Presentation prepared for the event: Inequality in a Lower Growth Latin America Monday, January 26, 2015 Woodrow Wilson International Center for Scholars Washington, D.C.
Inequality in LAC: Explaining the Recent Past to Shed Light on the Future Chief Economist Office Latin America and the Caribbean World Bank Augusto de la Torre Based on ongoing research by a team including Guillermo Beylis, Julián Messina, Eduardo Levy Yeyati, and Carlos Rodríguez-Castelán Wilson Center, Washington, DC January 2015
The great EM deceleration is affecting particularly intense in the case of Latin America Source: WDI, WEO (October 2014) and Consensus Forecasts (December 2014) 2
How big of a threat is it to social progress and, in particular, to the declining trend in inequality? Poverty and Inequality in LAC 60% 0.70 50% 0.65 40% 0.60 30% 0.55 20% 10% 0.50 0% LAC Andean Region South Cone extended Mexico and CA 0.45 Poverty 2003 Poverty 2008 Poverty 2010 Poverty 2012 2014 f Gini (RHS) 3
Income inequality reduction in LAC is rather unique but Household Income Inequality Across Regions Sources: Luxemburg Income Study, SEDLAC, PovCalNet. 4
does it reflect measurement limitations? Comparability: LAC uses income surveys, other regions use consumption Top earners have a higher than average non-response rate Capital income is hardly captured By being a measure of nominal income inequality, the Gini assumes that the consumption basket is the same for all income groups 5
What do we do about the measurement issue? Complement household surveys with tax-based records to capture top earners and capital income Analyze the shares of GDP going to labor (after adjusting for the income of self-employed workers) as a robustness test Estimate the inequality of purchasing power by correcting for decilespecific inflation rates Bottom line: the LAC inequality reduction story holds, but with some caveats 6
Adding the top earners changes the level but not the evolution of income inequality in non-crisis periods Colombia: Basic and Augmented Gini Index 7
The share of labor in GDP shows only part of the picture We check for consistency between the evolution of the labor share in national income and that of the survey-based Gini The expectation is that, since capital income is concentrated at the top, as the labor share declines inequality should increase. We find that the decline in labor shares (which in LAC was milder than elsewhere) was associated with a rise in income at the top 1% but that there is no systematic association with changes in the Gini Labor shares don t tell us how labor income is distributed In sum, changes in labor shares are a good proxy for changes of income at the top but not for changes in the distribution elsewhere 8
Nominal vs. purchasing power inequality The consumption basket of the rich and the poor are very different Inflation rates across products are different e.g., food prices The basket used to calculate the CPI is biased toward the rich Consumption Basket Across the Expenditure Distribution: Brazil - 2009 Decile 2 Decile 9 Plutocratic 31% 14% 16% 16% 9% 28% 26% 12% 11% 9
a similar movie with less action Change in Nominal and Deflated Gini for Selected LAC Countries circa 2001-2012 Note: El Salvador 2000 to 2012, Colombia 2001 to 2012, Mexico 2002 to 2012, Brazil and Chile: 2001 to 2011, Ecuador 2006 to 2012, and Nicaragua 2005 to 2009. 10
What are the drivers of the decline in household income inequality in the region? Big shift in social policy is a natural candidate Expansion of CCT programs Expansion of pension coverage, especially via non-contributory systems but its impact was mainly on poverty reduction It does not seem to have been the main driver of household income inequality reduction where the decrease in labor income inequality seems to have played the biggest role 11
