WPS6568 Policy Research Working Paper 6568 Growth Still Is Good for the Poor David Dollar Tatjana Kleineberg Aart Kraay The World Bank Development Research Group Macroeconomics and Growth Team August 2013 Policy Research Working Paper 6568 Abstract Incomes in the poorest two quintiles on average increase the poorest quintiles. These findings hold across most at the same rate as overall average incomes. This is regions and time periods and when conditioning on because, in a global dataset spanning 118 countries over a variety of country-level factors that may matter for the past four decades, changes in the share of income of growth and inequality changes. This evidence confirms the poorest quintiles are generally small and uncorrelated the central importance of economic growth for poverty with changes in average income. The variation in changes reduction and illustrates the difficulty of identifying in quintile shares is also small relative to the variation specific macroeconomic policies that are significantly in growth in average incomes, implying that the latter associated with the relative growth rates of those in the accounts for most of the variation in income growth in poorest quintiles. This paper is a product of the 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 akraay@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 Growth Still Is Good for the Poor David Dollar (Brookings Institution) Tatjana Kleineberg (Yale) Aart Kraay (World Bank) Keywords: growth, inequality JEL Classification Codes: O4, O11, I3 _____________________ ddollar@brookings.edu, tatjana.kleineberg@yale.edu, akraay@worldbank.org. We would like to thank, without implication, Kaushik Basu, Stefan Dercon, Phil Keefer, Luis Serven, and Martin Ravallion for helpful comments. Financial support from the Knowledge for Change Program of the World Bank is gratefully acknowledged. The views expressed here are the authors' and do not reflect those of the Brookings Institution, the World Bank, its Executive Directors, or the countries they represent. 1. Introduction Absolute poverty has fallen sharply in the developing world over the past three decades. In 1980, 52 percent of the world’s population lived below the World Bank’s $1.25/day poverty line. By 1990, the incidence of poverty had fallen to 42 percent, and to 21 percent in 2010. Much of this reduction has been due to rapid growth in large and initially poor developing countries such as China and India. But in all regions of the world, rapid growth has been systematically associated with sharp declines in absolute poverty. This success in poverty reduction has meant that low global absolute poverty lines, like the World Bank's $1.25/day standard, have become less relevant for many developing countries where today only a small fraction of the population lives below this austere threshold. This led the World Bank to put a new institutional emphasis on tracking “shared prosperity�, in addition to monitoring absolute poverty. “Shared prosperity� is defined in terms of the growth rate of incomes in the bottom 40 percent of households, and the World Bank has made a public commitment to supporting policies that foster “shared prosperity� in the developing world. 1 Concerns about “shared prosperity� are also widespread in advanced economies, where many fear that growth no longer benefits the bottom half of the income distribution. 2 This emphasis on “shared prosperity� naturally raises the question of the extent to which it differs from simply “prosperity�, where the latter could be defined as overall aggregate income growth. In this paper, we address this question, updating and elaborating on our earlier work in Dollar and Kraay (2002). In that paper, we studied the relationship between growth in average incomes of the poorest 20 percent of the population, and growth in average incomes, using a large cross-country panel dataset on average incomes and inequality. Our main findings in that paper were that (i) incomes in the poorest quintile on average increase equiproportionately with average incomes, reflecting the lack of a systematic correlation between growth and changes in the first quintile share, and (ii) this relationship is very strong, reflecting the fact that most of the variation in growth in incomes in the poorest quintile 1 See World Bank (2013). 2 As an example of this, in a recent speech at Knox College in Galesburg, Illinois on July 24, 2013, President Barack Obama described the US economy as “... a winner-take-all economy where a few do better and better, while everybody else just treads water�. More systematically, a recent Pew Global Survey found that a strong majority of respondents in 14 advanced economies felt that the gap between rich and poor was increasing in recent years. The fraction holding this view ranged from a low of 58 percent in Japan to a high of 90 percent in Spain (Pew Research Center, 2013). 2 reflected growth in average incomes, rather than changes in the share of income accruing to the poorest quintile. Over the past 15 years since we began work on that paper, the quality and quantity of available household survey data on income distribution have improved dramatically, providing rich new information that can be used to revisit the evidence on the relationship between overall growth and growth in the poorest quintiles. We work with a large cross-country dataset of high-quality survey-based measures of average incomes and income distributions, drawing on the POVCALNET database 3 of the World Bank for developing countries, and the Luxembourg Income Study (LIS) data 4 for advanced economies. Using this combined dataset, which covers 118 countries for which household surveys are available for at least two years since the 1970s, we revisit the relationship between growth in average incomes and growth in the poorest quintiles. Updating the work in Dollar and Kraay (2002), we consider growth rates of the poorest 20 percent of the population, and given the new emphasis on “shared prosperity�, we also consider growth rates of the poorest 40 percent of the population. Echoing our earlier work, this expanded and updated dataset reveals a very strong equiproportionate relationship between average incomes in the poorest quintiles, and overall average incomes. In our preferred benchmark specification, covering 299 non-overlapping within-country growth episodes at least five years long, the slope of the relationship between growth in average incomes in the poorest quintiles and growth in overall average incomes is very close to – and not significantly different from – one. Moreover, a standard variance decomposition indicates that 62 percent (77 percent) of the cross-country variation in growth in incomes of the poorest 20 percent (40 percent) of the population is due to growth in average incomes. These findings for the most part hold across different regions and over time, and across a variety of different robustness checks. This basic result underscores the central importance of overall growth for improvements in living standards among the poorest in societies. Although the portion of the variation in growth in incomes in the poorest quintiles due to changes in inequality is -- on average -- both small and uncorrelated with growth in average incomes, it is nevertheless important to understand its other correlates. In particular, if one combination of macroeconomic policies and institutions that support a given aggregate growth rate also leads to an increase in the share of incomes accruing to the poorest quintiles, while another combination did the opposite, then the former would be preferable from the standpoint of promoting shared prosperity. We 3 See PovcalNet Database (2013). 4 See Luxembourg Income Study (LIS) Database (2013). 3 therefore investigate how growth in incomes of the poor correlates with a variety of country-level variables commonly thought to matter for growth (e.g. financial depth, financial openness, inflation rate, budget balance, trade openness, life expectancy, measures of internal and external conflicts, population growth, life expectancy and civil liberties), as well as a number of variables often considered to matter directly for inequality (e.g. primary school enrollments, inequality in educational attainment, government expenditure in education and health, and agricultural productivity). In the spirit of data description, we use Bayesian Model Averaging to systematically document the partial correlations between these variables and growth in incomes of the poor, conditional on growth in average incomes, for all possible combinations of these variables. We find at best very modest evidence that any of the policies and institutions reflected in these variables are significantly correlated with growth in incomes of the poor, beyond any direct effect of these variables on growth itself. These findings illustrate the difficulty in using cross-national data to identify specific macro policy reforms that disproportionately support growth in the poorest income quintiles. Moreover, the particularly strong relationship between growth in incomes of the bottom 40 percent and growth in average incomes, and the lack of evidence of systematic correlates of the difference between the two, underscores the central importance of rapid growth in average incomes as a means to achieving “shared prosperity�. The rest of this paper proceeds as follows. Section 2 describes our empirical framework, as well as the cross-country panel of household survey data on which our results are based. Section 3 presents our core results on the bivariate relationship between incomes of the poor and average incomes, and subjects them to a variety of robustness checks. Section 4 considers the additional impact of a variety of policy and institutional variables on the income share of the poor. Section 5 concludes. 2. Empirical Strategy and Data 2.1. Basic Setup Our starting point is the identity that relates incomes of the poor to average incomes: (1) 𝑌 𝑃 = 𝑆 𝑃 𝑌 where 𝑌 𝑃 denotes average income in either the bottom 20 or 40 percent of the income distribution; 𝑆 𝑃 𝑄1 denotes the income share of the first quintile divided by 0.2 ( ) or the share of the bottom two 0. 2 4 𝑄1 +𝑄2 quintiles divided by 0.4 ( 0. 