.010 0.010 -0.010 -0.010 -0.030 -0.030 -0.050 -0.050 -0.070 -0.070 -0.090 -0.090 -0.110 -0.110 1995 2000 2005 2010 1995 2000 2005 2010 Gini Theil Gini Theil Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. The decomposition follows Foguel and Azevedo (2007). 32 Figure 5. Decomposition of Labor Income (individual hourly wages) inequality changes, 1995-2010: 90/10 and 80/20 labor income ratios 15.0 15.0 Observed Quantities Changes in the Gini from 2000 10.0 10.0 5.0 5.0 0.0 0.0 -5.0 -5.0 -10.0 -10.0 -15.0 -15.0 1995 2000 2005 2010 1995 2000 2005 2010 90/10 80/20 90/10 80/20 15.0 Prices 15.0 10.0 Unobservables 10.0 5.0 5.0 0.0 0.0 -5.0 -5.0 -10.0 -10.0 -15.0 1995 2000 2005 2010 -15.0 1995 2000 2005 2010 90/10 80/20 90/10 80/20 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. The decomposition follows Foguel and Azevedo (2007). 33 Table 4. Decomposition of changes in the labor income Gini Coefficient, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -0.032 -0.002 -0.027 -0.002 Bolivia -0.042 -0.011 -0.041 0.010 Brazil -0.052 -0.001 -0.032 -0.019 Chile -0.011 -0.013 -0.015 0.016 Colombia -0.039 0.004 -0.032 -0.011 Costa Rica 0.027 0.014 0.016 -0.003 Dominican Rep. -0.019 -0.017 0.001 -0.003 Ecuador -0.069 0.000 -0.016 -0.053 El Salvador -0.027 0.005 -0.019 -0.013 Honduras 0.047 -0.001 0.001 0.047 Mexico -0.050 -0.005 -0.033 -0.013 Panama -0.019 0.016 -0.031 -0.003 Paraguay 0.001 -0.002 -0.017 0.019 Peru -0.067 -0.009 -0.034 -0.023 Uruguay 0.026 -0.003 0.013 0.015 LAC -0.045 -0.003 -0.029 -0.014 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. Table 5. Decomposition of changes in the labor income Theil Index, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -0.041 -0.005 -0.043 0.007 Bolivia -0.059 -0.058 -0.082 0.082 Brazil -0.099 -0.001 -0.071 -0.026 Chile -0.038 -0.071 -0.023 0.056 Colombia -0.109 0.005 -0.077 -0.037 Costa Rica 0.053 0.024 0.032 -0.002 Dominican Rep. -0.052 -0.045 0.007 -0.014 Ecuador -0.163 -0.004 -0.039 -0.120 El Salvador -0.049 0.000 -0.032 -0.017 Honduras 0.248 0.008 0.010 0.229 Mexico -0.121 -0.001 -0.083 -0.037 Panama -0.018 0.046 -0.064 0.000 Paraguay 0.055 -0.015 -0.027 0.096 Peru -0.253 -0.093 -0.095 -0.065 Uruguay 0.058 -0.007 0.023 0.043 LAC -0.101 -0.010 -0.068 -0.023 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. 34 Table 6. Decomposition of changes in the labor income 90/10 ratio, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -1.844 0.250 -2.582 0.488 Bolivia -13.122 -1.790 -10.904 -0.428 Brazil -10.599 2.148 -8.683 -4.064 Chile -4.323 -2.142 -2.700 0.519 Colombia -7.633 0.903 -7.056 -1.480 Costa Rica 0.676 0.558 1.601 -1.484 Dominican Rep. 0.338 -1.142 0.122 1.358 Ecuador -11.174 0.478 -2.976 -8.676 El Salvador -4.410 0.879 -2.575 -2.714 Honduras 14.473 1.050 0.504 12.919 Mexico -11.197 -1.481 -7.692 -2.024 Panama -7.461 1.462 -6.321 -2.601 Paraguay 0.600 0.653 -4.254 4.200 Peru -20.745 -1.826 -12.269 -6.650 Uruguay 3.769 0.012 1.328 2.430 LAC -9.646 0.498 -7.381 -2.763 