WPS6015 Policy Research Working Paper 6015 Do Middle Classes Bring Institutional Reforms? Norman Loayza Jamele Rigolini Gonzalo Llorente The World Bank Latin America and the Caribbean Region Office of the Chief Economist & Development Research Group Macroeconomics and Growth Team March 2012 Policy Research Working Paper 6015 Abstract The paper examines the link between poverty, the middle corruption) also improves. This trend does not occur class and institutional outcomes using a new cross- at the expense of economic freedom, as a larger middle country panel dataset on the distribution of income class also leads to more market-oriented economic policy and expenditure. It uses an econometric methodology on trade and finance. These beneficial effects of a larger to gauge whether a larger middle class has a causal middle class appear to be more robust than the impact of effect on policy and institutional outcomes in three lower poverty, lower inequality or higher gross domestic areas: social policy in health and education, market- product per capita. That may be linked to the evolution oriented economic structure and quality of governance. of the middle class: they are more enlightened, more The analysis finds that when the middle class becomes likely to take political actions and have a stronger voice. larger (measured as the proportion of people earning They also share preferences and values for policy and more than US$10 a day), social policy on health and institutional reforms, as well as higher stakes in property education becomes more progressive, and the quality rights and wealth accumulation. of governance (democratic participation and official This paper is a product of the Office of the Chief Economist, Latin America and the Caribbean Region; and Macroeconomics and Growth Team, Development Research Group. 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 authors may be contacted at jrigolini@worldbank. org and nloayza@worldbank.org, respectively. 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 Do Middle Classes Bring Institutional Reforms? Norman Loayza Jamele Rigolini Gonzalo Llorente 1 The World Bank The World Bank and IZA The World Bank JEL: D3, H5, O1, O4 Keywords: Poverty, Middle Class, Income, Institutions, Development 1 The authors gratefully acknowledge financial support from the Spanish Fund for Latin America and the Caribbean (SFLAC). We are also grateful to Nancy Birdsall, Francisco Ferreira, Luis Servén and participants to the Barcelona workshop on Socio-Economic Mobility and Middle Class in Latin America for insightful comments and suggestions. Tomoko Wada provided excellent research assistance. 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. 1. Introduction How the middle class affects institutions and the social contract has been subject to widespread attention. Numerous empirical analyses have demonstrated that lower inequality and a larger “class in the middleâ€? leads to better institutional outcomes (Barro, 1999; Easterly, 2001; Easterly et al., 2006).2 Yet, middle class status is as much defined by people’s relative position on the distribution of income as by occupational status and absolute income (Goldthorpe and McKnight, 2004; Lopez- Calva and Ortiz-Juarez, 2011), and many of the positive impacts of middle class societies depend on people exiting poverty and earning higher incomes. Theoretical models postulate a positive impact of wealthier middle class societies on economic and social outcomes through several channels. A first channel stresses the impact of middle class endowments, preferences and values on economic growth. The seminal studies of Banerjee and Newmann (1993) and Galor and Zeira (1993) suggest, for instance, that members of the middle class are less vulnerable to the credit market imperfections and fixed costs in physical and human capital accumulation that prevent the poor from investing and growing.3 Another strand of literature highlights the importance of domestic markets for industrialization and the higher demand of the middle class for quality goods (Murphy, Shleifer and Vishny, 1989a and 1989b; Matsuyama, 1992 and 2002; Foellmi and Zweimueller, 2006). Doepke and Zilibotti (2005, 2008) also argue that middle class families are in occupations that require skills and experience, thus they develop work ethics and patience. A related vein of literature also explores the link between wealthier middle classes and institutional outcomes. Particular attention has been given to the “modernization theoryâ€? (Lipset, 1959), which looks at the extent to which more affluent societies favor the creation and consolidation of democracies (and, more generally, good institutions). Conceptually, higher incomes may reduce conflict over its distribution (Benhabib and Rustichini, 1996; Benhabib and Przeworski, 2006), and citizens with higher human capital may be more effective in sustaining good institutions (Glaeser et al., 2004). While data do show a clear correlation, there is a debate on the extent to which the relationship can be interpreted causally (Benhabib et al., 2011; Acemoglu et al., 2008 and 2009; Epstein et al., 2006; Glaeser et al., 2004). 