4 ); and 𝑌 denotes overall average income. As discussed below, in roughly half of the surveys in our dataset, the relevant welfare measure is consumption expenditure, while in the other half it is income. However, for terminological convenience we will refer only to income. Also, while our dataset is an unbalanced and irregularly-spaced panel of country-year observations where survey data are available, for notational convenience we will suppress country and year subscripts. Taking log differences over time results in the following expression for growth in incomes of the poor: (2) ∆ln𝑌 𝑃 = ∆ ln 𝑆 𝑃 + ∆ln𝑌 That is, increases in incomes of the poor can mechanically be decomposed into increases in average incomes, and increases in the share of income accruing to the poor. In order to investigate these two factors, we begin by estimating a series of regressions of growth in incomes of the poor on growth in average incomes. The slope coefficient from this regression is 𝐶𝑂𝑉 (∆ln𝑌 𝑃 , ∆ln𝑌) 𝐶𝑂𝑉 ( ∆ ln 𝑆 𝑃 , ∆ln𝑌) (3) =1+ 𝑉 (∆ln𝑌) 𝑉(∆ln𝑌) where the equality follows from the definition of growth in incomes of the poor. When this estimated slope coefficient is equal to one, incomes of the poor increase on average at the same rate as overall average incomes. This is because the income share of the poorest does not vary systematically with 𝐶𝑂𝑉 (∆ ln 𝑆 𝑃 , ∆ln𝑌) changes in average income, i.e. = 0. If however the estimated slope coefficient is 𝑉 (∆ln𝑌) greater (less) than one, incomes of the poor rise faster (slower) than average incomes, reflecting a positive (negative) correlation between growth and the income share of the poor. A related question has to do with the relative importance of these two sources of growth in average incomes of the poor. We document this using a standard variance decomposition, which defines the share of the variation of growth in incomes of the poorest due to growth in average incomes as: 𝑉 (∆ln𝑌) + 𝐶𝑂𝑉 (∆ ln 𝑆 𝑃 , ∆ln𝑌) (4) 𝑆 = 𝑉 (∆ ln 𝑆𝑃 + ∆ln𝑌) 5 In the data, we shall see that 𝐶𝑂𝑉 (∆ ln 𝑆 𝑃 , ∆ln𝑌) is small in most specifications, and so this variance share primarily reflects the relative variances of average incomes and incomes of the poor. When the variation in changes in the poorest quintile shares is small, then the share of the variation in growth in incomes of the poor due to growth in average incomes will be close to one. 5 We report this variance decomposition in all of the tables of results that follow, as a useful summary of the relative importance of growth and changes in inequality in driving growth in incomes of the poor. In the last part of our empirical results, we report a series of regressions of growth in average incomes of the poor on growth in average incomes, augmented by various combinations of variables intended to capture a range of policies and institutions that may matter for growth and changes in inequality. The estimated slope coefficients capture the partial correlations between these variables and growth in the income share of the poorest, conditional on growth in average incomes. Given the identities above, this is equivalent to regressing changes in a particular measure of inequality, the income share of the poor, on growth in average incomes and a set of additional variables. If these additional variables are not significant, this means that they are not systematically associated with changes in the income share of the poor, conditional on overall growth. 2.2. Measuring Growth in Average Income and Income of the Poor Our starting point is a large dataset of 963 country-year observations for which household surveys are available, covering a total of 151 countries between 1967 and 2011. This dataset is the merger of data available in two high-quality compilations of household survey data: the World Bank’s POVCALNET database, covering primarily developing countries, and the Luxembourg Income Study (LIS) database, covering primarily developed countries. The POVCALNET database is the dataset underlying the World Bank's widely known global poverty estimates. Its data on average incomes and income distribution are based on primary household survey data. In most cases, surveys are representative for the whole country. 6 Roughly half of the surveys in the POVCALNET database report income and its distribution, while the other half report consumption expenditure and its distribution. As noted earlier, 5 See Klenow and Rodriguez-Clare (1997) for a more formal justification of this variance decomposition in a growth context. This variance share is closely related to the R-squared from a basic regression of growth in average 𝑉 (∆ln𝑌 𝑃 ) incomes of the poor on growth in average incomes, i.e. 𝑅2 = 𝑆 2 . 𝑉 (∆ln𝑌 ) 6 In the case of Argentina and Uruguay, survey data is only available for urban areas; however, due to high urbanization rates (over 90%) this seems to be an acceptable proxy for the national income distribution. 6 however, for terminological convenience we will refer only to income. All survey means are expressed in constant 2005 US dollars adjusted for differences in purchasing power parity. For countries that are not covered in POVCALNET, we rely on the LIS database. 7 This expands our sample by adding 19 OECD economies. For these countries we construct mean income and income shares of the poorest directly from the micro data at the household level. The underlying surveys are nationally representative and intended to be comparable over time. We focus on the LIS measure of household disposable income, which is expressed in the raw data in current local currency units. We convert the survey means to constant 2005 USD and then apply the 2005 purchasing power parity for consumption from the Penn World Table, in order to be consistent with the POVCALNET data. Figure 1 gives an overview of the annual data availability from these two sources. LIS survey data starts earlier, going back to 1967, while POVCALNET observations start in the 1980s. Both databases have better country coverage in more recent years. For our empirical analysis, we organize the data into “spells�, defined as within-country changes in variables of interest between two survey years. Specifically, we calculate average annual log differences of average incomes, incomes of the poor, and quintile shares for each spell, recognizing that different spells cover periods of different length, depending on the availability of household survey data. We work with three sets of spells corresponding to different time horizons. The first set consists of all possible consecutive non-overlapping spells, beginning with the first available survey for each country. This largest sample consists of 735 spells in 123 countries, with a median spell length of 2 years. A drawback of this sample is that the time period covered by many spells is quite short, and moreover a small number of countries with high frequency availability of surveys are over-represented in this sample. In order to be able to study the relationship between incomes of the poor and average incomes over longer horizons, we work with two additional sets of spells. The second consists of all possible consecutive non-overlapping spells by country, but imposing a minimum length of five years for each spell. This results in a set of 299 spells and a smaller set of 117 countries. The median spell length is 6 years. The third sample considers only the longest available spell for each country. This results in 118 spells with a median spell length of 16 years. 8 7 A handful of countries have surveys available both through POVCALNET and LIS. For these countries we use only the POVCALNET data, i.e. we do not switch within countries between POVCALNET and LIS. 8 In all three sets of spells, we trim extreme observations using the following criteria: (i) we trim the distribution of th growth rates of income shares of the bottom 20 and 40 percent at the first and 99 percentile in each sample, and 7 The minimum-five-year-spell sample is our preferred sample. As noted above, the all-spells sample overweighs those countries in which surveys are more frequent; furthermore, the year-to-year changes in inequality may have a less favourable signal-to-noise ratio than those observed over longer intervals. The long-spell sample has the disadvantage that it does not include any within-country variation in growth rates. We report results for both the all-spells and long-spells to ensure the robustness of the results, but focuse primarily on the minimum-five-year-spell sample. Appendix Table A1 summarizes the country coverage and data availability. Table 1 provides summary statistics on annual growth in overall average incomes, the first quintile share, and the sum of the first two quintile shares. The basic story is clear from the summary statistics. Consider for example Panel 1: for the 299 observations in the minimum-five-year-spell sample, the mean growth rate of average income is 1.4 percent per year and the mean change in the share of the bottom 40 percent is 0 percent per year. This implies that the growth rate of income of the bottom 40 percent is also 1.4 percent per year on average. Furthermore, the correlation of the change in the bottom 40 percent share and mean income growth is 0.007, which is insignificantly different from zero. Finally, growth rates in average incomes vary considerably more across spells than growth rates of the income share of the bottom 40 percent: the standard deviations of these two growth rates are 4.7 versus 2.5 percent. This implies that the bulk of the variation in growth in incomes of the poor is attributable to growth in average incomes. The second panel of Table 1 reveals some interesting heterogeneity by disaggregating the five- year spells by geographical region (the assignment of countries to geographical regions is noted in Appendix Table A1). Unsurprisingly growth rates in average incomes vary greatly across regions, ranging from near zero percent per year in the Middle East North Africa sample, to a high of 3.4 percent per year in East Asia. East Asia also stands out in the sense that rising incomes are correlated across spells with rising inequality: the correlation of the growth rate of the first (first two) quintile shares with growth in average incomes is around -0.5. Nevertheless, growth in average incomes of the poor according to either definition (i.e. the sum of the first and fourth, and first and seventh columns of Table 1) is substantially higher in this region compared with any other. (ii) we trim the distribution of the difference between the growth rate of the survey mean and the corresponding th growth rate of private consumption from the national accounts, also at the first and 99 percentiles. This results in the small changes in the number of countries represented in each sample noted in the main text. In addition to data cleaning, one country (Bhutan) is dropped from the minimum-five-year-spell sample as data is only available for four years. However, the minimum five-year criterion is not imposed in the long-spells sample, which therefore includes one more country than the five-year spells sample. 