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. Table 7. Decomposition of changes in the labor income 80/20 ratio, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -0.979 0.278 -1.247 -0.009 Bolivia -4.594 -0.080 -4.496 -0.018 Brazil -5.099 -0.178 -3.272 -1.649 Chile -1.666 -0.737 -1.317 0.387 Colombia -3.216 0.439 -2.830 -0.825 Costa Rica 0.774 0.490 0.650 -0.366 Dominican Rep. -0.221 -0.566 -0.054 0.399 Ecuador -4.507 0.062 -1.181 -3.387 El Salvador -1.893 0.284 -1.202 -0.975 Honduras 4.509 0.531 0.127 3.852 Mexico -3.958 -0.538 -2.639 -0.781 Panama -2.509 0.647 -2.420 -0.737 Paraguay -0.238 0.496 -1.796 1.062 Peru -7.234 -0.804 -4.075 -2.355 Uruguay 1.512 0.040 0.602 0.869 LAC -4.074 -0.203 -2.738 -1.132 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. 35 Table 8. Typology of Changes in the Gini coefficient of Labor Income (individual hourly wages), between 2000 and 2010 Inequality-reducing Inequality-increasing Argentina Colombia Bolivia Costa Rica Brazil Ecuador Chile El Salvador Dom. Rep. Panama Quantity effect Honduras Mexico Paraguay Peru Uruguay Argentina Costa Rica Bolivia Dominican Rep. Brazil Honduras Chile Uruguay Colombia Ecuador Price effect El Salvador Mexico Panama Paraguay Peru Argentina Bolivia Brazil Uruguay Colombia Chile Costa Rica Paraguay Dominican Rep. Honduras Other Factors Ecuador El Salvador Mexico Panama Peru Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). 36 Figure 6. Mean returns to years of education, experience and other factors, 1995-2010 105.0 100.0 95.0 Index 2000=100 90.0 85.0 80.0 75.0 70.0 1995 2000 2005 2010 Education Experience Other Factors Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using population-weighted averages. The aggregation of the returns follows Foguel and Azevedo (2007). Table 9. Mean returns to education, experience and other factors, 1995-2010 (2000=100) Education Experience Unobservables 1995 2000 2005 2010 1995 2000 2005 2010 1995 2000 2005 2010 Argentina 52 100 114 72 93 100 90 66 94 100 108 102 Bolivia 146 100 108 40 216 100 194 165 93 100 98 97 Brazil 112 100 87 63 103 100 97 85 102 100 97 94 Chile 89 100 57 132 113 100 90 81 108 100 102 101 Colombia 106 100 83 76 79 100 88 80 103 100 101 99 Costa Rica 159 100 114 121 111 100 89 78 97 100 97 94 Dom. Rep. 117 100 109 89 105 100 95 95 109 100 104 104 Ecuador 138 100 130 94 127 100 109 68 87 100 91 89 El Salvador 99 100 71 69 85 100 74 81 96 100 100 91 Honduras 97 100 111 95 106 100 94 147 94 100 119 106 Mexico 85 100 102 71 104 100 82 69 104 100 100 99 Panama 76 100 104 100 107 100 101 81 90 100 103 96 Paraguay 86 100 58 25 131 100 145 98 108 100 109 102 Peru 87 100 68 69 99 100 141 108 103 100 94 96 Uruguay 136 100 118 158 106 100 96 92 103 100 106 106 LAC 101 100 92 71 103 100 96 83 102 100 99 97 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. 