2 There is a rich related literature on the institutional and growth effects of inequality. See, for instance, Alesina and Rodrik (1992), Alesina and Perotti (1996), Keefer and Knack (1994), Perotti (1996), and Persson and Tabellini (1994). 3 These works were followed by a number of related studies. See, among many others, Banerjee and Duflo (2003), Galor and Moav (2004), Voitchovsky (2005), and Foellmi and Oechslin (2008). 2 Despite the relevance and strong interest in the issue, analyses of the impacts of wealthier middle classes face constraints in measuring income directly, and most use GDP per capita as a proxy for income. Only a handful of cross country datasets report headcount indexes for income thresholds other than poverty, and the ones that exist span too short time periods to exploit both cross country and time series variations. As a result, current analyses fail to investigate directly whether, as postulated by the literature, “critical massesâ€? of people overcoming certain income thresholds can affect institutional outcomes. This paper makes two contributions. First, it presents a new panel dataset that contains information about households’ mean income and expenditures, inequality, and headcount indexes for several income and expenditures thresholds, which complements existing large cross-country datasets reporting poverty and inequality measures. The dataset spans 672 yearly observations across 128 countries. To build the dataset, we draw both from a collection of nationally representative household surveys, and from parameterized distributions of income and expenditures using parameters from the World Bank’s PovCal database. Second, we use this new dataset to test the extent to which policy and institutional outcomes are affected by an expansion of the middle class, which in the paper is measured as the proportion of the population who have achieved income above 10 US dollars a day. The joint endogeneity of the size of the middle class (and of the income distribution in general) is a concern. So is the likely presence of unobserved country-specific effects. We use an econometric methodology that, by taking advantage of the panel nature of the data, attempts to gauge the causal effect of expansions in the middle class while controlling for country-specific effects. Specifically, the estimation draws from the generalized method of moments (GMM) estimator for panel data developed by Arellano and Bond (1991) and Arellano and Bover (1995). We estimate the impact of the middle class on policies and institutions in three broad categories: social policy regarding public expenditures on health and education, market-oriented economic structure on international trade and finance, and quality of governance regarding democratic participation and absence of official corruption. Our analysis thus also relates to the literature investigating the association between poverty, institutions and growth (Dollar and Kraay, 2002; Lopez and Serven, 2009). We find that when the size of the middle class increases, social policy on health and education becomes more active, and the quality of governance regarding democratic participation and official corruption improves. This does not occur at the expense of economic freedom, as an expansion of 3 the middle class also implies more market-oriented economic policy on trade and finance. The impact of a larger middle class appears to be more robust than the impact on the same outcomes of lower poverty, lower inequality, and also of higher GDP per capita. Overall, our findings suggest that the development of the literature studying the association between income and socioeconomic factors may have been hindered by its limitation in measuring household income directly. The paper is organized as follows. Section 2 discusses the dataset; Section 3 presents the empirical approach; Section 4 discusses the results, and Section 5 concludes. 