8 The last two panels in Table 1 disaggregate the summary statistics by decade and by region, again focusing on the five-year spells. A practical challenge for data description here is that only a small fraction of spells fall entirely within a single decade, and so it is not obvious how to assign the remaining spells to decades. To circumvent this problem, for each spell we define three variables measuring the fraction of years in the spell falling in each of three decades. For example, a spell lasting from 1989 to 1994 would have one-fifth of its years in the 1980s and four-fifths in the 1990s, and none in the 2000s. We then report weighted summary statistics by decade, weighting each spell by the fraction of observations falling in each decade. The importance of overall growth for incomes of the poor can be seen by comparing the statistics for the 1980s and the 2000s: for the observations in the 1980s, mean income growth averaged -0.3 percent while there was a slight shift in favor of the income of the bottom 40 percent, resulting in zero income growth for the bottom 40 percent. In the 2000s, growth accelerated to an average of 3.0 percent; again there was a small shift in favor of the bottom 40 percent and their income grew at 3.4 percent per year. 3. Main Results Our baseline empirical specification consists of a simple OLS regression of growth in incomes of the poor on mean income growth. Table 2 documents these results for the three samples with different spell lengths as described above. Panel A provides the results for the poorest quintile and Panel B for the poorest two quintiles. For all three samples, we cannot reject the null hypothesis that the slope coefficient is equal to one, indicating the absence of a statistically significant relationship between growth in average incomes and growth in the income shares of the poorest. This holds both when the poor are defined as those in the bottom 20 percent, and in the bottom 40 percent, the latter corresponding to the “shared prosperity� measure advocated by the World Bank. In our preferred sample of spells at least five years long, the estimated slope coefficient is 1.06 for the bottom 20 percent, and 1.00 for the bottom 40 percent, indicating that average growth is reflected on average one-for-one in growth in incomes of the poor. In the samples of all spells, and long spells, the estimated slopes are slightly smaller than one, but again not significantly so. The top panel of Figure 2 shows the relationship between growth in average incomes (on the horizontal axis) and growth in incomes in the poorest two quintiles (on the vertical axis), focusing on our preferred sample of spells at least five years long. Consistent with the results in Table 2 , the slope of the fitted relationship is nearly indistinguishable from the 45-degree line. Moreover, it is clear that this relationship is very strong. The R-squared from the corresponding regression in Table 2 is 0.78, and the 9 share of the variance of growth in average incomes in the bottom 40 percent due to growth in average incomes is 77 percent. The bottom panel of Figure 2 shows the same relationship, in the three sets of spells. In all three sets of spells, the estimated slopes are close to one, and the corresponding R-squareds are large, ranging from 67 to 78 percent. We next investigate how this relationship varies across geographical regions and over time. Table 3 shows that our basic finding of a tightly estimated equiproportional relationship between growth in incomes of the poor, and growth in average incomes, holds in most regions, and particularly so for average incomes in the bottom 40 percent of the population. The main exception is the East Asia and Pacific region, where the estimated slopes are substantially smaller than one (and significantly so in the case of incomes of the bottom 40 percent). This indicates that in this region, spells with faster growth in average incomes were more likely to also have decreases in the income share of the poorest quintiles. However, this does not imply that those in the poorest quintiles fared particularly poorly in such spells. Recall from Table 1 that average incomes in East Asia grew fastest among all regions at 3.4 percent per year, and incomes in the poorest 40 percent rose at 3.2 percent per year on average, faster than in any other region. In Table 4 we investigate how the relationship between growth in average incomes and growth in incomes of the poor varies over time and by region. Combining all countries, the slope of the estimated relationship is close to one across the 1980s, 1990s, and 2000s, and in all three cases is not significantly different from one. The strength of the estimated relationship, and the corresponding share of the variance of growth in incomes of the poor due to overall growth, also does not vary much across decades, ranging from a low of 58 percent in the 2000s to a high of 66 percent in the 1980s for the poorest quintile. For the bottom 40 percent, the corresponding figures range from 75 to 77 percent. When we break the results down by region there is some interesting variation. The combined East and South Asia region has a slope coefficient substantially lower than 1.0 in both the 1990s and the 2000s (and significantly so in the 1990s). Here the fastest growing countries, notably China, have had increases in income inequality so the growth of income of the bottom 40 percent lags behind average income growth. Latin America shows the opposite tendency in the 2000s, with a slope coefficient significantly greater than 1.0. This means that in faster-growing Latin American countries, income shares of the bottom quintiles also increased more, so that growth in the bottom 20 and 40 percent outstripped growth in average incomes. This gap is substantial. Referring back to Table 1, growth in average incomes in Latin America in the 2000s was 1.2 percent per year on average, while the income share of the poorest 10 40 percent grew at 1.1 percent per year on average, for an overall growth rate for the poorest 40 percent of 2.3 percent per year. Still, income growth of the bottom 40 percent in Asia was at an even higher rate of 3.7 percent per year during the 2000s, because the overall average growth rate in Asia was so high. In all of our results so far, we have relied exclusively on household survey data to construct measures of average income growth and growth in incomes of the poor. However, many past studies, including our own work in Dollar and Kraay (2002), relied on national accounts growth rates to measure overall average income growth. A large literature has discussed substantial differences between growth in survey mean income and corresponding aggregates in the national accounts in some countries (see for example Deaton (2005) and Deaton and Kozel (2005) for the case of India in particular). These differences are illustrated in Figure 3, which plots average annual growth in household survey mean income (on the vertical axis), and growth in the same period taken from the national income accounts (on the horizontal axis). 9 From this figure, substantial differences in these two alternative measures of growth in average living standards are clearly apparent in the large deviations from the 45-degree line for many spells. Without taking a stand on relative merits of national accounts versus household surveys as a measure of average living standards, we perform some simple robustness checks to see how our findings change if we rely on national accounts growth rates instead of household survey mean growth rates. The results are presented in Table 5. The first panel reproduces our benchmark specification in the slightly smaller samples of spells for which both national accounts growth and household survey growth rates are available. Dropping these few spells makes very little difference for our benchmark results, which are quite similar to those in Table 2. The second panel reports results replacing household survey growth with the corresponding national accounts growth rate (and of course also using the national accounts growth rate plus the growth rate of the relevant quintile shares to compute growth in incomes of the poor). The estimated slope coefficients are slightly larger than when using the survey means, suggesting there is a more positive correlation between changes in the poorest quintile shares and national accounts growth rates than household survey mean growth rates. However, in all but one case, this relationship is not statistically significant, as the estimated slopes are not significantly different from one. The one exception is using the minimum five-year spells, and considering incomes of the bottom 20 percent. In the third panel of Table 5, we follow the approach suggested in Chen and 9 As we have noted earlier, the household survey data are a mix of income and consumption surveys. This raises the question of which national accounts aggregate is the closest corresponding measure. Here we compare with real private consumption growth in all countries, following Ravallion and Chen (2008). 11 Ravallion (2008), using a simple average of the household survey mean and national accounts growth rates. 10 Since household survey mean growth rates vary much more than consumption growth rates in the national accounts, they dominate these average growth rates. As a result, this mixed method leads to findings that are very similar to those in the first panel of Table 5. Overall, our findings show that the poor on average benefit equiproportionally from overall growth, and these findings hold across most regional and temporal disaggregations of the data, and across a variety of further robustness checks. In most cases this relationship is also fairly tightly estimated, particularly for income growth in the poorest 40 percent, where our benchmark findings suggest that nearly 80 percent of the variation in growth in average incomes of the poorest 40 percent is attributable to growth in average incomes. At the same time, however, it is important to recognize that these are in a sense “non-results�, because they simply confirm that growth is distribution-neutral on average, and that changes in relative incomes tend to be substantially smaller than growth in overall average income. 