The Quest for Pro-poor and Inclusive Growth: The Role of Governance

The Quest for Pro-poor and Inclusive Growth: The Role of Governance Djeneba DOUMBIA Paris School of Economics (PSE) Université Paris 1 Panthéon-Sorbonne E-mail : djeneba.doumbia@psemail.eu [Draft; please do not circulate] 1

Table of Contents Abstract.3 I. Introduction... 4 II. Governance, Growth, Poverty, and Income Distribution: A review of literature... 5 2.1. Growth, Poverty, and Income Distribution... 5 2.2. Governance and Pro-poor Growth... 7 III. Data and Trends in Poverty and Income Distribution... 8 3.1. Data... 8 3.1.1. Measuring Poverty and Inequality... 8 3.1.2. Defining and Measuring Governance... 10 3.2. Trends in Poverty and Income Distribution... 11 IV. Econometric Methodology... 16 V. Estimation Results... 18 5.1. Pro-poor Growth Regressions... 18 5.2. Inclusive Growth Regressions... 26 5.3. What determines how Pro-poor and Inclusive Growth is?... 32 5.4. Evidence of linearity or non-linearity... 38 VI. Panel Smooth Transition Regression: Robustness Check... 42 6.1. Estimation of Model (i)... 42 6.2. Estimation of Model (ii)... 44 VII. Conclusion and Discussion... 47 Appendix 49 2

Abstract This paper assesses the role of good governance in fostering pro-poor and inclusive growth. It relies on the Generalized Method of Moments in System (SGMM) following Arrellano and Bover (1995) and the Panel Smooth Transition Regression (PSTR) following Gonzalez et al. (2005) using a sample of 113 countries over the period 1975-2012. The main results support that growth is in general pro-poor per capita income growth reduces poverty. However, growth has not been inclusive at the global level as illustrated by a decline in the bottom quintile share of the income distribution. While good governance supports income growth and reduces poverty, it does less regarding inclusive growth. Investigating the determinants of pro-poor and inclusive growth highlights that health and education strategies, infrastructure improvement and financial development factors are the key factors for poverty reduction and inclusive growth. Key words: Poverty reduction, Pro-poor growth, Inclusive growth, Governance JEL Classification: C23, G28, H5, O11, O15, O57 3

I. Introduction Poverty remains widespread, particularly in developing countries, notwithstanding recent progress. While the aggregate worldwide poverty rate was reduced by about half between 1990 and 2010 mainly thanks to robust growth, the World Bank estimated that more than 1.22 billion people lived with less than $1.25 a day in 2010. To contrast the encouraging dynamic in poverty reduction, income inequality has risen across the world over the last two decades. How does these two divergent dynamics impact the income opportunities of the less fortunate, namely the poorest 20 percent of the population? This is an important policy question that has led to the development of new concepts for pro-poor growth and increased focus on income distribution with new studies on inclusive growth. Numerous empirical and statistical studies have identified economic growth as one of the main factors affecting poverty reduction (Dollar and Kraay (2002), Dollar, Kleineberg and Kraay (2013)). Moreover, there is a growing understanding that economic, political, legal and social institutions are critical for economic prosperity. Since the 1990s the concept of good governance has become central in the discussion and design of development policies. Since both governance and pro-poor growth are important in development policies agenda, the question arises as to whether and how they are related to each other. This paper provides a cross-country analysis investigating the role of economic growth in poverty reduction but adds two main contributions to the existing literature. First, it contributes to the recent and growing literature on inclusive growth by assessing how pro-poor and inclusive growth has been in different regions of the world. Second, it investigates the main structural factors that impact inclusive growth with a particular attention to an important channel that has received little attention so far: the quality of governance. The analysis could therefore shed some light on the role of governance in making growth more propoor and inclusive. Following Ravallion and Chen (2003), this paper defines growth as pro-poor simply if it reduces poverty or increases the income of the poor while inclusive growth refers to growth which is not associated with an increase in inequality (Rauniyar and Kanbur, 2010). The paper relies on panel fixed effect estimation and the Generalized Moments Method in System (SGMM) following Arrellano and Bover (1995). This method attempts to address endogeneity issues 4

related to potentially endogenous explanatory variables. Because of a possible non-linear relationship between poverty reduction and governance, through the level of economic development, a second empirical method in the study relies on the Panel Smooth Transition Regression (PSTR), following Gonzalez et al. (2005). The PSTR considers the speed of transition from one regime to the other, with the transition between the two regimes assumed to be gradual. The empirical analysis is based on a sample of 113 developed and developing countries covering all regions in the world (Europe and Central Asia, East Asia and Pacific, Middle East and North Africa, South Asia, Sub-Sahara, Latin America and Caribbean). The main findings are that (i) in general growth is pro-poor: the income of the poorest 20 percent increases with per capita income growth; (ii) the combination of political, economic and institutional features of good governance increases the income of the poor and reduces poverty; (iii) globally growth has been non-inclusive, but this result depends upon econometric specifications and still good governance leads to an increase in the bottom quintile share of the income distribution; (iv) when investigating what determines how pro-poor and inclusive growth is, the results suggest that health expenditure, spending in education, infrastructure improvement and some financial development factors are the key factors for poverty reduction. The rest of the paper is structured as follows. Section II reviews the literature analyzing the relationship between growth, poverty, income distribution, and governance. Section III describes that the data used in the study and discusses main trends in poverty and income distribution. Section IV explores the econometric methodology. Section V explains estimations results. Section VI provides robustness check and section VII concludes. II. Governance, Growth, Poverty, and Income Distribution: A review of literature 2.1. Growth, Poverty, and Income Distribution This section discusses cross-country analyses that investigate the relationship between growth, poverty and income distribution. An effective pro-poor growth policy may not have to only concentrate on economic growth, but it may also be combined with effective policy of income redistribution. Hence, the link between growth and inequality are important from a policy standpoint. It has been extremely debated since the 1950s. 5

Kuznets (1955) tries to answer two broad questions: Does inequality in the distribution of income increase or decrease in the course of a country s economic growth? What factors determine the secular level and trends of income inequalities? Kuznets explored the relationship between per capita income and inequality in a cross-section of countries. His findings display an inverted-u pattern that is, first inequality increased, and then declined, as per capita income augmented. Many other studies tested the hypothesized relationship found by Kuznets. Using a more accurate dataset than previous studies as well as individual countries data, Deininger and Squire (1998) found no evidence of an inverted-u relationship between per capita income and inequality in their sample. In addition, the authors did not find any systematic evidence in favour of a relationship between rapid growth and increasing inequality. Fast growth was associated with declining inequality as often as it was related to increasing inequality, or related to no changes at all. Similarly, Ravallion and Chen (1997) found that changes in inequality and polarization were uncorrelated using household surveys for 67 developing and transnational economies over the period 1981-1994. Income distribution enhanced as often as it worsened in growing economies, and negative growth was often more unfavorable to distribution than positive growth. Regarding the effect of economic growth on inequality, according to Goudie and Ladd (1999) first there is little persuasive evidence that growth systematically changes distribution and second in the nonexistence of clear link, countries pursue a policy growth-oriented. Few other studies have analyzed the impact of inequality on poverty. Deininger and Squire (1998) examined how initial inequality and concomitant changes in inequality impact poverty. They found that the poorest 20 percent suffer the most from growth decreasing effects of inequality. Initial inequality also hurts the poor via credit rationing and powerlessness to invest. Ravallion (2001) also shows that the poor might gain more from redistribution but suffer more than the rich from economic shrinkage. However, while some argue even with this strong relationship, it might be the case that countries with initially severe inequality may be less successful at reducing poverty; earlier models (Harrod-Domar model) envisaged that greater inequality would lead to higher growth rates. Forbes (2000) also challenged the belief on negative link between inequality and growth using data from 45 countries over the period 1966-1995. Forbes results using various estimation methods show that in short and medium term, an increase in income inequality has a significant positive effect on economic growth. 6