2. Data Description The analysis is based on a cross-country panel dataset that contains information about headcount indexes for various thresholds that we have purposely built for the analysis. The dataset spans 672 yearly observations across 128 countries, from 1967 to 2009 (around 90 percent of the observations are however from the 1990s and 2000s). To compute the headcount indexes we draw from various World Bank collections of harmonized nationally representative household surveys that contain information on income or expenditures, and from simulated distributions of income and expenditures from the World Bank’s PovCal database, whose parameters fit the distribution of nationally representative household surveys (see Table 1). Because of the nature of the primary data, 17 percent of the countries and 38 percent of the annual observations are from Latin America. The dataset is fairly balanced across levels of economic development: 21 percent of the observations are from high income countries, 37 percent from upper middle income, 30 percent from lower middle income, and 11 percent from low income countries. Because surveys tend to report information either for income or expenditures, we report for each country only one of the two measures: 57 percent of the sample reports information on income, and 43 percent on expenditures. All income and expenditures data are in 2005 PPP US dollars. For each survey, we first correct current units for inflation using the national CPIs, and then convert them into 2005 US dollars PPP using the International Comparison Program (ICP) PPP conversion. Since the ECAPOV, PovCal and SEDLAC surveys are used to compute World Bank poverty figures, we used for these surveys the same conversion, weights and methodology that has been used to compute internationally comparable poverty data. For the analysis in this paper, we have collapsed yearly observations into five-year averages. We have also dropped from the analysis countries with population of less than two million, and have 4 excluded Yemen because of an abnormal association between mean income and GDP per capita suggesting that data quality may be an issue. We are left with 343 observations over 110 countries. By taking five-year averages, the proportion of observations from Latin America also falls to 24 percent. We use the dataset to capture, in each country, the proportion of people living in poverty (below 2.5 US dollars a day in per capita PPP terms), the percentage of the population that lives with more than 10 US dollars a day (our “middle class thresholdâ€?), and overall income inequality as measured by the Gini coefficient. In using 10 dollars a day as our middle class threshold, we follow analyses that have looked at vulnerability to poverty as a prerequisite to middle class status. These analyses find that if the middle class threshold is set excessively close to the poverty threshold, people remain excessively vulnerable to shocks that would bring them back into poverty, and may not behave differently than the poor (see Goldthorpe and McKnight, 2004; and Lopez-Calva and Ortiz-Juarez, 2011). Using a 10 dollars a day threshold also brings us close to papers investigating middle class composition and trends at a global level (Milanovic and Yitzhaki, 2002; Kharas, 2010). Observe, also, that we refrain from using an income ceiling to limit the size of the middle class from above. As we pool together countries from all income levels, an income ceiling may lead to strange measurements by which the middle class in rich countries may be artificially too small. Our results should, correspondingly, be interpreted as the impacts of a growing proportion of people with sufficient income to undertake activities beyond constantly fighting poverty, instead of the impacts of a class “in the middle.â€? Table 2 shows the correlation between income distribution variables: while poverty and the middle class are negatively correlated, they do so imperfectly, suggesting that each measure carries some information. Other variables that we use in the analysis are GDP per capita in 2005 PPP terms, public expenditures in education and health, and mean applied tariffs, all from the World Development Indicators; credit market liberalization, from the Economic Freedom of the World project; the polity score for democracy, from Polity IV; and the corruption index from the International Country Risk Guide. 3. Econometric Methodology We estimate a set of regression equations of the form, ï?² y i ,t  ï?¢ ' X i ,t  ï?¨ i  ï?­ t  ï?