4. Policies, Institutions, and Growth in Incomes of the Poor The previous section has shown that average incomes of the poor tend to rise at the same rate as overall average incomes, implying that policies and institutions that stimulate higher growth benefit the poor equiproportionately on average. Moreover, we have seen that most of the cross-country variation in growth in incomes of the poor reflects growth in average incomes, rather than changes in the share of income captured by the poorest quintiles. Nevertheless, it is possible that growth from different sources or in different institutional contexts has a differentiated effect on the growth in incomes of the poor, to the extent that such policies and institutions are correlated with the part of the variation in growth in incomes of the poor that is due to changes in the income share of the poor. This information would be valuable for policy-makers seeking to pursue the goal of reducing inequality by promoting “pro-poor� growth or “shared prosperity�. In this section, we augment our basic specification to include two sets of variables that serve as proxies for a variety of policies and institutions that might matter for growth, and those that might be relevant for changes in relative incomes. The growth correlates include a measure of financial 10 Chen and Ravallion (2008) show that under certain strong assumptions (a lognormal distribution of growth rates and equal variance of measurement error across the two sources), treating national accounts data on consumption as a prior, and household surveys as data, the natural posterior estimate of mean living standards is an equally- weighted geometric average of the two. In log-differences this implies a simple average of the two growth rates. 12 development (M2 as percentage of GDP), the Sachs-Warner indicator of trade openness, the Chinn-Ito Index of financial openness, the inflation rate, the general government budget balance, life expectancy, population growth, the Freedom House measure of civil liberties and political rights, assassinations and revolutions per capita, as well as dummies for internal conflicts and war participation. Most of these variables have been identified as important correlates of growth in one or more of three prominent meta-analyses of growth determinants (Fernandez, Ley and Steel (2001a), Sala-i-Martin (2004) and Ciccone et al. (2010)). They are also time-varying, so that we can relate within-country changes in these variables to within-country changes in incomes of the poor. In a second set we include five variables that are intended to proxy for “pro-poor� policies that may matter for the distribution of income, and that have been found to be significant correlates of inequality in the much smaller existing cross-country literature on determinants of inequality. These consist of primary enrollment rates, a measure of educational inequality 11 (as emphasized by De Gregorio et. al. (2002)), public spending on health and on education (reflecting the emphasis on redistributive spending in Milanovic (2000), De Gregorio (2002) and Checchi (2008)), and finally the share of agriculture in GDP (as emphasized for example in Datt and Ravallion (2002)). 12 Table A1 provides a detailed description of the definitions and sources of all of these variables. Two comments about these variables are in order. First, distinguishing between those variables that might matter for growth and those that might matter for inequality is inevitably somewhat arbitrary. For example, Jaumotte et al. (2013) find that some variables closely related to some of our growth variables (for example, de facto measures of trade and financial openness) are also significantly correlated with changes in quintile shares in a large cross-country dataset, even though we classify them among our set of growth variables. Second, we emphasize that many papers in the empirical literature on inequality consider the cross-sectional relationship between levels of Gini coefficients and various explanatory variables. In our specifications, we will be considering a different measure of inequality (poorest quintile shares), and moreover we are looking at how changes within countries over time in 11 Specifically, we use data on educational attainment by different levels of attainment from the Barro-Lee dataset to construct a (grouped) Lorenz curve summarizing the distribution of the total number of years of education across individuals, and from this calculate a corresponding Gini coefficient. 12 We also considered several other variables found to be significant correlates of inequality in some papers in the literature, but did not include them in our analysis because data coverage was very poor for many of the developing countries in our sample. These included indicators of labour market regulation and progressivity of tax systems (Checchi et. al. (2008)), public sector employment (Milanovic (2000) ), and social transfers (Milanovic (2000), De Gregorio et. al. (2002)). 13 these inequality measures relate to changes within countries over time in these various candidate explanatory variables. 13 In the spirit of data description, we use Bayesian Model Averaging (BMA) to systematically document the partial correlations between various combinations of these covariates and growth in incomes of the poor. This approach follows a growing literature which relies on BMA to show the robustness of empirical findings in the cross-country growth literature across many model specification choices. 14 The basic idea of BMA is to consider the large set of 2𝐾 empirical models defined by all possible combinations of the set of 𝐾 = 17 variables added to our benchmark specification, rather than to base conclusions on just a few pre-selected models. Let j ϵ {1,2,…,2𝐾 } index the universe of potential models, and let 𝑋𝑗 denotes the particular set of regressors added to our benchmark specification in model 𝑗. Each model 𝑗 thus represents a variation of our benchmark specification, regressing growth in average incomes, ∆𝑌 𝑃 , on growth in average incomes, ∆𝑌, and the change in the corresponding potential determinants of average incomes and/or the poorest quintile share, ∆𝑋𝑗 , i.e.: (5) ∆ 𝑌 𝑃 = 𝛼0𝑗 + 𝛼1𝑗 ∆𝑌 + 𝛼2𝑗 ∆ 𝑋𝑗 + 𝜀𝑗 . The estimated slope coefficients in 𝛼2𝑗 capture the partial correlations between growth in incomes of the poor and the variables included in model 𝑗, conditional on growth in average incomes. And given the definition of average income of the poor, this is of course equivalent to regressing growth in the first (or first two) quintile shares on growth in average incomes, and on the set of variables included in model 𝑗. BMA provides an algorithm for assigning posterior probabilities to each model reflecting their relative likelihoods. These likelihoods in turn reflect the “fit� of the model as summarized by the R- squared, but with a model size penalty that rewards more parsimonious models with fewer regressors. These posterior model probabilities can then be used to combine inferences across different models in a way that reflects their relative likelihood. For each variable, we calculate the Posterior Inclusion Probability (PIP), which is the sum of the posterior model probabilities for each model in which the given 13 In this sense, this part of our analysis is most closely related to Jaumotte et al. (2013) who estimate country-year panel fixed-effects regressions that explain changes in inequality as a function of changes in the explanatory variables. 14 See Fernandez, Ley and Steel (2002) for the seminal application of this technique to cross-country growth empirics. 14 variable is included. High values of the PIP indicate that this variable appears in models that are relatively more likely. In addition, we calculate the posterior probability-weighted average of the estimated slope coefficient for each variable, averaging across all models, and averaging only across those models in which the variable is included. 15 Table 6 and Table 7 show the results, for growth rates in incomes of the poorest 20 percent and 40 percent, respectively. In both tables we focus on the sample of spells at least five years long. The rows of the table correspond to the seventeen variables included in the BMA analysis. In the first five columns we summarize the distribution of the estimated slope coefficients over all 217 = 131,072 models considered by the BMA procedure. Consider for example the first row, which reports the distribution of the estimated coefficient on growth in average incomes. The median estimated coefficient is very close to one, at 1.01 for the bottom 20 percent, and 0.963 for the bottom 40 percent. The range from the minimum to the maximum estimated coefficient is quite narrow (0.91 to 1.10 for the bottom 20 percent, and 0.88 to 1.03 for the bottom 40 percent). Moreover, this slope coefficient is not significantly different from one in any of the specifications considered for the bottom 20 percent and in only 3.5 percent of the specifications for the bottom 40 percent. This indicates that our basic finding of a one-for-one average relationship between growth in incomes of the poor and growth in overall incomes is robust to the inclusion of nearly all combinations of the 17 control variables in the model. Turning to the additional variables, in most cases the distribution of estimated slope coefficients is centered around zero, and most commonly includes many negative as well as positive values. A useful summary in this respect can be found in the sixth and seventh columns of the tables, which report the proportion of specifications in which the estimated slope coefficient is significantly positive, or significantly negative. Of the 17 control variables, only three are significant in more than five percent of the models in which they are included in Table 6 and in Table 7. This indicates that the large majority of these variables are not significantly partially correlated with changes in income share of the poorest quintiles, conditional on overall growth, and conditional on nearly all possible combinations of other variables included in the model. 15 We implement BMA using a standard g-prior for the parameters of each individual regression model, and a prior that assigns a equal probability of 𝜇 /𝐾 that each individual variable is included in a given model (see for example Fernandez, Ley and Steel (2001a) for a seminal application to cross-country growth empirics). We set 𝑔 = 0.01 and 𝜇 = 0.25𝐾 . Since the total number of models is not very large, we implement BMA by exhaustively estimating all possible models, rather than use common numerical algorithms to visit only a subset of relatively more likely models. 