Alesina and Rodrik (1994) illustrate in a political economy context that when inequality is high i.e. the poor have less voice and accountability, the median voter will push for distortionary taxes, which will have discouraging effects on savings and hamper growth. 2.2. Governance and Pro-poor Growth A large number of studies have investigated the role of good governance for economic development and poverty reduction. Kaufmann and Aart (2002) affirmed that per capita income and the quality of governance are strongly positively correlated across countries. Their empirical strategy allows separating this positive correlation into two parts. First, they show that a strong positive causal effect runs from better governance to higher per capita. This result displayed the importance of good governance for economic development. Second, they found a weak, even negative, casual effect running from in the opposite direction from per capita income to governance. Hence, there is no evidence of Virtuous Circles in which higher income leads to further improvement in governance. Based on a sample of 92 countries, Dollar and Kraay (2002) found that the rule of law may be related to a greater share of growth accruing to the poorest 20 percent of the population. Dollar and Kraay (2002) found that the rule of law indicators is positively and significantly correlated with growth in per capita incomes of the poorest quintile. Their conclusion is that greater rule of law may be associated with a greater share of growth accruing to the lowest 20 percent of the population. This is predominantly due to the indicator s influence through growth rather than through improving distribution. These results are somewhat covered by Kraay (2004). Resnick and Regina (2006) developed a conceptual framework that specified the relationship between different aspects of governance and pro-poor growth. Using this framework, the paper reviewed a range of quantitative cross-country studies that include measures of governance as independent variables and focuses on the dependent variable in at least two of three dimensions of pro-poor growth: poverty, inequality and growth. The review indicated that governance indicators, such as political stability and rule of law are associated with growth but provide mixed results regarding poverty reduction. On the other hand, governance indicators that refer to transparent political systems, such as civil liberties and political freedom, tend to conduce for poverty reduction, but the evidence is rather mixed and the relationship of these variables with growth remains unclear. 7

Lopez (2004) assessed whether policies that are pro growth are also pro-poor. He found that policies might not be poverty-reducing in the short run, but in the long run. Though, he argues that political economy constraints could prevent these policies from staying in place long enough to reach that poverty-reduction level. Furthermore, White and Anderson (2001) argued that the higher the initial Gini coefficient, the less the poor benefit from growth and there are apparent trade-offs between growth and distribution by examining sectorial patterns of growth. More civil liberties tend to have a less propoor impact while more political freedom tends to have a more pro-poor impact while ethnic fractionalization appears to increase the poor s participation in the growth process. Agricultural growth tends to be less pro-poor while the opposite is true for growth in the services sector. Kraay (2004) found that 60 percent to 95 percent of poverty changes are explained by growth in average income while changes in income distribution are relatively more significant in the short run. Additionally, he studied that rule of law and accountability are both positively correlated with growth and distributional changes while openness to international trade has a positive correlation with growth and correlated with poverty- reducing shifts in incomes. All these studies suggest that good governance is pro-poor in terms of increasing incomes and reducing the poverty headcounts. Yet, they are less clear about what the intervention mechanism is i.e., increased growth, improved equity, or a combination of both that leads to such outcomes. III. Data and Trends in Poverty and Income Distribution 3.1. Data 3.1.1. Measuring Poverty and Inequality This paper uses two measures to capture poverty in pro-poor growth regressions and one measure of inclusiveness. 8

The first proxy for poverty reduction in pro-poor growth regressions is the income of the bottom 20 percent in the income distribution (llllllll). We use Dollar-Kleineberg-Kraay s (DKK) 1 dataset to measure the average income of the bottom 20 percent in the income distribution. Indeed, a positive growth is important but not sufficient to assess whether the poor benefit or not. That is, it is necessary to evaluate how the benefits of growth are shared amongst the different income sets. Thus, it is relevant to use the income of the poorest 20 percent in examining the effects of income growth on the income of poor. We also use the same dataset to measure the income share of the first quintile (llllll). This dataset builds on a larger dataset of 963 country-year observations for which household surveys are available. It emerges from the fusion of data accessible in two high-quality editions of household surveys data: the Luxembourg Income Study (LIS) database, covering mostly developed countries, and the World Bank s POVCALNET database, covering essentially developing countries. The survey means are converted to constant 2005 USD in order to be consistent with POVCALNET data. DKK s dataset covers a total of 151 countries between 1967 and 2011. However, this paper covers the period 1975-2012, leading to 113 countries because of severe missing data issues and outliers 2 problems. Table A1 in Appendix A presents the list of countries. The second proxy for poverty reduction is the poverty headcount ratio at $2 a day in purchasing power parity. This measure is based on the percentage of the population living on less than $2 a day at 2005 international prices. We measure mean income per capita income as real per capita GDP 3 at purchasing power parity in constant 2005 international dollars, from the World Development Indicators (WDI) databank. The logarithm of the Gini index is our measure of inequality. The Gini index measures the extent to which the distribution of income or consumption expenditure among individuals or households within an economy deviates from a perfectly equal distribution. The Gini index measures the area between the Lorenz 4 curve and a hypothetical line of absolute equality, expressed as a percentage of the maximum area under the line. Thus a Gini index of 0 represents perfect equality, while an index of 100 implies perfect inequality. 1 See Growth Still Is Good for the Poor, The World Bank Development Research Group, 2013 2 Armenia seems to be an outlier so we end up with 113 countries instead of 114 countries. 3 The two «names» are equivalent in the rest of this paper 4 A Lorenz curve plots the cumulative percentages of total income received against the cumulative number of recipients, starting with the poorest individual or household. 9

3.1.2. Defining and Measuring Governance The concept of Governance is widely discussed among scholars and policymakers. It means different things to different people and there is as yet no consensus around its definition. Consequently, there are varying definitions of Governance. Theoretically, governance can be defined as the rule of the rulers, typically within a given set of rules. In the context of economic growth and poverty reduction, governance refers to essential parts of the wide-ranging cluster of institutions. The United Nations Development Program (UNDP, 1997) defines governance as the exercise of economic, political, and administrative authority to manage a country s affairs at all levels. It comprises mechanisms, processes, and institutions through which citizens and groups articulate their interests, exercise their legal rights, meet their obligations, and mediate their differences. According to the World Bank (1993), the governance is the process through which power is exercised in the management of a country s political, social and economic institutions for development. Kaufmann, Kraay and Zoido-Lobaton (1999) explain that the fundamental aspects of governance are graft, rule of law, and government effectiveness. Other dimensions are: voice and accountability, political instability and violence, and regulatory burden. Within this notion of governance, the evident interrogation is: what is good governance? This paper associates the quality of governance with democracy and transparency, with the rule of law and good civil rights, and with efficient public services. Also, the quality of governance is determined by the impact of this exercise of power on the quality of life enjoyed by the citizens. In order to measure this concept, we use the Worldwide Governance Indicators (WGI). The WGI have been proposed by the World Bank to estimate good governance. World Bank governance indicators, developed by Kaufmann et al. (2005), are a set of worldwide measures of six composite indicators of governance perception for 299 countries. Higher value of these indicators corresponds to better outcomes. The point estimates range from -2.5 (weak governance) to 2.5 (strong governance). There exist three dimensions of governance: political, economic and institutional dimensions. Thus, the governance indicators can be classified into three groups with two indicators in each cluster. First, the political feature of governance is proposed to capture the process by which government is nominated, supervised and replaced. The first indicator voice and accountability captures political, civil and human rights and independence of the media. It measures the extent to which citizens of a country are able to participate in the selection of government. Although, the 10