¥ i ,t (1) 5 where represents a given policy outcome, is a set of explanatory variables that include the size of the middle class, is an unobserved country-specific effect, is an unobserved time-specific effect, is the regression residual, and the subscripts i and t represent country and time period, respectively. The regression equation poses two main challenges for estimation. The first is that the explanatory variables are likely to be jointly endogenous with the policy outcomes; that is, may not be zero. The second challenge is the presence of unobserved time and country-specific effects, which may be correlated with the explanatory variables. To address these challenges, the estimation draws from the generalized method of moments (GMM) estimator for panel data developed by Arellano and Bond (1991) and Arellano and Bover (1995). We deal with unobserved time effects through the inclusion of period-specific intercepts. Dealing with unobserved country effects is not as simple given the possibility that the regression equation contains endogenous explanatory variables. The GMM estimator takes advantage of the panel nature of the data set in dealing with country-specific effects and endogenous explanatory variables. Unobserved country-specific effects are controlled for by differencing the regression equation and using instrumental variables based on previous observations of the explanatory variables. Differencing the regression equation also controls for potential level effects caused by some countries reporting income data, while others report expenditures. The method relies on similar instrumental variables to control for joint endogeneity. Specifically, the GMM estimator uses jointly the regression in levels and in differences. For the regression in levels, it uses as instruments the previous differences of the explanatory variables. For the regression in differences, it uses as instruments the previous levels of the explanatory variables. These are appropriate instruments under the following assumptions: (1) future realizations of the error term do not affect current values of the explanatory variables, (2) the error term is serially uncorrelated, and (3) changes in the explanatory variables are uncorrelated with the unobserved country-specific effect. As Arellano and Bond (1991) and Arellano and Bover (1995) show, this set of assumptions generates moment conditions that allow consistent estimation of the parameters of interest. For ï?›ï€¨  ï?? the regression in levels, these assumptions imply E X i ,t 1  X i ,t 2  ï?¨ i  ï?¥ i ,t  0 . For the ï?›  ï?? regression in differences, they imply E X i ,t 2  ï?¥ i ,t  ï?¥ i ,t 1  0 . Since typically the moment conditions over-identify the regression model, a Hansen-type test can be used to check the validity of the moment conditions and their underlying assumptions. 6 4. Empirical Results We begin the analysis by looking at the simple relation between economic development and institutions (Table 3). A substantial body of evidence finds that policies and institutions have a strong influence on economic development (see Barro, 1991; Easterly and Levine, 2001; and Chang, Kaltani and Loayza, 2009, among others). This is not, however, a unidirectional causal relationship but a symbiotic one, by which economic development also shapes the evolution of social policies and government institutions (Krueger, 1995). We follow most the literature and represent a country’s economic development by its GDP per capita (Acemoglu, Johnson, and Robinson, 2001; Easterly and Levine; 2003; and Caselli, 2005), and divide policies and institutions in three broad categories: social policy regarding public expenditures in health and education, market-oriented economic structure on international trade and finance, and quality of governance regarding democratic participation and absence of official corruption. A clear result emerges: GDP per capita significantly and beneficially affects the indicators of social policy, economic structure, and governance. Specifically, an increase in GDP per capita induces a rise in public health and education expenditures, a reduction in tariff rates on international trade, a liberalization of credit markets, an improvement in democratic participation, and a reduction in official corruption. To the extent that other aspects of the income distribution are also germane in shaping policies and institutions, it is important to consider them explicitly for both statistical and conceptual reasons. From a statistical perspective, it is unlikely that a regression model with only GDP per capita as explanatory variable be well specified. In fact, in the context of the GMM estimator, the Hansen test rejects the moment conditions in three of the six regressions. On conceptual considerations, we would like to know what aspect of the income distribution is most relevant: is it average output, income inequality, the prevalence of poverty, or, more specifically, the size of the middle class? As expected, GDP per capita is positively correlated with the size of the middle class and negatively so with the poverty headcount index and the Gini coefficient of income inequality (Table 2). A priori it is not clear however which of these variables would be most relevant in shaping different policies and institutions. In particular, how important is the size of the middle class in this regard? 