15 The three exceptions in Table 6 and in Table 7 are relative growth in agriculture, changes in life expectancy, and inflation. Consistent with existing findings in the literature, faster growth in agriculture is significantly associated with increase of the income share of the poorest 20 percent in 29 percent of the specifications considered. For the poorest 40 percent, faster growth in agriculture enters significantly in 11 percent of the specifications. This reflects the reality that many of the poor in developing countries work in agriculture, so that faster growth in this sector is likely to disproportionately benefit the poor. The results for changes in life expectancy and changes in inflation are somewhat puzzling. In about 25 (42) percent of specifications, increases in life expectancy are significantly associated with reductions in the income share of the poorest 20 (40) percent, while the results suggest in 39 (32) percent of specifications that increases in inflation are associated with a higher income share of the poorest 20 (40) percent. We should not take these puzzling results too seriously however, because the findings hold only for a relatively small set of models, moreover ones with low probabilities. The last three columns of Table 6 and Table 7 incorporate the information generated by BMA about the relative likelihood of the many different models corresponding to different combinations of control variables. By construction, the posterior inclusion probability is equal to one for growth in average incomes, since we include it in every specification. The posterior inclusion probabilities for the other 17 variables are all low, and are below five percent for all except one variable in Table 6 (population growth), and for all except two variables in Table 7 (population growth, internal conflict). This reflects the fact that adding various combinations of control variables to our basic specification does not do much to improve the explanatory power of the model. The BMA algorithm in turn interprets this as low model probabilities for those models that add regressors over the benchmark specification. 16 Another way to see this directly is to consider the distribution of R-squareds in the last row of Table 6 and Table 7. It is striking that the highest R-squared observed across all models is only 0.68 (in the case of the bottom 20 percent), and only 0.79 (in the case of the bottom 40 percent). This is only slightly better than the R-squareds of the corresponding benchmark regressions of growth in incomes of the poor on growth in average incomes alone reported in Table 2, which are 0.65 and 0.78 respectively. 16 The precise magnitudes of these posterior inclusion probabilities are somewhat sensitive to the choices of prior parameters in the BMA analysis. Specifically, smaller values of the prior parameter 𝑔 make the posterior model probabilities more sensitive to improvements in model fit as measured by R-squared. We set 𝑔 = 0.01 which is 1 1 actually larger than benchmark values recommended in the BMA literature such as 𝑔 = = 1/299 or 𝑔 = 2 = � 𝐾 1/172 . See Feldkirchner and Zeugner (2009) and Fernandez, Ley and Steel (2001b). 16 Overall, these results suggest that a large set of plausible macro variables are remarkably unsuccessful in explaining growth in incomes of the poor, beyond any effect that they might have on aggregate growth. This finding in turn implies that historical experience in a large sample of countries does not provide much guidance on which combinations of macroeconomic policies and institutions might be particularly beneficial for promoting “shared prosperity� as distinct from simply “prosperity�. 5. Conclusions Incomes of the bottom 20 percent and bottom 40 percent of the income distribution generally rise equiproportionally with mean incomes as economic growth proceeds. We establish this result in a data-set spanning 118 countries and four decades, updating and expanding the results of Dollar and Kraay (2002). The result holds across decades, including in the 2000s -- hence the conclusion that “growth still is good for the poor.� The shares of the bottom 20 percent and bottom 40 percent are measures of income inequality, and the foundation of our result is that changes in this particular measure of inequality generally are small and uncorrelated with economic growth. The finding is good news in the sense that we can expect economic growth to lift people out of poverty and lead to shared prosperity on average. The result also helps us understand how the rapid growth in the developing world in recent decades has led to such dramatic poverty reduction. A second important finding is that the income shares of the bottom 20 percent and bottom 40 percent show no systematic tendency to decline over time; that is, there is no worldwide trend towards greater inequality, using these measures on a country-by-country basis. During 299 minimum-five-year spells, the average annual growth rate in the income share of the bottom 40 percent is 0.000. Furthermore, there is no tendency for that result to change over time. The average change was 0.003 in the 1980s, -0.003 in the 1990s, and 0.004 in the 2000s. Our third result is that around three-quarters of the variation across countries and over time in growth rates of income of the bottom 20 percent or 40 percent can be explained by variation in growth rates of mean income, while the remainder comes from changes in quintile shares. The fact that changes in quintile shares are zero on average does not mean that there are not some striking changes in inequality in particular countries at particular time periods. We attempt to explain these changes in inequality with variables used in the empirical growth literature, such as measures of macroeconomic stability, trade openness, and political stability. We also include variables that might plausibly increase the income share of the poor (measures of agricultural productivity and government spending in health 17 and education). This part of our work essentially provides non-results: none of the macro country-level variables we consider robustly correlates with changes in the income shares of the poorest quintiles. So, if we are interested in “shared prosperity�, we have both good news and bad news. The good news is that institutions and policies that promote economic growth in general will on average raise incomes of the poor equiproportionally, thereby promoting “shared prosperity�. The bad news is that, in choosing among macroeconomic policies, there is no robust evidence that certain policies are particularly “pro-poor� or conducive to promoting “shared prosperity� other than through their direct effects on overall economic growth. A final interesting puzzle is raised by the recent experiences of Latin America and Asia. In parsing the data by region and time period, there are almost no cases in which growth is significantly pro-poor or pro-rich. The exceptions are Latin America in the 2000s, in which income growth of the bottom 40 percent is 1.2 times mean growth; and Asia in the 1990s and 2000s, where income growth of the bottom 40 percent is only about 0.6 of mean growth. In both cases the coefficients are statistically different from 1.0. So, it would be interesting to understand better how Latin America achieved such inclusive growth while Asia is going in the opposite direction. At the same time it is important to keep in mind that growth of income of the bottom 40 percent has been much faster in Asia than in Latin America because the overall growth rate has been so much higher. References Balakrishnan Ravi, Chad Steinberg, and Murtaza Syed. (2013). “The Elusive Quest for Inclusive Growth: Growth, Poverty, and Inequality,� IMF Working Paper WP/13/152. Checchi, Daniele; Garcia-Penalosa, Cecilia. (2008). “Labour Market Institutions and Income Inequality�, Economic Policy issue 56, pp. 601-34, 640-49. Chen, Shaohua, and Martin Ravallion. (2008). “The Developing World Is Poorer Than We Thought, But No Less Successful in the Fight against Poverty,� World Bank Policy Research Working Paper 4703. Chen, Shaohua, and Martin Ravallion. (2010). “The Developing World Is Poorer Than We Thought, But No Less Successful in the Fight against Poverty,� The Quarterly Journal of Economics. Ciccone, Antonia and Marek Jacocinski. (2010). “Determinants of Economics Growth: Will Data Tell?,� American Economic Journal: Macroeconomics 2, 2:4, 222-246. 18 Datt, Gaurav and Martin Ravallion (2002). “Why Has Economic Growth Been More Pro-Poor in Some States of India than Others?�. Journal of Development Economics. 68: 381-400. Deaton, Angus, and Valerie Kozel. (2005). “Data and Dogma: The Great Indian Poverty Debate�, Oxford University Press 20:177-199. De Gregorio, Jose; Lee, Jong-Wha. (2002). “Education and Income Inequality: New Evidence from Cross- Country Data�, Review of Income and Wealth, v. 48, iss. 3, pp. 395-416. Dollar, David, and Aart Kraay. (2002). “Growth is Good for the Poor,� Journal of Economic Growth, 7, 195- 225. Feldkircher, M., and Zeugner, S. (2009). Benchmark Priors Revisited: On Adaptive Shrinkage and the Supermodel Effect in Bayesian Model Averaging. IMF Working Paper No. 09/202, International Monetary Fund. Fernandez, C., Ley, E., and Steel, M.F.J. (2001a). Model Uncertainty in Cross-Country Growth Regressions. Journal of Applied Econometrics, 16(5), 563-576. Fernandez, C., Ley, E., and Steel, M.F.J. (2001b). Benchmark prior for Bayesian model averaging. Journal of Econometrics, 100(2), 381-427. Jaumotte, Florence, Subir Lall, and Chris Papageorgiou. (2013). “Rising Income Inequality: Technology, or Trade and Financial Globalization,� IMF Economic Review, v. 61, no. 2, pp. 271-309. Klenow, Peter and Andres Rodriguez-Clare (1997). “The Neoclassical Revival in Macroeconomics – Has It Gone Too Far?�, in Ben Bernanke and Julio Rotemberg, eds. NBER Macroeconomics Annual. Cambridge, MIT Press, pp. 72-103. Luxembourg Income Study (LIS) Database (2013). http://www.lisdatacenter.org (multiple countries; May 2013). Luxembourg: LIS. Milanovic, Branko. (2000). “Determinants of Cross-Country Income Inequality: An 'Augmented' Kuznets Hypothesis�, Essays in honour of Branko Horvat, pp. 48-79. Pew Research Center (2013). “Economies of Emerging Markets Better Rated During Difficult Times�. http://www.pewglobal.org/2013/05/23/economies-of-emerging-markets-better-rated-during-difficult- times/. PovcalNet Database (2013). The on-line tool for poverty measurement developed by the Development Research Group of the World Bank, http://iresearch.worldbank.org/PovcalNet/index.htm; May 2013. Sala-i-Martin, Xavier, Gernot Doppelhofer, and Roland I. Miller. (2004). “Determinants of Long-Term Growth: A Bayesian Averaging of Classical Estimates (BACE) Approach.