second indicator, political stability, measures whether the government in power will be destabilized by possibly unconstitutional or violent means, including for instance military cop, assassination, terrorism. The second dimension that is the Economic Governance includes Government effectiveness and Regulatory quality. The government effectiveness represents the capacity of the government to effectively formulate and implement sound and complete policies. The regulatory quality includes measures of the incidence of market unpromising policies such as price control or inadequate bank supervision, as well as the perceptions of the worries imposed by excessive regulation in area such as foreign trade and business development. The third dimension represents the institutional feature of governance. It involves rule of law and control of corruption indicators. The rule of law summarizes several indicators that measure the extent to which agents have confidence in the rules of the society. It measures the quality of contract enforcement. These indicators also measure a society s success in developing environment in which fair and predictable rules form basis for economic and social interactions. Control of corruption reveals perceptions conventionally defined as the exercise of public power for private gains, including both petty and grand forms of corruption, as well as "capture" of the state by elites and private interests. In addition to the main variable of interest described above, a set of control and structural variables will also be considered in this paper. This set includes variables related to inequality, health, human capital, gender parity, infrastructure, openness to trade, employment and financial variables. These control variables reflect the state of the empirical literature on the determinants of economic growth and poverty reduction. Table A2 in Appendix A summarizes the description and source of the variables. 3.2. Trends in Poverty and Income Distribution The purpose of this section is to provide an overview of how the poverty and the distribution of income have evolved around the world. The table below presents the number of people living on less than $1.25 per day by region. One observes that across the world the poverty headcount at $1.25 declined from 1990 to 2008, but it remains high essentially in Sub-Saharan Africa and Asia. 11

In addition to the remaining high poverty levels in most of the developing countries, income inequality has risen across the world over the last two decades even in developed countries. According to the OECD, income inequality measured by the Gini coefficient rose by 10 percent from the mid-1980s to the late 2000s in OECD countries, while the ratio of top income to bottom income decile reached its highest level in 30 years. Graph 1 below displays the evolution of the Gini coefficient for the East Asia and Pacific. One observes that inequality has been particularly growing in China and India because the rich benefited more from robust growth than both low and middle-income households while some countries such as Cambodia, Thailand, Malaysia, have experienced reductions in inequality. According to official 5 estimates of the Asian Development Bank, China s Gini increased from 37 percent in the mid 1990s to 49 percent in 2008. Graph 1: Evolution of inequality in East Asia and Pacific region 5 See http://www.adb.org/data/main 12

Despite of these rises in inequality, incomes of the poor and the income share of the poor (Graph 2 in Appendix A) have increased across the world, with South Asia experiencing the greatest growth in the income of the poor. Graph 3: Evolution of the income of the poor in South Asia As a preliminary analysis of our data, Table A3, in Appendix A, presents the statistic summary of the main within and between transformed variables. 13

To capture pictures in data, on the relationship between poverty reduction and GDP per capita, we plot Figure 1 that shows the tight link between poverty reduction and GDP per capita (overall per capita income). For both transformed between and within variables, we see that income growth increases with the income of the poor. In countries, which are in the fitted values line, such as Ukraine and Czech Republic, the average income of the poor grows equi-proportionately with the average income (GDP per capita). Moreover, Figure 2 (in Appendix A) confirms that income growth decreases poverty. Figure 1: Growth and the income of the poor (a) Between transformed variables 14

(b) Within transformed variables Furthermore, in order to illustrate with our data how good governance impacts average incomes and the income of the poor, Figure 3 (in Appendix A) presents correlation matrices for both the between and within transformed variables between the GDP per capita, the income of the poor and the control 15

of corruption. The top graph, in Figure 3, shows that there is a positive relationship between the three variables. However, there is a weak positive link when regarding the between transformed variables. IV. Econometric Methodology For the empirical investigation, we split the sample period 1975-2012 into 10 non-overlapping 4- year periods (the latter period is the mean of two years), in order to control for business cycle fluctuations. This method is popular in the majority of empirical works, which are especially related to economic growth study. In this paper, we examine, using Fixed Effects 6 models, the following equation depending on the different effects the paper is investigating. (1) where YY iiii is a vector of three variables depending on what the paper is investigating. This vector includes first yyyy iiii - the income of the poorest 20 percent in the income distribution in country ii at time tt, second PP iiii poverty headcount ratio at $2 a day (PPP) in percentage of population in country ii at time tt, and finally a measure of inclusiveness QQ iiii that denotes the bottom quintile share (the income share of the poorest 20 percent) of the income distribution in country ii at time tt. llllll iiii is the logarithm of GDP per capita, llllllllllll iiii denotes the logarithm of the Gini index. GGGGGG denotes a set of the six governance indicators plus the aggregated indicator of governance that we create using the Principal Component Analysis (PCA 7 ), and XX represents the set of control variables. Finally, ii indicates a country-specific effect, μμ tt specifies a time-specific effect, and εε iiii is the timevarying error term. The regression model follows the empirical literature that argues while per capita income growth is a key factor in reducing poverty, it can produce very different rates of poverty reduction, meaning that other factors matter. Hence, following Ravallion and Chen (1997), this paper allows poverty and the income of the poor to also depend on the inequality denoted by the Gini index, which proxies for the main factors affecting the income distribution. Indeed, it is likely that growth in average income shifts the income distribution and variations in inequality change the shape of the 6 Fixed Effects models control for unobserved heterogeneity when this heterogeneity is constant over time and correlated with independent variables. Note: for benchmark results, we will also used the Between Effects model. 7 PCA is a procedure that transforms a number of correlated variables into a smaller number of uncorrelated variables called principal components. It is use for finding patterns in data of high dimension. 16

distribution, both of these phenomena can impact on the income of the poor and the poverty headcount ratio. As a starting point, the paper examines the impact of economic growth on the income of the poorest 20 percent and poverty headcount at $2 a day, in order to examine how pro-poor growth is, by examining equation (1) and using llnnnnnn iiii and llnnnn iiii as dependent variables with only llnnnn iiii and llllllllllll iiii as explanatory variables i.e. with ββ GGGGGG = 0 and ββ XX = 0. The coefficients of interest are ββ, which gives the impact of economic growth on poverty reduction as the equation is in logarithm terms, and γγ measures the effect of a change in the Gini index on poverty reduction. The second step consists in adding first the governance aspects, and then structural factors in the estimating equation, in order to capture the impacts of good governance on poverty reduction. The paper also investigates how inclusive growth is, by using equation (1) and considering llnnnn iiii - the logarithm of the bottom quintile share of the income distribution- as dependent variable. As for the pro-poor growth regressions, we follow two steps by allowing the income share to depend on inequality and then on governance and structural variables. Since, the paper considers growth as inclusive when income growth is associated with an increase in the bottom quintile share of the income distribution, growth is inclusive if ββ is greater than one. However, in order to avoid as much as possible the issues that complicate estimation with typical linear estimators, we consider the Generalized Method of Moments in System (SGMM) approach in addition to the fixed-effect estimator. This estimator handles well two issues that tend to complicate estimation with typical linear estimators such as FE model: First the relation between poverty reduction and some of its expected determinants is complex, with the possibility of reverse causality. For instance, infant mortality rate might have a negative impact on the poverty reduction while children from poor families are more likely to die before age 5. Second. Second, poverty reduction process may be dynamic. It is likely that the past levels of income of the poor as well as the past levels of poverty reduction impact on current levels. Consequently, we address these issues by using the Dynamic Panel Data (DPD) approach. Following Arrelano and Bover (1995), we will use the SGMM approach. The DPD implies that the FE and the instrumental variables approaches do not exploit all the information available in the sample. Let s note that the Generalized Method of Moments (GMM) can avoid this problem. It is specified as a system of equations in which the instruments are 17