7 Table 4 presents the results of the set of regressions on the indicators of social policy, economic structure, and governance, considering as explanatory variables not only GDP per capita but also measures of poverty, inequality, and the middle class. For this augmented model, the Hansen test does not reject the regression specification (i.e., moment conditions) for any of the regressions. When controlling for the size of the middle class, the coefficients corresponding to GDP per capita lose their significance, size, or even sign, depending on the regression. At the same time, the size of the middle class appears to now carry the coefficients’ sign and significance that GDP per capita used to have when it was the only explanatory variable. It is plausible therefore that the beneficial effect that had been attributed to changes in GDP per capita actually corresponds to the evolution of the middle class. An expansion of the middle class has a significant impact on social policy by inducing an increase of public health and education expenditures as a share to GDP. A larger middle class does not, however, mean a more state-driven economy. To the contrary, an increase in the size of the middle class produces a more market-oriented economy by reducing tariffs on international trade and liberalizing the financial sector. No less remarkable is the effect on the quality of governance. An expansion of the middle class induces an improvement in democratic participation and a decline in official corruption. The indicators of poverty and inequality are also relevant determinants for social policies, economic structure, and governance quality, but not always in the expected way or with the consistency shown by the middle class measure. For instance, a decrease in income inequality seems to produce a decline in official corruption (as possibly expected) but also a reduction in democratic participation (which may be harder to explain).4 Similarly, a decrease in the poverty headcount appears to induce a liberalization of international trade but also, surprisingly, a constriction of credit markets. 5. Conclusion This paper provides evidence that the expansion of the middle class may be a relevant aspect of economic development for reforming government policies and institutions. Using a GMM estimator on a panel of cross-country and time-series observations, the paper finds that when the size of the middle class increases, measured as the proportion of people with income above 10 US dollars a 4 For a review of potentially ambiguous effects of inequality see, among many, Bollen and Jackman (1995), Weede (1997), Piketty (1995), Benabou (2000), Benabou and OK (2001), Banerjee and Duflo (2003) and Esteban and Schneider (2008). 8 day, social policy on health and education becomes more active and the quality of governance regarding democratic participation and official corruption improves. This does not occur at the expense of economic freedom, as an expansion of the middle class also implies more market- oriented economic policy on trade and finance. Behind these effects, there may be a combination of characteristics associated with the evolution of the middle class: more enlightened and numerous political participation, stronger voice, larger similarity of preferences and values for policy and institutional reforms, and stronger stakes into property rights and wealth accumulation, among others. Distinguishing the relative importance of these mechanisms could be a worthy subject for future research. 9 References Acemoglu, Daron, Simon Johnson, and James A. Robinson. (2001). “The Colonial Origins of Comparative Development: An Empirical Investigationâ€? American Economic Review 91(5): 1369-1401. Acemoglu, D., Johnson, S., Robinson, J. A. and Yared, P. (2008). “Income and Democracy.â€? American Economic Review. 98(3). 808-42. Acemoglu, D., Johnson, S., Robinson, J. A. and Yared, P. (2009). “Reevaluating the Modernization Hypothesis.â€? Journal of Monetary Economics. 56. 