� American Economic Review, 94(4): 813-35. World Bank (2013). “The World Bank Goals: End Extreme Poverty and Promote Shared Prosperity�. http://www.worldbank.org/content/dam/Worldbank/document/WB-goals2013.pdf. 19 Figure 1: Availability of Household Survey Data (POVCALNET and LIS) 60 50 40 30 20 10 0 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 LIS Database (# obs) PCN Database (#obs) Notes: This figure shows number of household surveys available in each year, for the LIS and POVCALNET databases. 20 Figure 2: Growth rates of Incomes of Poorest 40 Percent (a) Sample of medium spell length (b) Samples of short, medium and long spells Notes: These figures show the correlation between growth in incomes of the poorest 40 percent and overall income growth. The top panel uses the sample of spells at least five years long. The bottom panel contrasts the findings in the three sets of spells: all available spells regardless of length, spells at least five years long, and the longest available spell for each country. 21 Figure 3: Comparison of National Accounts and Survey Mean Growth Rates I n c o m e G r o w t h in S u r v e y M e a n -. 2 -. 15 -. 1 -. 05 0 .0 5 .1 .1 5 .2 -.2 -.1 5 -.1 -.0 5 0 .0 5 . 1 .15 .2 C on su m p tio n G ro w t h in N A S d ata F itte d V a lue s 4 5 de gre e lin e d ln m e a n fyr Notes: This figure compares growth in real private consumption from the national accounts (horizontal axis) with household survey mean growth rates (vertical axis). Growth rates are average annual log differences. The sample consists of spells at least five years long. 22 Table 1: Descriptive Statistics Survey mean growth rate Growth rate in share (bottom 20%) Growth rate in share (bottom 40%) Std. Std. Corr with Std. Corr with Mean Nb obs Mean Mean deviation deviation mean deviation mean Panel 1 : Growth rates, sample pooled over time and regions All spells 0.020 0.081 735 0.004 0.071 -0.010 0.003 0.046 -0.105 Min-five-year spells 0.014 0.047 299 0.001 0.036 0.073 0.000 0.025 0.007 Long spells 0.018 0.028 118 0.005 0.025 -0.051 0.004 0.018 -0.103 Panel 2 : Growth rates by regions min-5-year-sample Europe & Central Asia 0.010 0.086 44 -0.007 0.034 0.291 -0.006 0.024 0.265 Latin America & Caribbean 0.009 0.045 66 0.006 0.045 0.030 0.004 0.028 -0.141 Middle East & North Africa 0.003 0.024 14 0.007 0.022 0.123 0.005 0.018 0.144 High Income 0.012 0.029 78 -0.002 0.030 0.172 -0.004 0.020 0.057 Sub-Saharan Africa 0.016 0.040 55 0.008 0.044 -0.012 0.005 0.034 -0.032 South Asia 0.020 0.014 17 -0.001 0.016 -0.203 -0.002 0.015 -0.147 East Asia and Pacific 0.034 0.034 25 -0.002 0.029 -0.499 -0.002 0.021 -0.542 Panel 3 : Growth rates by decades min-5-year-sample 1980-89 -0.003 0.049 86 0.003 0.034 0.067 0.002 0.027 0.012 1990-99 0.005 0.048 205 -0.003 0.037 0.087 -0.003 0.025 0.031 2000-10 0.030 0.040 174 0.004 0.034 -0.037 0.001 0.024 -0.093 Panel 4 : Growth rates by region and decades min-5-year-sample Europe & Centr. Asia 80-89 -0.122 0.086 8 -0.029 0.034 0.448 -0.020 0.023 0.584 Europe & Centr. Asia 90-99 -0.049 0.082 26 -0.015 0.038 0.219 -0.011 0.027 0.187 Europe & Centr. Asia 00-10 0.056 0.047 34 -0.001 0.030 0.082 -0.002 0.022 0.070 Latin America & Car. 80-89 0.003 0.054 18 0.016 0.045 -0.266 0.013 0.037 -0.376 Latin America & Car. 90-99 0.009 0.049 46 -0.008 0.045 -0.084 -0.005 0.028 -0.281 Latin America & Car. 00-10 0.012 0.037 35 0.019 0.040 0.398 0.011 0.020 0.348 High Income 80-89 -0.001 0.032 32 0.004 0.034 -0.059 0.002 0.026 -0.113 High Income 90-99 0.011 0.026 56 -0.009 0.025 0.322 -0.009 0.016 0.333 High Income 00-10 0.026 0.024 35 -0.004 0.016 -0.077 -0.005 0.012 -0.325 Middle East & Africa 80-89 -0.002 0.032 14 -0.006 0.032 0.210 -0.007 0.022 0.199 Middle East & Africa 90-99 0.009 0.036 50 0.016 0.042 0.079 0.012 0.030 0.087 Middle East & Africa 00-10 0.022 0.037 49 0.004 0.040 -0.115 0.001 0.032 -0.139 East and South Asia 80-89 0.018 0.028 14 0.004 0.013 -0.578 0.002 0.010 -0.340 East and South Asia 90-99 0.028 0.020 27 -0.009 0.017 -0.513 -0.007 0.014 -0.506 East and South Asia 00-10 0.036 0.034 21 0.002 0.034 -0.465 0.001 0.025 -0.526 Notes: This table reports descriptive statistics for growth rates in survey means and quintile shares. The first three columns report the mean, standard deviation, and number of spells. The next three columns report the mean and standard deviation of growth rates in the first quintile share, as well as its correlation with growth in average income. The last three columns provide the same information, but for the income share of the bottom 40 percent. Growth rates are calculated as average annual log differences over the length of each spell. Panel 1 combines all observations, for the three sets of spells. The remaining panels report results for sample splits by region, by decade, and by region-decade, only for the sample of spells at least five years long. See main text for description of how spells are assigned to decades. Note that in Panel 4 we combine Middle East North Africa and Sub-Saharan Africa into one group as well as East Asia and Pacific with South Asia due to small sample sizes within region-decade bins. 23 Table 2: Regression Results in the Benchmark Specification Dependent. var.: Growth in incomes of the poor (1) (2) (3) (1) (2) (3) Panel A: Bottom 20 percent Panel B: Bottom 40 percent Avg. growth - All spells 0.992*** 0.941*** (0.0509) (0.0367) Avg. growth - Min 5 year spells 1.057*** 1.004*** (0.0572) (0.0435) Avg. growth - Long spells 0.955*** 0.932*** (0.118) (0.0798) Number of Observations 735 299 118 735 299 118 Number of Countries 123 117 118 123 117 118 R-squared 0.557 0.653 0.533 0.734 0.776 0.666 Share of variance due to growth 0.562 0.618 0.558 0.780 0.773 0.714 P-value of wald test, slope=1 0.874 0.324 0.704 0.111 0.933 0.396 Notes: *** (**) (*) denotes significance at the 1 (5) (10) percent level. Heteroskedasticity-consistent standard errors clustered at the country level reported in parentheses. This table reports results from OLS regressions of growth in incomes of the poor on growth in average incomes. Growth rates are calculated as average annual log differences over the indicated definitions of spells. Columns (1)-(3) define the poor as those in bottom 20 percent of income distribution, while Columns (4)-(6) refer to bottom 40 percent of income distribution. In addition to the regular regression outputs, we document the variance decomposition which summarizes the part of the variation in income of the poor that is due to variation in overall incomes. We also report the p-value corresponding to a Wald test of the null hypothesis that the estimated slope is equal to one. 24 Table 3: Results by Region (1) (2) (3) (4) (5) (6) (7) Europe & Latin Middle Sub- Dependent. var.: Growth in High East Asia Central America & East & Saharan South Asia income of the poor Income and Pacific Asia Caribbean North Africa Panel A: Bottom 20 percent Avg. growth -Min- 5yr-spells 1.113*** 1.030*** 1.112*** 1.180*** 0.986*** 0.772*** 0.569** (0.0580) (0.147) (0.130) (0.186) (0.166) (0.137) (0.196) Number of Observations 44 66 14 78 55 17 25 R-squared 0.900 0.523 0.601 0.567 0.441 0.329 0.367 Share of variance due to growth 0.808 0.508 0.540 0.480 0.447 0.426 0.644 P-val. wald test, slope=1 0.0663 0.841 0.427 0.343 0.934 0.170 0.0556 Panel B: Bottom 40 percent Avg. growth -Min- 5yr-spells 1.074*** 0.915*** 1.110*** 1.039*** 0.972*** 0.844*** 0.662*** (0.0403) (0.104) (0.100) (0.138) (0.120) (0.153) (0.137) Number of Observations 44 66 14 78 55 17 25 R-squared 0.940 0.694 0.685 0.698 0.566 0.391 0.614 Share of variance due to growth 0.875 0.759 0.617 0.672 0.582 0.463 0.928 P-val. wald test, slope=1 0.0804 0.423 0.325 0.778 0.819 0.364 0.0354 Number of Countries 20 21 6 27 28 5 10 Standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1 Notes: *** (**) (*) denotes significance at the 1 (5) (10) percent level. Heteroskedasticity-consistent standard errors clustered at the country level reported in parentheses. This table reports results from OLS regressions of growth in incomes of the poor on growth in average incomes. Growth rates are calculated as average annual log differences over the indicated definitions of spells. Panel A defines the poor as those in bottom 20 percent of income distribution, while Panel B refers to bottom 40 percent of income distribution. In addition to the regular regression outputs, we document the variance decomposition which summarizes the part of the variation in income of the poor that is due to variation in overall incomes. We also report the p-value corresponding to a Wald test of the null hypothesis that the estimated slope is equal to one. The assignment of countries to geographical regions is documented in Appendix Table A1. 25 Table 4: Results Across Regions and Over Time Dependent. var.: Europe and Latin America and the High Income Countries Middle East and East Asia, Pacific and All regions Growth in income of Central Asia Caribbean (from all regions) Sub-Saharan Africa South Asia the poor 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 1980 1990 2000 Panel A: Bottom 20 percent Avg. growth by decade 1.046*** 1.067*** 0.969*** 1.176*** 1.101*** 1.052*** 0.781*** 0.924*** 1.422*** 0.936*** 1.316*** 0.948*** 1.207*** 1.090*** 0.878*** 0.722*** 0.574*** 0.546* (0.0881) (0.0647) (0.0834) (0.0960) (0.0701) (0.0807) (0.209) (0.131) (0.158) (0.310) (0.137) (0.115) (0.314) (0.182) (0.214) (0.0913) (0.143) (0.276) Number of Observations 86 205 174 8 26 34 18 46 35 32 56 35 14 50 49 14 27 21 R-squared 0.695 0.659 0.565 0.918 0.858 0.735 0.493 0.508 0.681 0.428 0.667 0.663 0.609 0.477 0.412 0.773 0.394 0.284 Share of variance due to growth 0.664 0.618 0.583 0.781 0.780 0.699 0.632 0.550 0.479 0.457 0.507 0.699 0.505 0.438 0.469 1.070 0.686 0.521 P-val. Wald test, slope=1 0.600 0.303 0.710 0.109 0.170 0.527 0.314 0.569 0.0158 0.838 0.0287 0.653 0.520 0.624 0.575 0.0161 0.0117 0.122 Panel B: Bottom 40 percent Avg. growth by decade 1.006*** 1.017*** 0.943*** 1.154*** 1.061*** 1.033*** 0.745*** 0.841*** 1.191*** 0.907*** 1.212*** 0.831*** 1.138*** 1.073*** 0.880*** 0.875*** 0.659*** 0.616*** (0.0709) (0.0508) (0.0588) (0.0639) (0.0493) (0.0626) (0.172) (0.0927) (0.0783) (0.247) (0.114) (0.0854) (0.229) (0.126) (0.156) (0.0855) (0.127) (0.177) Number of Observations 86 205 174 8 26 34 18 46 35 32 56 35 14 50 49 14 27 21 R-squared 0.777 0.787 0.711 0.967 0.917 0.827 0.584 0.705 0.844 0.549 0.803 0.741 0.736 0.622 0.516 0.864 