The fall in labor income inequality was more important than transfers for income inequality in most of LAC Decomposition of the Evolution of Household Income Inequality circa 2001-2012 2 Nicaragua El Salvador Dominican RepublicParaguay Argentina urban Bolivia Ecua Uruguay Chile urban Brazil Peru Guatemala Colombia Panama Honduras Mexico 0-2 -1.6-1.3-1.1-0.9 Cha -4-6 -6.4-5.6-5.4-5.3-4.7-3.9-3.4-2.5-8 -8.4-8.4-7.7-7.5 2003-2012 2007-2012 2003-2011 2004-2012 2006-2011 2004-2012 2003-2011 2003-2012 2003-2012 2005-2009 2004-2012 2003-2012 2003-2012 2004-2012 2004-2012 2003 Non-labor income Labor income Share of occupied Source: Cord, L., O. Barriga, L. Lucchetti, C. Rodriguez-Castelan, L. Sousa, D. Valderrama (2014) Inequality Stagnation in Latin America in the Aftermath of the Global Financial Crisis. World Bank Unpublished Document. 12
The well-tuned trio: co-movement between household income, labor income, and returns to education Labor and Total Income Gini and the Education Wage Gap (LAC Average) 13
is a little less well-tuned on a country by country basis Composition-Neutral Earning Premiums and Gini Coefficient of Labor Earnings Argentina Brazil Chile 1.7 60 1.7 60 1.7 60 Skill Premium - Tertiary to Primary 1.5 1.3 1.1.9.7 55 50 45 40 Labor Gini Skill Premium - Tertiary to Primary 1.5 1.3 1.1.9.7 55 50 45 40 Labor Gini Skill Premium - Tertiary to Primary 1.5 1.3 1.1.9.7 55 50 45 40 Labor Gini 1995 1997 1999 2001 2003 2005 2007 2009 2011 1995 1997 1999 2001 2003 2005 2007 2009 2011 1995 1997 1999 2001 2003 2005 2007 2009 2011 Tertiary to Primary Labor Gini Tertiary to Primary Labor Gini Tertiary to Primary Labor Gini Prediction with average age of cell Prediction with average age of cell Prediction with average age of cell Colombia Mexico Peru 1.7 60 1.7 60 1.7 60 Skill Premium - Tertiary to Primary 1.5 1.3 1.1.9 55 50 45 Labor Gini Skill Premium - Tertiary to Primary 1.5 1.3 1.1.9 55 50 45 Labor Gini Skill Premium - Tertiary to Primary 1.5 1.3 1.1.9 55 50 45 Labor Gini.7 40.7 40.7 40 1995 1997 1999 2001 2003 2005 2007 2009 2011 1995 1997 1999 2001 2003 2005 2007 2009 2011 1995 1997 1999 2001 2003 2005 2007 2009 2011 Tertiary to Primary Labor Gini Tertiary to Primary Labor Gini Tertiary to Primary Labor Gini Prediction with average age of cell Prediction with average age of cell Prediction with average age of cell 14
What if we focus on the falling returns to education? For starters, a common factor seems difficult to discard, given that the (trio) phenomenon was widespread across the region H1: increase in the supply of education (quantity) H2: decrease in the quality of education, especially tertiary (Lustig) Lower average quality of educational institutions ( garage universities ) Lower average readiness of students that enter tertiary education H3: a demand side story (congenial to the common factor story) By promoting growth of nontradable sectors, commodity boom increased demand of unskilled relative to skilled workers 15
H1: a dead end road if anything, the completion of tertiary education decelerated in the 2000s Education Coverage in LAC 16
H2: Accelerated entry into school and rise in educational attainment for lower income groups Average Years of Completed Education by Household Income Decile Did returns to tertiary education fall because of less prepared students at entry or because lower quality of the new tertiary education programs provided? 17
H2: A case study Colombia Average test score before entering tertiary education has been falling Most of the new entrants access education through new programs 75 70 mean 8000 Programs All year of entry 65 60 6000 4000 2000 55 0 1998 2001 2004 2007 2010 2013 year_primiparo Existing New IES New program in existing IES 2001 2003 2005 2007 2009 2011 year New 'branch' of an existing program New program in existing IES New IES Existing Graphs by all Controlling for student socio-economic characteristics and area of study, the effect of new programs weakens Source: Camacho, A; Messina, J and Uribe, J.P. (2014) Heterogeneous returns to higher education: Bad programs or bad students? World Bank unpublished paper. 18