37 Figure 7. Mean and Standard deviation of education and experience (1995-2010) Education Experience 120 105 115 104 Index (2000=100) Index (2000=100) 110 103 105 102 101 100 100 95 99 90 98 1995 2000 2005 2010 1995 2000 2005 2010 Standard Deviation Mean Standard Deviation Mean Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. Table 10. Mean education and experience, 1995-2010 (2000=100) Mean Years of Education Mean Years of Experience 1995 2000 2005 2010 1995 2000 2005 2010 Argentina 96 100 106 110 98 100 100 100 Bolivia 98 100 99 117 101 100 102 99 Brazil 88 100 109 119 99 100 100 100 Chile 101 100 108 111 94 100 103 104 Colombia 88 100 101 107 98 100 102 103 Costa Rica 97 100 107 114 96 100 103 104 Dom. Rep. 102 100 103 98 99 100 102 109 Ecuador 122 100 103 108 89 100 103 106 El Salvador 93 100 108 109 99 100 99 100 Honduras 100 100 104 105 100 100 104 106 Mexico 93 100 110 117 100 100 102 102 Panama 105 100 104 109 93 100 102 104 Paraguay 96 100 111 119 96 100 97 97 Peru 87 100 98 105 99 100 104 104 Uruguay 97 100 108 105 99 100 102 102 LAC 91 100 107 114 99 100 101 102 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. 38 Table 11. Standard deviation of education and experience, 1995-2010 (2000=100) Standard Deviation of Education Standard Deviation of Experience 1995 2000 2005 2010 1995 2000 2005 2010 Argentina 98 100 99 95 99 100 103 103 Bolivia 103 100 100 98 100 100 102 102 Brazil 99 100 99 98 100 100 102 102 Chile 101 100 91 89 101 100 102 103 Colombia 99 100 103 104 99 100 101 103 Costa Rica 97 100 100 103 101 100 102 104 Dom. Rep. 99 100 97 98 101 100 99 101 Ecuador 96 100 101 101 98 100 100 101 El Salvador 99 100 100 98 101 100 99 99 Honduras 99 100 103 101 99 100 99 100 Mexico 99 100 100 97 100 100 99 99 Panama 97 100 100 99 100 100 102 104 Paraguay 98 100 105 104 99 100 102 104 Peru 94 100 101 100 101 100 101 103 Uruguay 100 100 102 100 102 100 100 100 LAC 99 100 100 98 100 100 101 102 Source: Authors’ calculations with data from SEDLAC (CEDLAS and The World Bank). LAC values refer to labor population weighted averages. Figure 8. Decomposition of Variance of Annual Earnings 1.4 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 1995 2000 2005 2010 Var hours worked Var hourly wage Covarience Hours-Wage 39 ANNEX Table A1. The circa criteria Country Circa 1995 Circa 2000 Circa 2005 Circa 2010 Argentina 1995 2000 2005 2010 Bolivia 1997 2000 2005 2008 Brazil 1995 2001 2005 2009 Chile 1996 2000 2006 2009 Colombia 1996 2002 2005 2010 Costa Rica 1995 2000 2005 2009 Dominican Rep. 1996 2000 2005 2010 Ecuador 1995 2003 2006 2010 El Salvador 1995 2000 2005 2009 Honduras 1995 1999 2005 2009 Mexico 1996 2000 2005 2010 Panama 1995 2001 2005 2009 Paraguay 1995 1999 2005 2010 Peru 1997 2002 2005 2010 Uruguay 1995 2000 2005 2010 Table A2. Surveys in the Sample Country Circa 1995 Circa 2000 Circa 2005 Circa 2010 Argentina EPH EPH EPH-C EPH-C Bolivia ENE ECH ECH ECH Brazil PNAD PNAD PNAD PNAD Chile CASEN CASEN CASEN CASEN Colombia ENH-FT ECH ECH GEIH Costa Rica EHPM EHPM EHPM EHPM Dominican R. ENFT ENFT ENFT ENFT Ecuador ECV ENEMDU ENEMDU ENEMDU El Salvador EHPM EHPM EHPM EHPM Honduras EPHPM EPHPM EPHPM EPHPM