0.564 0.497 Share of variance due to growth 0.772 0.774 0.754 0.838 0.864 0.801 0.784 0.839 0.709 0.605 0.663 0.891 0.647 0.580 0.586 0.988 0.855 0.806 P-val. Wald test, slope=1 0.930 0.742 0.334 0.0470 0.235 0.603 0.163 0.102 0.0255 0.710 0.0755 0.0595 0.558 0.566 0.449 0.181 0.0203 0.0484 Notes: *** (**) (*) denotes significance at the 1 (5) (10) percent level. Heteroskedasticity-consistent standard errors clustered at the country level reported in parentheses. This table reports results from weighted OLS regressions of growth in incomes of the poor on growth in average incomes, in the indicated region-decade bins, with weights corresponding to the fraction of observations in each spell falling in the indicated decade. Growth rates are calculated as average annual log differences over spells at least five years long. Panel A defines the poor as those in bottom 20 percent of income distribution, while Panel B refers to bottom 40 percent of income distribution. In addition to the regular regression outputs, we document the variance decomposition which summarizes the part of the variation in income of the poor that is due to variation in overall incomes. We also report the p-value corresponding to a Wald test of the null hypothesis that the estimated slope is equal to one. Table 5: Robustness Across Alternative Measures of Average Growth Survey-based National Accounts Mixed Measure (1) (2) (3) (4) (5) (6) (7) (8) (9) Dependent. var.: Growth in Survey-based welfare measure Real private consumption per capita Mixing survey-based and national income of the poor (income or consumption) (national accounts data) accounts' welfare measures Panel A: Bottom 20 percent Avg. growth - All spells 0.979*** 1.009*** 0.983*** (0.0534) (0.0499) (0.0616) Avg. growth - Min 5 year spells 0.971*** 1.109*** 1.036*** (0.0627) (0.0536) (0.0657) Avg. growth - Longest spells 0.854*** 0.935*** 0.856*** (0.114) (0.0973) (0.128) Number of Observations 710 282 106 710 282 106 710 282 106 R-squared 0.546 0.593 0.510 0.351 0.577 0.552 0.382 0.526 0.434 Share of variance due to growth 0.558 0.610 0.597 0.348 0.520 0.591 0.388 0.508 0.507 P-value of wald test, slope=1 0.689 0.649 0.202 0.858 0.0445 0.503 0.779 0.581 0.264 Panel B: Bottom 40 percent Avg. growth - All spells 0.930*** 1.009*** 0.935*** (0.0384) (0.0373) (0.0456) Avg. growth - Min 5 year spells 0.939*** 1.064*** 0.989*** (0.0477) (0.0356) (0.0488) Avg. growth - Longest spells 0.863*** 0.942*** 0.868*** (0.0758) (0.0700) (0.0855) Number of Observations 710 282 106 710 282 106 710 282 106 R-squared 0.727 0.737 0.655 0.566 0.719 0.688 0.576 0.673 0.582 Share of variance due to growth 0.781 0.785 0.759 0.562 0.675 0.730 0.616 0.681 0.671 P-value of wald test, slope=1 0.0731 0.203 0.0742 0.819 0.0730 0.408 0.159 0.816 0.124 Notes: *** (**) (*) denotes significance at the 1 (5) (10) percent level. Heteroskedasticity-consistent standard errors clustered at the country level reported in parentheses. This table reports results from OLS regressions of growth in incomes of the poor on growth in average incomes. Growth rates are calculated as average annual log differences over the indicated definitions of spells. Panel A defines the poor as those in bottom 20 percent of income distribution, while Panel B refers to the bottom 40 percent of income distribution. Columns 1-3 use household survey means, in the slightly smaller sample of spells where national accounts growth rates are also available. Columns 4-6 use national accounts growth rates as a measure of average income growth and to construct average income growth of the poor. Columns 7-9 use a simple average of survey mean and national accounts growth rates. We also report the p-value corresponding to a Wald test of the null hypothesis that the estimated slope is equal to one. Table 6: Bayesian Model Averaging Results (Bottom 20 Percent) Dependent Variable: Income Significance of Estimated Growth Bottom 20% Distribution Of Estimated Slopes Slopes BMA Post. Inclusion Probability Expected slope Min. 5th perc. Median 95th perc. Max. Signif > 0 Signif < 0 prob. weighted slope cond. on incl. ∆ Average income 0.905 0.949 1.010 1.062 1.096 100.0% 0.0% 1.000 1.056 1.056 ∆ Financial depth (M2 % GDP) -0.002 -0.002 -0.001 0.000 0.001 0.0% 0.0% 0.000 0.000 -0.001 ∆ Inflation rate -0.071 0.057 0.198 0.450 0.547 38.9% 0.0% 0.000 0.000 -0.026 ∆ Budget Balance -0.196 -0.053 0.116 0.340 0.462 0.0% 0.0% 0.000 0.000 0.141 ∆ Trade Openness 0.019 0.043 0.062 0.101 0.131 3.8% 0.0% 0.000 0.000 0.039 ∆ Population growth -0.021 -0.002 0.015 0.046 0.084 0.1% 0.0% 0.053 0.001 0.020 ∆ Life expectancy -0.037 -0.029 -0.015 -0.008 0.000 0.0% 25.6% 0.032 0.000 -0.002 ∆ Assassinations per pop. -0.130 -0.101 0.019 0.093 0.148 0.0% 0.0% 0.000 0.000 -0.076 ∆ Revolutions per pop. -0.015 0.006 0.071 0.111 0.140 0.0% 0.0% 0.000 0.000 0.014 ∆ Civil Liberties / Democracy -0.016 -0.010 -0.004 0.002 0.009 0.0% 0.0% 0.000 0.000 0.000 ∆ Internal conflict (dummy) -0.014 0.010 0.039 0.067 0.087 0.0% 0.0% 0.035 0.001 0.024 ∆ War participation (dummy) -0.162 -0.127 -0.083 -0.010 0.035 0.0% 0.0% 0.032 0.000 0.004 ∆ Fin. openness (Chinn-Ito) -0.010 -0.003 0.005 0.015 0.024 0.0% 0.0% 0.000 0.000 0.003 ∆ Primary school enrollment rate -0.003 -0.002 -0.001 0.000 0.001 0.0% 0.0% 0.000 0.000 0.000 ∆ Education Gini -0.869 -0.546 -0.265 0.141 0.560 0.0% 0.0% 0.000 0.000 -0.624 ∆ Gov Expend Educ (% GDP) -0.043 -0.029 -0.014 -0.001 0.009 0.0% 0.2% 0.000 0.000 -0.014 ∆ Gov. Expend Health (% GDP) -0.006 0.000 0.010 0.023 0.030 0.2% 0.0% 0.000 0.000 0.002 ∆ Agriculture (% GDP) 0.067 0.102 0.138 0.187 0.228 29.0% 0.0% 0.000 0.000 0.154 Distribution of Sample Size 113 122 164 234 299 Distribution of R-squared 0.487 0.525 0.569 0.629 0.676 Notes: This table summarizes the results of the Bayesian Model Averaging exercise described in Section 4 of the paper. The first five columns summarize the distribution of the estimated slope coefficients across the 131,072 regression models defined by all possible combinations of the seventeen control variables listed in the first column. The next two columns report the fraction of estimated slope coefficients significantly greater (less than) zero across all models. The posterior inclusion probability is the sum of the posterior probabilities of all models including the indicated variable. The probability-weighted slope coefficient is the expected value of the slopes, weighting each by the posterior probability of the corresponding model in which it was estimated, and treating the estimated slope as zero in those models in which it is not included. The last column reports the same information, but conditional on the variable being included. Table 7: Bayesian Model Averaging Results (Bottom 40 Percent) Dependent Variable: Income Significance of Estimated Growth Bottom 40% Distribution Of Estimated Slopes Slopes BMA Post. Inclusion Probability Expected slope Min. 5th perc. Median 95th perc. Max. Signif > 0 Signif < 0 prob. weighted slope cond. on incl. ∆ Average income 0.877 0.915 0.963 1.006 1.031 100.0% 0.0% 1.000 1.003 1.003 ∆ Financial depth (M2 % GDP) -0.002 -0.001 0.000 0.000 0.001 0.0% 0.0% 0.000 0.000 0.000 ∆ Inflation rate -0.046 0.044 0.137 0.275 0.338 32.1% 0.0% 0.000 0.000 -0.017 ∆ Budget Balance -0.193 -0.080 0.019 0.112 0.183 0.0% 0.0% 0.000 0.000 0.031 ∆ Trade Openness 0.010 0.025 0.039 0.062 0.083 1.5% 0.0% 0.000 0.000 0.027 ∆ Population growth -0.024 -0.009 0.004 0.027 0.053 0.0% 0.0% 0.067 0.001 0.017 ∆ Life expectancy -0.028 -0.023 -0.012 -0.006 0.002 0.0% 41.5% 0.032 0.000 -0.001 ∆ Assassinations per pop. -0.058 -0.033 0.025 0.071 0.108 0.0% 0.0% 0.000 0.000 -0.013 ∆ Revolutions per pop. 0.058 0.077 0.122 0.145 0.170 0.1% 0.0% 0.000 0.000 0.086 ∆ Civil Liberties / Democracy -0.005 -0.001 0.003 0.008 0.011 0.0% 0.0% 0.000 0.000 0.004 ∆ Internal conflict (dummy) 0.016 0.029 0.046 0.065 0.072 2.5% 0.0% 0.082 0.003 0.041 ∆ War participation (dummy) -0.078 -0.050 -0.029 0.020 0.051 0.0% 0.0% 0.036 0.001 0.022 ∆ Fin. openness (Chinn-Ito) -0.007 -0.002 0.004 0.011 0.017 0.0% 0.0% 0.000 0.000 0.004 ∆ Primary school enrollment rate -0.002 -0.001 -0.001 0.000 0.001 0.0% 0.0% 0.000 0.000 0.000 ∆ Education Gini -0.644 -0.430 -0.246 0.044 0.282 0.0% 0.0% 0.000 0.000 -0.515 ∆ Gov Expend Educ (% GDP) -0.028 -0.020 -0.009 0.002 0.010 0.0% 0.4% 0.000 0.000 -0.008 ∆ Gov. Expend Health (% GDP) -0.005 0.000 0.007 0.016 0.022 0.4% 0.0% 0.000 0.000 0.002 ∆ Agriculture (% GDP) 0.041 0.059 0.088 0.122 0.155 10.9% 0.0% 0.000 0.000 0.096 Distribution of Sample Size 113 122 164 234 299 Distribution of R-squared 0.633 0.671 0.709 0.756 0.790 Notes: This table summarizes the results of the Bayesian Model Averaging exercise described in Section 4 of the paper. The first five columns summarize the distribution of the estimated slope coefficients across the 131,072 regression models defined by all possible combinations of the seventeen control variables listed in the first column. The next two columns report the fraction of estimated slope coefficients significantly greater (less than) zero across all models. The posterior inclusion probability is the sum of the posterior probabilities of all models including the indicated variable. The probability-weighted slope coefficient is the expected value of the slopes, weighting each by the posterior probability of the corresponding model in which it was estimated, and treating the estimated slope as zero in those models in which it is not included. The last column reports the same information, but conditional on the variable being included. 29 Appendix: Table A1: Data availability by country Sample min- Sample Total First year Last year Sample all Country Region Database 5-year- longest observations available avail spells (diff.) spells (diff.) spell (diff.)