1043-58. Alesina, A. and Rodrik, D. (1994). “Distributive Politics and Economic Growth.â€? Quarterly Journal of Economics. 109. 465-90. Alesina, A. and Perotti, R. (1996). “Income Distribution, Political Instability, and Investment.â€? European Econonomic Review. Arellano, M. and Bond, S. (1991). “Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations.â€? Review of Economic Studies. 58(2). 277-97. Arellano, M. and Bover, O. (1995). “Another look at the instrumental-variable estimation of error-components models.â€? Journal of Econometrics. 68(1). 29-52. Banerjee, A. V. and Newman, A. F. (1993). “Occupational Choice and the Process of Development.â€? Journal of Political Economy. 101(2). 274-98. Banerjee, A. V. and Duflo, E. (2003). “Inequality and Growth: What Can the Data Say?â€? Journal of Economic Growth. 8. 267-299. Barro, R. J. (1991). “Economic Growth in a Cross Section of Countries.â€? The Quarterly Journal of Economics. 106(2). 407-43. Barro, R. J. (1999). “Determinants of Democracy.â€? Journal of Political Economy. 107(6). 158-183. Benhabib, J. and Rustichini, A. (1996). “Social Conflict and Growth.â€? Journal of Economic Growth. 1(1). 125-42. Benhabib, J. and Przeworski, A. (2006). “The Political Economy of Redistribution under Democracy.â€? Economic Theory. 29(2). 271-290. Benhabib, J., Corvalan, A. and Spiegel, M. M. (2011). “Reestablishing the Income-Democracy Nexus.â€? NBER Working Paper. No. 16832. Benabou, R. (2000). “Unequal Societies: Income Distribution and the Social Contract,â€? American Economic Review. 90(1). 96-129. Benabou, R. and Ok, E. A. (2001). “Social Mobility And The Demand For Redistribution: The POUM Hypothesis.â€? Quarterly Journal of Economics. 116(2). 447-487. Bollen, K. and R. Jackman (1995). “Income Inequality and Democratization Revisited: Comment on Muller.â€? American Sociological Review. 60(6): 983-989. Caselli, F. (2005). “Accounting for Cross-Country Income Differences.â€? Handbook of Economic Growth 1 (1): 679-741. Chang, R., Kaltani, L. and Loayza, N. (2009). “Openness Can Be Good for Growth: The Role of Policy Complementarities.â€? Journal of Development Economics. 90(1). 33-49. Doepke. M. and Zilibotti, F. (2005). “Social Class and the Spririt of Capitalism.â€? Journal of the European Economic Association. 3(2-3). 516-24. Doepke. M. and Zilibotti, F. (2008). “Occupational Choice and the Spirit of Capitalism.â€? Quarterly Journal of Economics. 123(2). 747-93. Dollar, D. and Kraay, A. (2002). “Growth Is Good for the Poor.â€? Journal of Economic Growth. 7(3). 195-225. Easterly, W. (2001). “The Middle Class Consensus and Economic Development.â€? Journal of Economic Growth. 6. 317-335. Easterly, W. and Levine, R. (2001). “It’s Not Factor Accumulation: Stylized Facts and Growth Models.â€? World Bank Economic Review 15(2). 177-219. Easterly, W. and Levine, R. (2003). “Tropics, Germs, and Crops: How Endowments Influence Economic Development.â€? Journal of Monetary Economics. 50(1). 3-39. Easterly, W., Ritzen, J. and Woolcock, M. (2006). “Social Cohesion, Institutions, and Growth.â€? Economics & Politics. 18(2). 103-20. 10 Epstein, D. L., Bates, R., Goldstone, J., Kristensen, I. and O’Halloran, S. (2006). “Democratic Transitions.â€? American Journal of Political Science. 50(3). 551-69. Esteban, J. and Schneider, G. (2008). “Polarization and Conflict: Theoretical and Empirical Issues.â€? Journal of Peace Research. 45(2). 131-141. Foellmi, R. and Zweimueller, J. (2006). “Income Distribution and Demand-Induced Innovations.â€? Review of Economic Studies. 73(4). 941-60. Foellmi, R. and Oechslin, M. (2008). Why Progressive Redistribution Can Hurt the Poor.â€? Journal of Public Economics. 92. 738-47. Galor, O. and Zeira, J. (1993). “Income Distribution and Macroeconomics.â€? Review of Economic Studies. 60. 35- 52. Galor, O. and Moav, O. (2004). “From Physical to Human Capital Accumulation: Inequality and the Process of Development.â€? Review of Economic Studies. 71(4). 1001-26. Glaeser, E. L., La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (2004). “Do Institutions Cause Growth?â€? Journal of Economic Growth. 9. 271-303. Goldthorpe, J. H., and McKnight, A. (2004). “The Economic Basis of Social Class.â€? CASE Paper. 80. London School of Economics. Keefer, P. and S. Knack (2002). “Polarization, Politics and Property Rights: Links between Inequality and Growth,â€? Public Choice. 111(1-2): 127-54. Kharas, H. (2010). “The emerging middle class in developing countries.â€? OECD Development Centre Working Paper. No. 285. Krueger, A. O. (1995). “Policy Lessons from Development Experience since the Second World War.