H2: Colombia controlling for student characteristics and area of study, the effect of new programs weakens (1) (2) (3) (4) (5) (6) (7) (8) New program in new institution -0.2790 *** [0.008590] -0.2735 *** [0.008592] -0.1495 *** [0.008546] -0.1078 *** [0.009320] -0.04406 *** [0.009056] -0.04762 *** [0.009105] -0.02952 ** [0.009616] -0.03584 *** [0.01025] New programs in an existing institution -0.1439 *** [0.004358] -0.1443 *** [0.004347] -0.08048 *** [0.004187] -0.05886 *** [0.004314] -0.03472 *** [0.004189] -0.04015 *** [0.004215] -0.03816 *** [0.004243] -0.03437 *** [0.004627] Constant 14.022 *** [0.003019] 8.1593 *** [0.4102] 10.022 *** [0.4027] 12.333 *** [0.3965] 12.748 *** [0.3706] 12.797 *** [0.3703] 12.828 *** [0.3685] 12.759 *** [0.4037] Year FE No Yes Yes Yes Yes Yes Yes Yes Age gender FE No Yes Yes Yes Yes Yes Yes Yes Test Score in HS No No Yes Yes Yes Yes Yes Yes School FE No No No Yes Yes Yes Yes Yes Area No No No No Yes Yes Yes Yes Level of education No No No No Yes Yes Yes Yes Institution No No No No No Yes Yes Yes characteristics Departamanent FE No No No No No No Yes Yes Familly Income& No No No No No No No Yes Parents education Observations 127073 127073 127073 127073 127073 127073 127073 107449 Adjusted R 2 0.021 0.043 0.121 0.187 0.267 0.269 0.271 0.273 Source: Camacho, A; Messina, J and Uribe, J.P. (2014) Heterogeneous returns to higher education: Bad programs or bad students? World Bank unpublished paper. 19
H3: low demand for skills? Mixed results, so far - Sectors that grew the most include services, which have more educated workers and pay higher wages, on average - The notion that services are low skilled is questionable there is heterogeneity - but does this reflects a rise in skill-job mismatches? - It would be better to look at the occupation content of sectors 1.8 Skill Premium and TOT Change in Share of Employment and Initial Education Intensity 1.6 1.4 Wage Premium Skilled-Unskilled 1.2 1 0.8 y = -0.0025x + 1.3035 0.6 0.4 50 70 90 110 130 150 170 190 TOT (2000=100) Note: On the left, the scatter represent a pooling of countries from the LAC region; on the right, each dot represents a sector in a country of LAC. Each dot in both panels corresponds to a year in 2000-2010. 20
H3: but when we look at skill content of occupations the story is highly nuanced Change in Employment Share by Skill Decile Panel A. Brazil Panel B. Chile Panel C. Mexico Panel D. United States Source: Fernandez-Sierra, Manuel (2014) Employment Polarization and Wage Inequality in Latin America: The Cases of Brazil, Chile and Mexico 21
So, what was behind the fall in labor earnings inequality? The fall in the skill premium, proxied by a reduction in the returns to education, appears as a primary candidate But other gaps were closing e.g., reductions in gender (or ethnic) pay differentials (Ñopo, 2012) For a given skill premium, changes in the composition of the labor force may have affected income inequality Older and better educated workers are better paid, except under fast skill obsolescence Controlling for skills, female workers tend to receive lower wages than men, especially after the marriage penalty Changes in the age and education structure of the labor force, and in female labor force participation can mechanically change wage inequality in either direction Institutional changes e.g., increases in the minimum wage 22
Thank you! 23
Supply factors? Changes in the composition of the labor force are not a good explanation 35% Education, Female LFP and Wage Inequality Female LFP Secondary Tertiary Labor Income Gini (rhs) 0.52 in % of Working Age Population 30% 25% 20% 15% 10% 5% 0.51 0.50 0.49 0.48 0.47 0.46 0% 0.45 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Gini Index Source: LCRCE from SEDLAC. 24
Supply factors? There seems to be evidence in favor of the degradation of tertiary education hypothesis 1.6 1.58 Within Goup Wage Inequality by Education in LAC 1990s 2000s 1.56 1.54 1.52 1.5 1.48 1.46 1.44 1.42 1.4 Secondary Tertiary Source: Fernández-Sierra & Messina (2012). 25