* Albania ECA PCN 5 1997 2008 4 2 1 Algeria MENA PCN 2 1988 1995 1 1 1 Argentina LAC PCN 22 1986 2010 20 4 1 Armenia ECA PCN 10 1996 2008 8 1 1 Australia HIINC LIS 6 1981 2003 5 3 1 Austria HIINC LIS 6 1987 2004 5 2 1 Azerbaijan ECA PCN 3 1995 2008 2 2 1 Bangladesh SA PCN 8 1984 2010 7 4 1 Belarus ECA PCN 12 1988 2008 7 3 1 Belgium HIINC LIS 6 1985 2000 5 2 1 Belize LAC PCN 7 1993 1999 5 1 1 Bhutan SA PCN 2 2003 2007 1 1 Bolivia LAC PCN 11 1991 2008 8 2 1 Bosnia and Herzegovina ECA PCN 3 2001 2007 2 1 1 Botswana SSA PCN 2 1986 1994 1 1 1 Brazil LAC PCN 26 1981 2009 25 5 1 Bulgaria ECA PCN 8 1989 2007 6 3 1 Burkina Faso SSA PCN 4 1994 2009 3 2 1 Burundi SSA PCN 3 1992 2006 2 2 1 Cambodia EAP PCN 4 1994 2008 3 1 1 Cameroon SSA PCN 3 1996 2007 2 2 1 Canada HIINC LIS 11 1971 2007 10 5 1 Central African Republic SSA PCN 3 1992 2008 2 2 1 Chile LAC PCN 10 1987 2009 9 4 1 China EAP PCN 9 1981 2005 7 3 1 Colombia LAC PCN 12 1992 2010 11 3 1 Costa Rica LAC PCN 23 1981 2009 22 5 1 Cote d'Ivoire SSA PCN 9 1985 2008 8 3 1 Croatia HIINC PCN 7 1988 2008 5 1 1 Czech Republic HIINC PCN 3 1988 1996 2 1 1 Denmark HIINC LIS 5 1987 2004 4 2 1 Dominican Republic LAC PCN 16 1986 2010 15 4 1 Ecuador LAC PCN 13 1987 2010 10 4 1 Egypt, Arab Rep. MENA PCN 5 1991 2008 4 2 1 El Salvador LAC PCN 15 1989 2009 13 2 1 Estonia HIINC PCN 9 1988 2004 7 2 1 Ethiopia SSA PCN 4 1982 2005 3 3 1 Fiji EAP PCN 2 2003 2009 1 1 1 Finland HIINC LIS 5 1987 2004 4 2 1 France HIINC LIS 7 1979 2005 5 5 1 Gambia, The SSA PCN 2 1998 2003 1 Georgia ECA PCN 12 1996 2008 10 2 1 Germany HIINC LIS 5 1994 2010 4 2 1 Ghana SSA PCN 5 1988 2006 4 2 1 Greece HIINC LIS 5 1995 2010 4 2 1 Guatemala LAC PCN 8 1987 2006 6 2 1 Guinea SSA PCN 4 1991 2007 2 1 1 Guinea-Bissau SSA PCN 3 1991 2002 1 1 1 Guyana LAC PCN 2 1993 1998 1 1 1 Honduras LAC PCN 20 1989 2009 14 4 1 Hungary HIINC PCN 10 1987 2007 7 2 1 India SA PCN 5 1978 2005 4 4 1 Indonesia EAP PCN 8 1984 2005 7 3 1 Iran, Islamic Rep. MENA PCN 5 1986 2005 4 2 1 Ireland HIINC LIS 6 1987 2004 5 2 1 Israel HIINC LIS 6 1986 2007 5 3 1 Italy HIINC LIS 11 1986 2010 10 4 1 Jamaica LAC PCN 7 1988 2004 6 3 1 Jordan MENA PCN 7 1987 2010 6 4 1 Kazakhstan ECA PCN 11 1988 2009 9 3 1 Kenya SSA PCN 4 1992 2005 2 1 1 Sample min- Sample Total First year Last year Sample all Country Region Database 5-year- longest observations available avail spells (diff.) spells (diff.) spell (diff.)* Kyrgyz Republic ECA PCN 10 1988 2009 8 2 1 Lao PDR EAP PCN 4 1992 2008 2 2 1 Latvia ECA PCN 11 1988 2008 9 3 1 Lesotho SSA PCN 4 1987 2003 2 2 1 Lithuania ECA PCN 9 1988 2008 7 3 1 Luxembourg HIINC LIS 6 1985 2004 5 3 1 Macedonia, FYR ECA PCN 9 1998 2009 8 2 1 Madagascar SSA PCN 7 1980 2010 6 4 1 Malawi SSA PCN 2 1998 2004 1 1 1 Malaysia EAP PCN 9 1984 2009 8 4 1 Maldives SA PCN 2 1998 2004 1 Mali SSA PCN 4 1994 2010 3 2 1 Mauritania SSA PCN 6 1987 2008 5 3 1 Mexico LAC PCN 13 1984 2010 10 3 1 Moldova ECA PCN 15 1988 2010 11 2 1 Montenegro ECA PCN 4 2005 2008 3 Mozambique SSA PCN 3 1996 2008 2 2 1 Namibia SSA PCN 2 1993 2004 1 1 1 Nepal SA PCN 4 1985 2010 2 2 1 Netherlands HIINC LIS 6 1983 2004 5 3 1 Nicaragua LAC PCN 4 1993 2005 3 2 1 Niger SSA PCN 4 1992 2008 3 1 1 Nigeria SSA PCN 5 1986 2010 4 3 1 Norway HIINC LIS 6 1979 2004 5 3 1 Pakistan SA PCN 8 1987 2008 7 3 1 Panama LAC PCN 14 1979 2010 12 3 1 Paraguay LAC PCN 14 1990 2010 13 2 1 Peru LAC PCN 16 1986 2010 14 3 1 Philippines EAP PCN 9 1985 2009 8 4 1 Poland HIINC PCN 17 1985 2009 14 4 1 Romania ECA PCN 14 1989 2009 11 2 1 Russian Federation ECA PCN 13 1988 2009 11 3 1 Rwanda SSA PCN 4 1985 2011 3 3 1 Senegal SSA PCN 4 1991 2005 3 1 1 Serbia ECA PCN 8 2002 2009 6 1 1 Seychelles SSA PCN 2 2000 2007 1 1 Slovak Republic HIINC PCN 9 1988 2009 7 2 1 Slovenia HIINC PCN 6 1987 2004 4 2 1 South Africa SSA PCN 5 1993 2009 4 2 1 Spain HIINC LIS 7 1980 2010 6 4 1 Sri Lanka SA PCN 5 1985 2007 4 4 1 Swaziland SSA PCN 3 1995 2010 2 2 1 Sweden HIINC LIS 8 1967 2005 7 6 1 Switzerland HIINC LIS 5 1982 2004 4 2 1 Tajikistan ECA PCN 5 1999 2009 4 2 1 Tanzania SSA PCN 3 1992 2007 2 2 1 Thailand EAP PCN 13 1981 2009 12 4 1 Timor-Leste EAP PCN 2 2001 2007 1 1 1 Trinidad and Tobago HIINC PCN 2 1988 1992 1 1 Tunisia MENA PCN 5 1985 2005 4 4 1 Turkey ECA PCN 9 1987 2008 8 3 1 Turkmenistan ECA PCN 3 1988 1998 1 1 Uganda SSA PCN 7 1989 2009 6 3 1 Ukraine ECA PCN 13 1988 2009 11 3 1 United Kingdom HIINC LIS 7 1991 2010 6 3 1 United States HIINC LIS 10 1974 2010 9 6 1 Uruguay LAC PCN 18 1981 2010 17 5 1 Venezuela, RB LAC PCN 13 1981 2006 11 4 1 Vietnam EAP PCN 6 1993 2008 5 2 1 West Bank and Gaza MENA PCN 2 2007 2009 1 1 Yemen, Rep. MENA PCN 2 1998 2005 1 1 1 Zambia SSA PCN 6 1993 2006 4 2 1 Notes: Region codes refer to World Bank categories with the exception that all High income countries were pooled by pulling observations from the geographical regions: HIINC= High Income countries, ECA= Europe and Central Asia, MENA= Middle East & North Africa, LAC = Latin America and the Caribbeans, SSA=Sub-Saharan Africa, SA= South Asia and EAP=East Asia and Pacific. Database indicates whether the data come from POVCALNET (PCN) or LIS. Total observations, first year, and last year refer to the number and timing of household surveys in our combined dataset. The last three columns indicate the number of spells included in each of the three definitions of spells. Note that these spells refer to the sample used in the regression, following the removal of extreme observations as noted in the text. This is why there are some blank entries in the last three columns. 31 Table A2: Explanation of control variables Variable Source Description / Adjustments Survey means POVCALNET, POVCALNET measures welfare by income or consumption as determined in the LIS surveys. For LIS, we calculate survey means of disposable income directly from the micro survey data on household level. Household per WDI Household final consumption expenditure (constant LCU) divided by population. capita consumption Covariates used in Bayesian Model Averaging: Population WDI Population growth in percentage points growth Life expectancy WDI Life expectancy in years Financial depth; WDI Money and quasi-money (M2) as percent of GDP M2 as % of GDP Inflation rate WDI Inflation measure is calculated by taking log-differences from the WDI reported GDP deflator (local currency units). Budget balance WEO and data Data series on Budget Balance from Easterly, Levine, Roodman (2004) was used from Easterly, when available, after last available year, used WEO data. Levine, Roodman (2004) Assassination; Cross-National Assassinations and revolutions as percentage per 100,000 habitants. Revolution Time Series Source: Banks, Arthur S., Wilson, Kenneth A. 2013. Cross-National Time-Series Data Archive. Databanks International. Jerusalem, Israel; see http://www.databanksinternational.com Trade Openness Wacziarg- Wacziarg-Welch (2008) extension of the initial Sachs-Warner (1995) openness Welch (2008); measure is available through 2001. We update the series to 2010 using underlying extended data on tariffs, black market premium and export marketing boards. A country is through 2010. considered as closed if it has one of the following: Average tariff rates over 40 percent, black market exchange rate over 20 percent lower than the official http://www.a exchange rate, or a state monopoly on major exports (export marketing board). nderson.ucla.e 1. Tariffs: (Francis K.T. Ng “Trends in average applied tariff rates in developing and du/faculty_pa industrial countries, 1980-2006�; http://go.worldbnka.org/LGOXFTV550). No ges/romain.wa countries had tariffs beyond the 40 percent threshold at any time after 2000. cziarg/papersu 2. Black market premium: (Economic Freedom in the World 2012 report and m.html database from the Fraser Institute (http://www.freetheworld.com)). Data reports a 0-10 ranking where 10 implies no black market premium and 0 implies a premium of 50 percent or more. The black market premium is defined as the percentage difference between the official and the black market exchange rate. We assume that a score of 0-6 implies a premium of 20 percent or greater. 3. Export marketing board: In 2001 Wacziarg-Welch identified 12 countries as having an export marketing board based on various underlying data and sources. Clemens et al. update the classification through 2005, identifying three further countries has having liberalized or abolished their export marketing boards (Senegal (2002), Chad and Papua New Guinea (2005)). In our update we assume that none of the remaining 9 countries (Central African Rep, Congo Dem. Rep, Congo Rep., Gabon, Russia, Togo, Ukraine) abolished or liberalized their export marketing board through 2010. As neither of these countries have tariffs over 40 percent or black market premiums over 20 percent, they would be considered 32 “open� when liberalizing their export marketing board. Internal conflict; UCDP-PRIO Data from UCDP dataset allows constructing one dummy for internal conflict and war participation Dataset one for war participation. In the latter, we consider a country to be participating in a war only if it is listed either as the country of location, or a major participant (side A or B), omitting countries that are listed as allies. Civil liberties, Freedom Sum of the civil liberties and the political rights indicator, both measured on a 1-7 political rights House scale. http://www.freedomhouse.org/report/freedom-world-2012/methodology Financial Chinn-Ito The Chinn-Ito index (KAOPEN) is an index measuring a country's degree of capital Openness Index account openness. KAOPEN is based on the binary dummy variables that codify the tabulation of restrictions on cross-border financial transactions reported in the IMF's Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER). http://web.pdx.edu/~ito/Chinn-Ito_website.htm Primary WDI Gross primary school enrollment rates (percent of population) schooling Gini coefficient Barro-Lee The Barro-Lee dataset provides data on the percentage of the population that on educational dataset attained different levels of education: No education (0 years), complete primary (6 attainment years), complete secondary (12 years), and complete tertiary (16 years). For non- complete primary, secondary, or tertiary we assume respectively 3 years, 9 years, and 14 years of schooling. With this information, we can construct a Lorenz curve measuring which percentage of population attained which percentage of total years of schooling. With this information, we construct a Gini coefficient that measures educational inequality analogous to the standard income inequality measure. Government IMF social Government expenditure on health and education is retrieved from various expenditure on spending data, sources. We prioritize the data from Nozaki et al. (2011), we use WDI data for health and WDI, IMF GFS countries where the WDI coverage is better than the former, and as a third source education we use the IMF Government Finance Statistics (GFS) for countries where this (percent of GDP) source offers the best coverage. We merge data sources only across not within countries. Source: Nozaki Masahiro, Clements, Benedict and Gupta, Sanjeev. (2011). “What Happens to Social Spending in IMF-Supported Programs?�. http://www.imf.org/external/pubs/cat/longres.aspx?sk=25190.0 Agricultural WDI WDI Indicator: NV.AGR.TOTL.ZS, “Agriculture, value added (% of GDP)�. productivity Constructing the log-difference provides a measure of change in agricultural productivity. 33