â€? In Handbook of Development Economics, Vol. 3, ed. Jere Behrman and T. N. Srinivasan, 2497-2550. Amsterdam: North-Holland. Lipset, S. M., (1959). “Some Social Requisites of Democracy: Economic Development and Political Legitimacy.â€? American Political Science Review. 53(1). 69-105. Lopez, H. and Serven, L. (2009). “Too Poor to Grow.â€? World Bank Policy Research Working Paper. No. 5012. Lopez-Calva, L. F., and Ortiz-Juarez, E. (2011). “A Vulnerability Approach to the Definition of the Middle Class.â€? World Bank Policy Research Working Paper. No. 5902. Matsuyama, K. (1992). “The Market Size, Entrepreneurship, and the Big Push,â€? Journal of the Japanese and International Economies. 6(4). 347-64. Matsuyama, K. (2002). “The Rise of Mass Consumption Societies.â€? Journal of Political Economy. 110(5). 1035- 1070. Milanovic, B. and Yitzhaki, S. (2002). “Decomposing world income distribution: Does the world have a middle class?â€? Review of Income and Wealth. 48(2). 155-178. Murphy, K. M., Shleifer, A. and Vishny, R. W. (1989a). “Industrialization and the Big Push.â€? Journal of Political Economy. 97(5). 1003-26. Murphy, K. M., Shleifer, A. and Vishny, R. W. (1989b). “Income Distribution, Market Size, and Industrialization.â€? Quarterly Journal of Economics. 104(3). 537-64. Perotti, R. (1996). “Growth, Income Distribution, and Democracy: What the Data Say.â€? Journal of Economic Growth. 1: 149–187. Persson, T. and Tabellini, G. (1994). “Is inequality harmful for growth?â€? American Economic Review. 84: 600– 621. Piketty, T. (1995). “Social Mobility and Redistributive Politics.â€? Quarterly Journal of Economics. 110(3). 551- 584. Voitchovsky, S. (2005). “Does the Profile of Income Inequality Matter for Economic Growth? Distinguishing Between the Effects of Inequality in Different Parts of the Income Distribution.â€? Journal of Economic Growth. 10. 273-96. Weede, E. (1997). “Income inequality, democracy and growth reconsidered.â€? European Journal of Political Economy. 13(4): 751–764 11 Table 1: Data Sources Number of Number of Income or Source Focus countries observations expenditures Latin America and SEDLAC 22 254 Income the Caribbean Eastern Europe ECAPOV 14 77 Expenditures and Central Asia LIS 21 102 Income High income Expenditures Low and Middle PovCal 71 239 (88%) income Sources: SEDLAC: Socio-Economic Database for Latin America and the Caribbean, http://sedlac.econo.unlp.edu.ar/eng/; ECAPOV: World Bank’s internal database for Eastern Europe and Central Asia; LIS: Luxembourg Income Study, http://www.lisdatacenter.org/; PovCal: http://iresearch.worldbank.org/PovcalNet/povcalSvy.html. Table 2: Correlations among Income/Expenditures Distribution Variables Middle Class Poverty Inequality Output per capita Middle Class 1.000 Poverty -0.749 *** 1.000 Inequality -0.320 *** 0.201 *** 1.000 Output per capita 0.857 *** -0.923 *** -0.190 *** 1.000 Table 3: Social Outcomes and Economic Development Social Policy Economic Structure Governance Health Education Expenditures Expenditures Mean Applied Credit Mkt / GDP / GDP Tariff Liberalization Polity Score Corruption Output per capita 1.209*** 0.717*** -2.179*** 0.585*** 2.727*** -0.656*** (ln of GDP per capita) [6.447] [3.183] [-3.882] [3.897] [4.563] [-4.423] Observations (5 year averages) 269 192 265 294 318 285 Number of countries 107 97 103 100 106 92 Hansen Test - p value 0.0349 0.330 0.0340 0.308 0.230 0.0450 Notes: 1. z-statistics in brackets 2. *** p<0.01, ** p<0.05, * p<0.1 3. For explanatory variables, USD are in PPP adjusted, constant 2005 prices 12 Table 4: The Middle-Class Effect Social Policy Economic Structure Governance Health Education Expenditures Expenditures Mean Applied Credit Mkt / GDP / GDP Tariff Liberalization Polity Score Corruption Middle Class 2.054*** 2.918** -10.945*** 1.357*** 6.431*** -1.764*** (% of population with income above 10 USD) [3.849] [2.337] [-3.072] [2.799] [4.068] [-4.767] Poverty -0.019** -0.042** 0.203*** 0.047*** -0.042** -0.011* (2.5 USD a day Poverty Headcount) [-2.411] [-2.472] [2.874] [3.383] [-2.345] [-1.825] Inequality -3.716** 3.028 20.736*** -0.209 9.262** 4.083*** (Gini Index) [-2.456] [1.360] [3.373] [-0.165] [2.164] [3.866] Output per capita 0.121 -0.922 5.485** 1.292*** 0.280 -0.470** (ln of GDP per capita) [0.416] [-1.310] [2.439] [3.009] [0.334] [-2.218] Observations (5 year averages) 269 192 265 294 318 285 Number of countries 107 97 103 100 106 92 Hansen Test - p value 0.174 0.640 0.934 0.469 0.701 0.451 Notes: 1. z-statistics in brackets 2. *** p<0.01, ** p<0.05, * p<0.1 3. For explanatory variables, USD are in PPP adjusted, constant 2005 prices 13