Mexico ENIGH ENIGH ENIGH ENIGH Panama EH EH EH EH Paraguay EH EIH EPH EPH Peru ENAHO ENAHO ENAHO ENAHO Uruguay ECH ECH ECH ECH 40 Figure A1. Gini coefficient and Theil index of total household income per capita in LAC, 1995-2010 0.75 0.70 0.67 0.66 Gini and Theil coefficients 0.65 0.61 0.60 0.57 0.57 0.54 0.55 0.55 0.52 0.50 0.45 0.40 1995 2000 2005 2010 Theil Gini Figure A2. Shares of income sources in total household income per capita in LAC, 1995-2010 1.000 0.128 0.112 0.113 0.108 0.900 0.026 0.036 0.017 0.029 0.800 0.076 0.106 0.113 0.116 0.700 0.600 0.500 0.400 0.780 0.756 0.745 0.740 0.300 0.200 0.100 0.000 1995 2000 2005 2010 labor pensions transfers other Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank). 41 Figure A3. Marginal effect on inequality by income source, 1995-2010 2.0% 1.4% 1.5% 1.3% 1.1% 1.2% 1.1% 1.0% 0.9% 0.7% 0.5% 0.0% 0.0% -0.5% -0.3% -0.4% -0.6% -1.0% -0.9% -0.9% -1.0% -1.5% -1.4% -2.0% -2.5% -2.2% Labor Pensions Transfers Other 1995 2000 2005 2010 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank). 42 Table A3. Female: Decomposition of changes in the labor income Gini Coefficient, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -0.006 -0.003 -0.007 0.004 Bolivia 0.002 0.023 -0.054 0.032 Brazil -0.044 0.001 -0.027 -0.017 Chile 0.011 0.004 -0.010 0.017 Colombia -0.027 -0.004 -0.022 0.000 Costa Rica 0.012 0.003 0.020 -0.011 Dominican Rep. 0.003 -0.019 0.022 0.001 Ecuador -0.074 -0.010 -0.014 -0.049 El Salvador -0.022 0.012 -0.018 -0.016 Honduras 0.031 0.011 -0.032 0.052 Mexico -0.017 0.009 -0.025 -0.002 Panama 0.015 0.016 -0.028 0.026 Paraguay 0.067 0.004 -0.009 0.073 Peru -0.036 0.003 -0.022 -0.016 Uruguay 0.020 0.002 0.016 0.002 LAC -0.028 0.002 -0.022 -0.008 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. Table A4. Male: Decomposition of changes in the labor income Gini Coefficient, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -0.050 0.000 -0.042 -0.008 Bolivia -0.070 -0.012 -0.045 -0.013 Brazil -0.054 0.000 -0.035 -0.019 Chile -0.018 -0.016 -0.018 0.016 Colombia -0.044 0.009 -0.034 -0.019 Costa Rica 0.033 0.018 0.017 -0.003 Dominican Rep. -0.030 -0.020 -0.004 -0.006 Ecuador -0.070 -0.002 -0.015 -0.053 El Salvador -0.030 -0.001 -0.018 -0.012 Honduras 0.057 -0.003 0.012 0.048 Mexico -0.059 -0.001 -0.040 -0.018 Panama -0.032 0.013 -0.029 -0.016 Paraguay -0.024 -0.007 -0.026 0.008 Peru -0.086 -0.012 -0.045 -0.029 Uruguay 0.031 -0.006 0.015 0.021 LAC -0.052 -0.001 -0.034 -0.017 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. 43 Table A5. Formal: Decomposition of changes in the labor income Gini Coefficient, 2000-2010 Gini Coefficient Country Observed Quantities Prices Unobservables Argentina -0.049 -0.008 -0.038 -0.002 Bolivia -0.044 -0.029 -0.034 0.019 Brazil -0.042 -0.005 -0.022 -0.016 Chile -0.014 -0.011 -0.010 0.007 Costa Rica 0.026 0.011 0.018 -0.002 Dominican Rep. -0.031 -0.029 -0.003 0.000 Ecuador -0.068 0.003 -0.021 -0.050 El Salvador -0.021 -0.003 -0.009 -0.009 Honduras 0.036 -0.006 -0.015 0.058 Mexico -0.032 -0.014 -0.026 0.008 Panama -0.003 0.016 -0.026 0.008 Paraguay -0.007 -0.017 -0.007 0.017 Peru -0.073 -0.019 -0.043 -0.011 Uruguay 0.002 0.004 0.021 -0.023 LAC -0.039 -0.009 -0.024 -0.006 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. Table A6. Informal: Decomposition of changes in the labor income Gini Coefficient, 2000-2010 Country Observed Quantities Prices Unobservables Argentina -0.009 0.001 -0.012 0.002 Bolivia -0.005 0.003 -0.008 0.000 Brazil -0.057 0.007 -0.036 -0.028 Chile 0.034 -0.004 -0.007 0.045 Costa Rica 0.029 0.012 0.002 0.016 Dominican Rep. -0.024 -0.010 -0.004 -0.011 Ecuador -0.099 -0.009 -0.016 -0.074 El Salvador -0.037 0.003 -0.010 -0.031 Honduras 0.028 -0.001 0.006 0.023 Mexico -0.047 0.011 -0.029 -0.029 Panama -0.002 0.016 -0.015 -0.004 Paraguay 0.059 0.005 -0.004 0.059 Peru -0.049 -0.004 -0.006 -0.039 Uruguay 0.043 -0.021 -0.007 0.070 LAC -0.044 0.005 -0.025 -0.024 Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. 44 Figure A4. Mean returns to education, experience and others, 1995-2010, by gender and sector 110 Female 110 Male 105 105 100 100 Index 2000=100 Index 2000=100 95 95 90 90 85 85 80 80 75 75 70 70 65 65 1995 2000 2005 2010 1995 2000 2005 2010 Education Experience Other Factors Education Experience Other Factors 110 Formal 110 Informal 105 105 100 100 Index 2000=100 Index 2000=100 95 95 90 90 85 85 80 80 75 75 70 70 65 65 1995 2000 2005 2010 1995 2000 2005 2010 Education Experience Other Factors Education Experience Other Factors Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. 45 Figure A5. Mean and Standard deviation of education (1995-2010), by sector and gender 125 Female 125 Male 120 120 115 115 Index (2000=100) Index (2000=100) 110 110 105 105 100 100 95 95 90 90 85 85 1995 2000 2005 2010 1995 2000 2005 2010 Standard Deviation Mean Standard Deviation Mean 125 Formal 125 Informal 120 120 115 Index (2000=100) 115 Index (2000=100) 110 110 105 105 100 100 95 95 90 90 85 85 1995 2000 2005 2010 1995 2000 2005 2010 Standard Deviation Mean Standard Deviation Mean Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. 46 Figure A6. Mean and Standard deviation of experience (1995-2010), by sector and gender 105 105 Female Male 104 104 103 103 Index (2000=100) Index (2000=100) 102 102 101 101 100 100 99 99 98 98 97 97 1995 2000 2005 2010 1995 2000 2005 2010 Standard Deviation Mean Standard Deviation Mean 105 105 Formal Informal 104 104 103 103 Index (2000=100) Index (2000=100) 102 102 101 101 100 100 99 99 98 98 97 97 1995 2000 2005 2010 1995 2000 2005 2010 Standard Deviation Mean Standard Deviation Mean Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) using labor population-weighted averages. Figure A7. Growth in number of workers in LAC (2000-2010), percent 140% 120% 100% 80% 60% 40% 20% 0% 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 -20% Males Females Source: Author’s calculations with data from SEDLAC (CEDLAS and The World Bank) for workers 15 and older. 47