appropriate to each equation. These instruments contain lags of the levels of endogeneous as well as strictly exogenous variables. However, Arrelano and Bover (1995) and Blundell and Bond (1998) revealed a potential limitation of the Arrelano-Bond estimator. The lagged levels are often poor instruments for especially first difference variables. Thus, the SGMM method may provide more efficient estimates since in this method the variables in levels are instrumented with their own first differences. Theoretically, the SGMM estimator symbolizes the following assumptions: The process may be dynamic, with current realizations of the dependent variable influenced by past ones. Some explanatory variables may be endogenous There may be subjectively distributed fixed individual effects. The idiosyncratic disturbances (from the fixed effects) may have individual -specific patterns of heteroskedasticity and serial correlation. The idiosyncratic disturbances are uncorrelated across individuals. The number of time periods of data: T may be small and we may have large N Some regressors can be not strictly exogenous, that is, some regressors can be influenced by their past values. The SGMM represents a system of two equations: one differenced and one in levels. Thus, equation (1) can be rewritten as system (2) considering the SGMM approach: (2) V. Estimation Results 5.1. Pro-poor Growth Regressions Our baseline empirical specification consists in a simple FE regression of the logarithm of the income of the poorest 20 percent and poverty headcount ratio at $2 on the logarithm of the GDP per capita. Tables 2 and Table 3 below document these results. 18

Since, following Ravallion and Chen (1997), we define growth as pro-poor simply if it reduces poverty, results suggest that growth is in general pro-poor, leading to significant increases in the income of the poor. These findings hold for the two measures of poverty reduction. Especially, while a 1 percent increase in real GDP per capita leads to about a 1.42 percent increase in the income of the poor (Table1, column (5), SGMM), the same one percent increase in real GDP per capita leads to a decrease of about 2.25 percent in the poverty headcount (Table 2, Column (4), SGMM). All these coefficients are highly significant at the 1 percent level though the impacts are small with FE estimators. However, one percent increase in the Gini coefficient more or less directly counterweights the positive impact on poverty reduction of the same increase in the per capita income for all the specifications. Let s note that, when using the Between Effects (BE) model, the coefficients for pro-poor growth are greater than the ones of the Fixed Effects models while they are smaller than SGMM coefficients. Table 2: Pro-poor Growth Regressions- Income of the poorest 20 percent (1) (2) (3) (4) (5) (6) Variables lnyp lnyp lnyp lnyp lnyp lnyp Log of GDP per capita 0.607*** 0.661*** 1.11*** 0.98*** 1.42*** 1.02*** (0.05) (0.05) (0.05) (0.03) (0.14) (0.08) Log of Gini Index -1.37*** -2.01*** -1.64*** (0.13) (0.2) (0.26) Constant 1.47*** 5.79*** -2.87*** 5.6*** -5.53*** 3.82** (0.49) (0.67) (0.44) (0.89) (1.27) (1.29) Observations 517 426 517 426 517 426 19

R-squared 0.21 0.4 0.8 0.9 AR(1) test 0.66 0.51 AR(2) test 0.3 0.23 P-Value Hansen test 0.11 0.2 Number of countries 112 109 112 109 112 109 Model FE FE BE BE SGMM SGMM Note: Robust standard errors in parentheses: *** p<0.01,** p<0.05, * p<0.1. Diagnostic tests 8 (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use from lag 1 to lag 4 of explanatory variables as instruments for (5) and only lag 1 for (6). Table 3: Pro-poor Growth Regressions- Poverty Headcount ratio at $2 (1) (2) (3) (4) (5) (6) Variables lnp lnp lnp lnp lnp lnp Log of GDP per capita -1.024*** -1.15*** -6.25*** -2.25*** -1.667*** -9.04*** (0.13) (0.12) (1.07) (0.44) (0.21) (1.87) Log of Gini Index 2.53*** -9.05*** 3.93*** -13.8*** (0.33) (2.44) (0.66) (4.17) Gini*y 1.42*** 2.09*** (0.29) (0.52) Constant 11.17*** 2.84* 44.19*** 21.29*** 1.88 64.4*** (1.13) (1.52) (8.76) (3.64) (3.01) (14.79) Observations 424 421 421 424 421 421 R-squared 0.14 0.27 0.31 AR(1) test 0.9 0.81 0.68 AR(2) test 0.57 0.1 0.21 P-Value Hansen test 0.13 0.2 0.5 Number of countries 92 92 92 92 92 92 Model FE FE FE SGMM SGMM SGMM Note: Robust standard errors in parentheses: *** p<0.01,** p<0.05, * p<0.1. Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use from lags 1 and 2 of explanatory variables as instruments for (4) and only lag 1 for (5) and (6). As explained in section 4, in the rest of this paper the preferred estimation will be the SGMM approach. The literature review highlighted well a growing understanding that economic, political, and institutional governance is critical for economic success and failure of countries. This poses the question of whether the positive impact of good governance on a nation s economic condition leads to a decrease in poverty reduction. 8 Correlations between endogeneous variables and instruments (variables in lag and difference) are presented in Appendix (Table A4) 20

To this end, we estimate system (1) considering the two measures of poverty reduction, governance indicators but without the structural variables (ββ XX = 0). The results of the impact of governance indicators on pro-poor growth are presented in Table 4 for the income of the poor. In Table 4, the sign of all governance indicators are positive and significant except the indicator of political stability and absence of violence that is not significant. A one percent increase of the aggregated governance index, that combines political, economic and institutional features of good governance using the PCA, increases the income of the poor by 0.14 percent. Indeed, good governance reforms positively impact poverty reduction by providing better opportunities to the poor. This is mostly likely to happen through the pro-poor services, which will be well distributed in presence of good governance reforms, but also through the protection of property rights and the rule of law, through anti-corruption policies and democratization. All of these factors permit the poor to protect their rights better, guarantee that a greater part of the public goods that they are allowed to are in fact provided. Also, this result confirms the growing logic that in their strategies for economic development, donors are increasing their focus on governance issues. Institutional governance is represented by rule of law and control of corruption. The results suggest that a better rule of law (0.24) and control of corruption (0.39) significantly increases the income of the poor, leading to the conclusion that institutions matter. This is consistent with previous empirical findings. This dimension of governance has somewhat founded the preeminence of institutions for effective market economy and it is interconnected with the other features of governance. The good political governance dimension that refers to a country s voice and accountability (0.25 percent) and political stability (0.08), has positive impact on the income of the poor. This indicates that a more developed political system is pro-poor in terms of increasing incomes and reducing poverty. Definitely, democracy eases progresses in economic integration. Societies that have well-defined rules, strong legal systems including political mechanisms that are transparent, efficient and follow the law, attract investment. This is also the case when governments effectively provide security and prevent violence. Thus, investors are confident that their investments are secure. In this line, Acemoğlu and Robinson, in Why Nations Fail, argue that less developed countries such as Egypt are poor because It [Egypt] has been ruled by a narrow elite that have organized society for their own benefit at the expense of the vast mass of people. Political power has been 21

narrowly concentrated, and has been used to create great wealth for those who possess it. They defend that developed countries such as the United Kingdom and the United States grew successful because they created inclusive institutional and political institutions that benefit society as a whole. Bad economic governance (non effectiveness of government and bad quality of regulation) is a major issue that can be a cause and a consequence of poverty around the world. It affects the poorest the most, in advanced or developing nations, since corruption undermines political progress, democracy, economic growth, and more. It worsens inequality and disadvantages smaller domestic firms. Corrupt governments can distort decision-making in favor of projects that profit the few rather than the many. Our results (columns (3) and (5)) show that an improvement in government effectiveness positively impacts the income of the poor by about 0.35 percent, while an enhancement in regulatory quality increases their income by about 0.42 percent. Undeniably, the regulation of economic institutions reveals countries capacity to succeed in boosting economic activity by different sectors. Therefore, good economic governance including fighting against corruption is a way of supporting public trust and promotes widespread benefits. Our results confirm this: a 1 percent increase in the quality of control of corruption policy leads to an increase of about 0.39 percent in the income of the poor. Table 4: Governance indicators and Pro-poor Growth Regressions (1) (1) (2) (3) (4) (5) (6) (7) Variables lnyp lnyp lnyp lnyp lnyp lnyp lnyp Log of GDP per capita 0.94*** 0.75*** 0.87*** 0.9*** 0.85*** 0.88*** 0.83*** (0.15) (0.14) (0.13) (0.08) (0.12) (0.1) (0.7) Log of Gini Index -1.21*** -1.65*** -1.4*** -1.41*** -1.33*** -1.38*** -1.67*** (0.42) (0.33) (0.37) (0.31) (0.38) (0.3) (0.29) Governance 0.14** (0.07) Control of Corrup 0.39*** (0.13) Gov Effectiveness 0.35** 22

(0.15) Political Stability 0.08 (0.08) Regulatory quality 0.42*** (0.12) Rule of law 0.24* (0.13) Voice and Account 0.25*** (0.08) Constant 2.94 6.22*** 4.26** 3.46** 4.06** 4.08** 5.52*** (2.45) (2.14) (2.2) (1.51) (1.96) (1.53) (1.43) Observations 286 286 286 286 286 286 286 AR(1) test 0.43 0.45 0.49 0.24 0.41 0.41 0.21 AR(2) test 0.49 0.07 0.33 0.18 0.89 0.33 0.05 P-Value Hansen test 0.05 0.013 0.14 0.05 0.04 0.15 0.07 Number of countries 107 107 107 107 90 107 107 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM Note: Robust standard errors in parentheses: *** p<0.01,** p<0.05, * p<0.1. Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use from lag 1 to lag 4 of explanatory variables as instruments for (1) (2) and (3) and from lag 1 to lag 3 for (4) (5) (6) (7). Table 5 (in Appendix B) presents the results of the impact of governance indicators on poverty headcount ratio. Results suggest that growth still is pro-poor since for every specification, an increase in income per capita decreases the poverty headcount ratio by more than 1 percent. Moreover, estimated effects of governance indicators on the poverty headcount ratio are not all significant. One observes that control of corruption; government effectiveness and rule of law are no longer significant in reducing poverty. Besides, a one percent increase in the aggregated governance indicator reduces poverty by 0.3 percent. Estimates of the other governance indicators are significant and go from 0.3 for political stability to -0.4 for voice and accountability. As explained before, generally good governance improves the income of the poor and decreases poverty. Moreover, it seems that good governance reduces the impact of inequality on poverty reduction. Furthermore, the results vary across regions. Not surprisingly, Table 6 shows that growth is more pro-poor in countries with high income. For all the three specifications, rises in per capita income increase more than proportionately the income of the poor. The impact is particularly greater (specification (3): -2.15) when adding the control of corruption, even through the control of corruption is not significant. Yet, in all the other regions, growth is pro-poor following our simple definition. For some regions, however, the estimated effects of the control of corruption on the income of the poor are somewhat unexpected. Especially, in Europe and Central Asia, East Asia and Pacific, as well as high-income 23

countries, the control of corruption has a negative but non-significant effect on the income of the poor. The effect is significant in South Asian countries. This can be explained by the fact that the level of political corruption, which refers to the corruption operated by politicians, in South Asian countries especially in Bangladesh, India, Nepal and Sri-Lanka is among the highest in the sample. Consequently, it is challenging for the political structures to credibly engage in anti-corruption undertakings in the society. In Latin America and the Caribbean as well as in the Middle East and Africa, the control of corruption has a significant impact on the income of the poor. Still, this effect is not significant for the latter region. Democratization and economic reforms in the region of Latin America and the Caribbean, which enhance the outcomes of control of corruption policies, permit us to explain these results. 24

Table 6 9 : Results by Regions-Control of Corruption and Pro-poor growth regressions Europe& Latin MENA& High South Asia East Asia& Central America& Sub-Sahara Income Pacific Regions Asia Caribbean Africa (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) Variables lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp Log of GDP per capita 0.87*** 0.72*** 0.67*** 0.24 0.29 0.38*** 0.79*** 0.86*** 0.65*** 1.65*** 1.86*** 2.15*** 0.49*** 0.5*** 0.62*** 0.63*** 0.6*** 0.66*** (0.12) (0.09) (0.09) (0.22) (0.19) (0.14) (0.15) (0.05) (0.09) (0.15) (0.13) (0.35) (0.1) (0.07) (0.09) (0.07) (0.07) (0.09) Log of Gini Index -1.26*** -1.18*** -0.91-2.41*** -2.04*** -2.36*** -1.49** -1.78*** -0.57** -0.12 0.06-0.2 (0.33) (0.33) (0.9) (0.59) (0.45) (0.14) (0.54) (0.62) (0.17) (0.4) (0.46) (0.56) Control of corrupt -0.13 0.27** 0.14-0.04-0.16** -0.02 (0.20) (0.1) (0.15) (0.27) (0.07) (0.1) Constant -0.82 4.89*** 4.98 3.99** 7.15** 12.38*** -0.51 6.7*** 9.5*** -7.74*** -4.94* -6.81** 1.99** 3.89*** 1.37 0.82 0.84 1.27 (1.08) (1.43) (1.65) (1.96) (3.92) (1.96) (1.16) (1.82) (1.85) (1.6) (2.5) (3.48) (0.73) (0.58) (1.79) (0.55) (1.27) (1.87) Observations 85 85 63 114 111 69 117 117 78 128 42 33 24 24 15 49 47 28 AR(1) test 0.26 0.39 0.82 0.31 0.23 0.68 0.46 0.08 0.17 0.07 0.67 0.21 0.24 0.49 0.16 0.14 0.13 0.13 AR(2) test 0.05 0.64 0.13 0.3 0.57 0.13 0.33 0.93 0.81 0.06 0.21 0.99 0.31 0.74 0.11 0.16 0.12 0.66 P-Value Hansen test 0.14 0.78 0.65 0.21 0.75 0.6 0.11 0.33 0.57 0.1 0.99 0.8 0.95 1 1 0.72 1 1 Number of Countries 18 18 18 21 21 21 32 32 31 25 23 22 6 6 6 10 9 9 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM Note: Robust standard errors in parentheses: *** p<0.01,** p<0.05, * p<0.1. Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use only lag 1 of explanatory variables as instruments for all the specifications. 9 We keep the control of corruption as good governance indicator following Kaufmann et al. (1999). 25

5.2. Inclusive Growth Regressions In this section, we follow Dollar and Kraay (2002), and examine the relationship between GDP per capita (per capita income) and a broader definition of inclusiveness the bottom quintile share of the income distribution. Debates on inclusiveness usually focus on the incidence of poverty and the income distribution among individuals and households in the society. Income shares are conventional metrics for gauging the distributive impact of policies. If income growth is not associated with a decrease in the income share of the bottom quintile, then growth is considered as inclusive. Specifically, we use equation (1) and system (2). The results are shown in Table 7. By using FE models, we don t exactly get the known Dollar- Kraay result that average incomes of the poorest 20 percent in the society, rise proportionately with per capita income because the coefficient is negative even though not significant (column (1)). Column (2) displays results when controlling for inequality, and we find a weak evidence of inclusive growth because of the positive but not significant coefficient. However, once we consider SGMM approach (column (4)), growth is inclusive with a positive and significant coefficient. When controlling for inequality, an increase in per capita income leads to a decrease in the income share of the bottom quintile that is growth is not inclusive. This result holds when we add the interaction term between inequality and per capita income. Also, inequality tends to rise with economic growth. The idea that the benefits of economic growth do not trickle down automatically may run counter to traditional economic wisdom. For instance, the period of economic growth that preceded the financial crisis in 2007 was characterized by growing inequalities of income and opportunities. As our preferred estimation is the SGMM, findings (columns (5) and (6)) explain the need to reverse the negative trends in inequality and to ensure that future economic expansions benefit more the poor than the richer in the society. The ideal is to generate growth and then use it in such way as to bring equality of income and chances as well as fairness in opportunity. Table 7: Inclusive Growth Regressions 26

(1) (2) (3) (4) (5) (6) Variables lnq lnq lnq lnq lnq lnq Log of GDP per capita -0.015 0.034-0.456** 0.13* -0.05** -0.8** (0.03) (0.02) (0.21) (0.08) (0.02) (0.4) Log of Gini Index -1.412*** -2.52*** -1.47*** -3.27*** (0.06) (0.48) (0.08) (0.98) Giniy 0.137** 0.21* (0.05) (0.11) Constant -2.77*** 2.02*** 6.02*** -4.07*** 3.03*** 9.32** (0.33) (0.31) (1.75) (0.71) (0.41) (3.5) Observations 522 426 426 522 426 426 R-squared 0.01 0.60 0.60 AR(1) test 0.33 0.73 0.65 AR(2) test 0.18 0.62 0.54 P-Value Hansen test 0.03 0.11 0.05 Number of countries 112 109 109 112 109 109 Model FE FE FE SGMM SGMM SGMM Note: Robust standard errors in parentheses: *** p<0.01,** p<0.05, * p<0.1. Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator.. We use from lag 1 to lag 4 of explanatory variables as instruments for (4) and from lag 1 to lag 3 for (5) and (6). Inclusiveness goes beyond poverty and income distribution and involves other dimensions such as governance. To making growth inclusive, we need building effective institutions. This raises the question about what are key governance factors and mechanisms that can facilitate growth and reduce poverty. To this end, we estimate system (1) using the income share of the poor as dependent variable, with interaction terms between governance indicators and per capita income and without structural factors. Results are presented in Table 8. All coefficients of our governance indicators are significant and positive. High and sustained growth, equal access to opportunities, strong social protection for the most vulnerable and poor, are major pillars reinforced by good governance. Findings are striking: good governance is key in achieving growth in the income share of the poorest but there is no evidence of significant effect of a growth in per capita income on the income share of the bottom quintile. Rises in income of the poor are smaller when growth in income per capita gets greater. This leads to the question whether economic growth was 27

complemented with liable and transparent public administration, and nondiscriminatory redistribution of the gains of growth. Political stability (political governance), regulatory quality (economic governance) and rule of law (institutional governance) are the governance features that have the most positive impact on the income share of the bottom quintile. A nation s political stability provides a good signal to investors and creates new job opportunities for the poor, and good economic governance is needed. Also, confidence in the rules of the society is essential in attaining inclusive growth. 28

Table 8: Governance indicators and Inclusive Growth Regressions (1) (2) (3) (4) (5) (6) (7) Variables lnq lnq lnq lnq lnq lnq lnq Log of GDP per capita -0.02-0.03-0.03 0.03 0.01-0.003-0.01 (0.03) (0.03) (0.03) (0.05) (0.05) (0.06) (0.04) Log of Gini Index -1.69*** -1.7*** -1.66*** -1.76*** -1.65*** -1.65*** -1.63*** (0.11) (0.11) (0.11) (0.15) (0.14) (0.14) (0.85) Governance 0.35*** (0.12) Governance*y -0.04*** (0.01) Control of Corrup 0.77*** (0.25) Control of Corrup*y -0.08*** (0.02) Gov Effectiveness 0.74*** (0.23) Gov Effectiveness*y -0.08*** (0.02) Political Stability 1.07** (0.57) Political Stability*y -0.14** (0.06) Regulatory quality 1.01** (0.42) Regulatory quality*y -0.12*** (0.04) Rule of law 0.86** (0.38) Rule of law*y -0.1** (0.04) Voice and Account 0.85** (0.04) Voice and Account*y -0.1** (0.04) Constant 3.6*** 3.75*** 3.61** 3.33*** 3.16*** 3.27*** 3.3*** (0.56) (0.57) (0.58) (0.77) (0.75) (0.84) (0.63) Observations 286 286 286 286 286 286 286 AR(1) test 0.41 0.28 0.45 0.34 0.47 0.43 0.64 AR(2) test 0.27 0.17 0.25 0.16 0.46 0.21 0.3 P-Value Hansen test 0.28 0.25 0.35 0.55 0.28 0.32 0.18 Number of countries 107 107 107 107 107 107 107 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM Note: Robust standard errors in parentheses: *** p<0.01,** p<0.05, * p<0.1. Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use from lag 1 and lag 2 of explanatory variables as instruments for (1) (2) (4) (7) and only lag 1 for the remaining specifications. 29

Besides, these results vary significantly across regions. When considering specification (1) of Table 9, growth has been significant and inclusive in Europe and Central Asia while in the other regions the elasticity is negative and to some extent positive for the Middle East and Sub-Saharan Africa but not significant for the latter. Inclusive growth in Europe and Central Asia is related to the successful reduction in income-based poverty and improving living standards other the past decade. Yet, inequality nullifies this inclusive growth effect. Essentially, for all the regions, as inequality cancels or negates inclusive growth, it remains an important issue for policy makers. Moreover, control of corruption is not significant for any region except High Income countries where it has a negative effect on the income share of the bottom quintile. It is likely that in these countries, the proportion of the poor in the total population is much smaller than that of the rich. Thus, outcomes of anti-corruption policies are felt less. In the Middle East and North Africa region and Sub-Saharan Africa, growth becomes inclusive when we add the inequality dimension. Recently Africa has enjoyed fairly rapid economic growth; this growth may create and expand economic opportunities and allow people to contribute and to benefit from the development process. Besides, the control of corruption has found to decrease the income share of the poor, raising interrogations about the effective implementation of these policies. 30

Table 9: Results by Regions: Control of Corruption and Inclusive Growth Regressions Europe& Latin MENA& High South Asia East Asia& Central America& Sub-Sahara Income Pacific Regions Asia Caribbean Africa (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) Variables lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq lnq Log of GDP per capita 0.13** 0-0.01-0.23-0.11 0.04 0.04 0.04* -0.03* -0.22* -0.28*** -0.14-0.18** -0.01-0.04-0.18*** -0.04-0.02 (0.05) (0.01) (0.02) (0.15) (0.11) (0.07) (0.09) (0.02) (0.01) (0.13) (0.05) (0.08) (0.1) (0.02) (0.02) (0.05) (0.04) (0.02) Log of Gini Index -1.16*** -1.17*** -2.06*** -2.87*** -1.81*** -1.76*** -1.09*** -0.82*** -1.03** -1.09*** -0.99*** -1.43*** (0.16) (0.09) (0.61) (0.35) (0.13) (0.07) (0.15) (0.18) (0.14) (0.11) (0.22) (0.14) Control of corrupt -0.01 0.05 0.02-0.13** -0.03-0.03 (0.46) (0.1) (0.03) (0.27) (0.02) (0.02) Constant -3.69*** 1.45** 1.63*** -1.36 5.7** 7.5*** -3.25*** 3.58*** 3.98*** -0.5 3.94*** 1.78** -1.12** 1.28** 1.7** -1.26*** 1.27 2.73*** (0.46) (0.59) (0.46) (1.35) (2.39) (1.26) (0.7) (0.48) (0.32) (1.6) (0.79) (0.85) (0.46) (0.5) (0.52) (0.37) (1.96) (0.45) Observations 85 85 63 114 111 69 117 117 78 133 42 33 24 24 15 49 47 28 AR(1) test 0.33 0.24 0.23 0.76 0.2 0.57 0.48 0.1 0.14 0.13 0.6 0.27 0.3 0.18 0.26 0.96 0.4 0.22 AR(2) test 0.43 0.68 0.19 0.42 0.38 0.08 0.31 0.36 0.26 0.27 0.06 0.73 0.22 0.23 0.44 0.06 0.32 0.45 P-Value Hansen test 0.25 0.78 0.88 0.39 0.94 0.81 0.37 0.37 0.5 0.63 1 0.92 0.97 1 0.98 0.69 1 1 Number of Countries 18 18 18 21 21 21 32 32 31 25 23 22 6 6 6 10 9 9 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM Note: Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use from lag 1 of explanatory variables as instruments for all the specifications. 31

5.3. What determines how Pro-poor and Inclusive Growth is? In this section, we try to uncover which factors drive how pro-poor and inclusive growth is. To this end, we estimate respectively pro-poor growth and inclusive growth systems by using SGMM approach. First, for the pro-poor growth regressions we use the income of the poorest 20 percent as dependent variable. Table 10 shows the results of the SGMM estimations. For all specifications, increases in per capita income significantly and positively impact the income of the poor - growth is pro-poor. In terms of what governs how pro-poor growth is, regressions that add each structural variable one by one indicate that health expenditure, spending in education, secondary school enrolment, improvement in sanitation infrastructure, employment in services, all significantly increase the income of the poor. Besides, gender parity, access of improved water and investment have positive (but not significant) impact on poverty reduction. Though, some other factors undermine this effect infant mortality, prevalence of HIV, openness. The latter factor being driven by a possible unequal share of opportunities that may weaken prospect for the poor to be better off. In the literature, results are ambiguous regarding trade openness. Dollar and Kraay (2002) found that trade openness, measured as the ratio of exports plus imports to GDP, is correlated with growth and reduces poverty. Lopez (2004) suggested that the impact of trade openness on the poor might vary according to the sectors in which the poor are concentrated. Measuring the trade openness as the volume of trade adjusted by a country s size and population, he found that while trade openness appears to increase poverty in the short run, it is negatively correlated with poverty in the long run. Besides, our results suggest that this has a negative effect on the income of the poor increases poverty- even though the coefficient is not significant. In column (17), we provide a combination of factors spending in education, financial development measured by M2 and financial openness have significant and positive impacts on the income of the poor. We find no significant effect of health expenditure, gender parity, and investment, on the income of the poor even though these factors have positive effects. 32

Above all, the control of corruption leads to significant increases in the income of the poor for all specifications, which confirms the benchmark results. In addition, these results hold using the poverty headcount ratio. Human capital, infrastructure, employment in agriculture and industry, and investment, are found to significantly reduce poverty. Results are displayed in Table 11 in the Appendix B. 33

Table 10: Structural determinants of How Pro-poor Growth Is (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) Variables lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp lnyp Log of GDP per capita 0.65*** 0.40** 0.55*** 0.91*** 0.59*** 0.63*** 0.46*** 0.55*** 0.73*** 0.72*** 0.88*** 0.79*** 0.92*** 0.89*** 0.71*** 0.55*** 0.45*** (0.1) (0.18) (0.07) (0.15) (0.16) (0.1) (0.18) (0.1) (0.1) (0.13) (0.09) (0.14) (0.1) (0.14) (0.08) (0.11) (0.14) Log of Gini Index -1.73*** -1.61*** -1.08*** -1.73*** -0.92** -1.68*** -1.43*** -1.54*** -1.89*** -1.61*** -1.51*** -1.54*** -2.08*** -1.87*** -1.7*** (0.25) (0.09) (0.34) (0.35) (0.53) (0.35) (0.3) (0.31) (0.37) (0.34) (0.23) (0.32) (0.27) (0.26) (0.4) Control of Corrup 0.35*** 0.62*** 0.012 0.33** 0.43*** 0.08 0.5*** 0.52*** 0.42*** 0.39*** 0.22** 0.28*** 0.32*** 0.23** 0.44*** 0.4*** 0.26* (0.12) (0.14) (0.13) (0.14) (0.13) (0.12) (0.16) (0.09) (0.11) (0.13) (0.11) (0.1) (0.12) (0.11) (0.1) (0.12) (0.1) Hegdp 0.06* 0.03 (0.03) (0.04) Mortality5-0.006** (0.003) pvih -0.02* (0.01) SpendingEdu 0.17*** 0.06* (0.06) (0.03) SchoolSec 0.007** (0.004) GenderParity 0.12 (0.46) Sanitation 0.013** 0.005 (0.007) (0.004) Water 0.004 (0.007) Inflation 0* 0.02 (0.0006) (0.001) M2 0.001 0.006** (0.002) (0.002) Openness -0.002-0.002 (0.002) (0.002) EmploymentA -0.004 (0.007) EmploymentI -0.03*** (0.01) EmploymentS 0.001 (0.007) Investment -0.01*** (0.006) 34

FinOpenness 0.09 0.09* (0.06) (0.05) Constant 7.1*** 3.53** 7.56*** -1.94*** 4.88*** 7.23*** 5.02** 7.57*** 5.57*** 5.95*** 6.14*** 5.83** 5.02*** 4.45*** 8.55*** 8.65*** 7.92*** (1.29) (1.71) (1.11) (1.34) (1.78) (1.57) (2.61) (1.66) (1.75) (1.97) (1.69) (2.39) (1.19) (1.85) (1.24) (1.54) (1.66) Observations 286 328 202 287 251 123 284 282 276 284 286 222 222 222 275 283 236 AR(1) test 0.3 0.9 0.23 0.42 0.85 0.44 0.47 0.54 0.4 0.43 0.33 0.87 0.86 0.93 0.36 0.53 013 AR(2) test 0.17 0.6 0.04 0.9 0.13-0.15 0.09 0.12 0.11 0.08 0.08 0.06 0.08 0.06 0.04 0.56 P-value Hansen test 0.09 0.102 0.51 0.2 0.104 0.5 0.096 0.102 0.109 0.09 0.13 0.16 0.26 0.13 0.11 0.108 0.27 Number of ccountries 107 110 70 103 102 76 106 106 104 107 107 94 94 94 103 105 95 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM Note: Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use only lag 1 of explanatory variables as instruments for all the specifications. 35

Turning to the inclusive growth regressions, Table 12 presents results of two multivariate specifications with the aggregated governance and control of corruption indicators. The results show that growth is inclusive since growth in per capita income leads to an increase in the income share of the poor. Besides, inequality strongly negatively impacts on this income share while good governance seems to support policies for inclusive growth. Especially, the control of corruption significantly impacts on the income share of the poor. Moreover, the pro-health policy reflected by improvement in health expenditure is good for the poor. Another point is that financial openness and inflation impacts negatively on the income share of the poor even though this effect cannot nullify the inclusive growth effect. Table 12: Inclusive Growth and Structural Variables (1) (2) Variables lnq lnq 36

Log of GDP per capita 0.07 0.09* (0.05) (0.05) Log of Gini Index -1.75*** -1.76*** (0.09) (0.1) Governance 0.029 (0.023) Control of Corrup 0.1** (0.04) HeGDP 0.32*** 0.4*** (0.09) (0.11) He*y -0.04*** -0.05*** (0.01) (0.01) SpendingEdu -0.01-0.02 (0.014) (0.01) Sanitation -0.0003-0.000 (0.001) (0.001) Openness 0.0008 0.0009 (0.0007) (0.0007) FinOpenness -0.04* -0.04** (0.02) (0.018) Inflation -0.001** -0.001** (0.0005) (0.0005) M2-0.0004-0.0008 (0.0006) (0.0006) Constant 3.08*** 3.02*** (0.47) (0.44) Observations 236 236 AR(1) test 0.45 0.58 AR(2) test 0.16 0.104 P-Value Hansen test 0.43 0.16 Number of countries 95 95 Model SGMM SGMM Note: Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use only lag 1 of explanatory variables as instruments for (1) and (2). 37

5.4. Evidence of linearity or non-linearity This section discusses the evidence of linearity or non-linearity from two interesting perspectives: the effects of growth and the ones of good governance on poverty reduction. A simple test consists of exogenously splitting the sample according to the median level of the variables of interest as a threshold point. As a first point, we examine the impact of good governance 10 on the income of the poor as a function of the level of development. Table 13 presents SGMM regressions results for countries below and above the median of the level of development measured as the logarithm of the GDP per capita. These results are outstanding. Especially, in specification (1), we find that the impact of the control of corruption on the income of the poor is positive and significant for all countries with high development level (above the median level) and it is nil or negative for countries with lower development level (below the median). These findings give insights on the way governments monitored good governance policies. Indeed, countries such as Côte d Ivoire or Mali tend to not have such strong institutions than developed countries as France or Luxembourg, to support anti-corruption policies. In addition, findings show that growth in per capita income has greater impact on the income of the poor in high-developed countries. Regarding inequality, the effects on the poor do not vary a great deal between the two groups of countries. Furthermore, specification (2) includes structural factors in addition of the base model. Only financial development (M2 as a percentage of GDP) is positive and significant for countries with high development level. Still, it is important to point out some stylized facts that the results confirm: while factors that enhance development (health expenditure, spending in education, improved access to sanitation) are more important in poverty reduction in less developed countries, financial factors (financial openness and M2 as a percentage of GDP) are likely to be more important for poverty reduction in more developed countries where the poor have a higher likelihood to have access to financial services. 10 Control of corruption a proxy for good governance here 38

Table 13: Impact of Governance on Income of the Poor Below and Above the Median 11 Levels of Development (1) (2) Level of Development Level of Development Below Above Below Above Log of GDP per capita 0.6*** 1.25*** 0.47*** 1.03*** (0.17) (0.14) (0.15) (0.18) Log of Gini Index -1.63*** -1.68*** -1.18*** -1.58*** (0.34) (0.26) (0.25) (0.36) Control of Corrup -0.02 0.21** -0.26** 0.25*** (0.15) (0.08) (0.12) (0.09) HeGDP 0.04 0 (0.02) (0.06) SpendingEdu 0.01 0.06 (0.02) (0.07) Sanitation 0.002 0.001 (0.002) (0.005 Openness 0-0.002 (0.002) (0.001) FinOpenness 0.01 0.009 (0.04) (0.04) Inflation 0.001 0.001 (0.000) (0.001) M2 0.002 0.004* (0.002) (0.002) Constant 7.09*** 1.58* 5.81*** 2.82 (2.04) (1.9) ( 1.41) (2.62) Observations 146 140 119 117 AR(1) test 0.25 0.23 0.18 0.81 AR(2) test 0.61 0.08 0.82 0.84 P-Value Hansen test 0.32 0.13 0.97 1 Number of Countries 58 59 52 51 Model SGMM SGMM SGMM SGMM Note: Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use only lag 1 of explanatory variables as instruments for (1) and (2). The second step consists in investigating the effect of growth on the income of the poor as a function of the quality of governance 12. As in the previous method, we split the sample in two groups of countries according to the median level of governance indicators. Countries that are below the median are those who have lower governance quality while those above the median have greater governance quality. 11 Median of the log of GDP per capita=8.5616 12 Quality of governance is measured by the aggregated governance and control of corruption indicators 39

Results are displayed in Table 14. Specifications (1) both for aggregated governance indicator and control of corruption provide no evidence of strong changes of the impact of income per capita growth on the income of the poor. Especially, regarding control of corruption, the effects of a 1 percent increase in income per capita on the income of the poor goes from 0.82 percent to 0.81 percent. Also, we observe that when anti-corruption policy is well managed, it has higher and significant impact on poverty reduction. However, with the aggregated indicator of governance, the impact of growth on the poor income goes from 0.74 to 0.67 percent. Let s note that for both indicators of governance, our findings reveal that inequality has less impact on the income of the poor in countries with higher quality of governance. Estimates from specifications (2), which include structural variables, tend to reinforce our previous findings and follow the theory on good governance. Indeed, for both governance indicators, the impacts of structural factors on the income of the poor are in general greater in countries with better quality of governance. 40

Table 14: Impact of Growth on Income of the Poor Below and Above the Median 13 Levels of Governance (1) (2) (1) (2) Governance Governance Control of Control of Corruption Corruption Below Above Below Above Below Above Below Above Log of GDP per capita 0.74*** 0.67*** 0.57*** 0.33*** 0.82*** 0.81*** 0.59*** 0.67*** (0.12) (0.21) (0.07) (0.1) (0.11) (0.13) (0.05) (012) Log of Gini Index -1.79*** -0.88*** -1.63*** -1.5*** -1.4*** -1.34*** -1.09*** -1.86*** (0.45) (0.3) (0.3) (0.26) (0.34) (0.35) (0.23) ( 0.3) Governance -0.005 0.43*** 0.003 0.33*** (0.09) (0.13) (0.06) (0.07) Control of Corrup 0.05 0.61*** -0.1 0.38*** (0.18) (0.17) (0.15) (0.1) HeGDP 0.02 0.05 0.01 0.06 (0.03) (0.04) (0.02) (0.05) SpendingEdu 0.02 0.01 0.03 0.07** (0.03) (0.04) (0.002) (0.03) Sanitation 0.001 0.001 0.003-0.004 (0.002) (0.002) (0.002) (0.004) Openness -0-0.005*** 0.001-0.005 *** (0.002) (0.001) (0.001) (0.001) FinOpenness 0.03 0.09** -0.01 0.09** (0.039) (0.04) (0.03) (0.04) Inflation 0 0 0 0 (0.000) (0.000) (0.000) (0.001) M2-0 0.007*** 0.001 0.008*** (0.001) (0.001) (0.002) (0.001) Constant 6.6*** 3.68* 7.15*** 8.61*** 4.57** 4.3** 4.63*** 7.15*** (2.27) (2.19) ( 1.42) (1.09) (1.74) (1.79) (0.87) (1.25) Observations 156 127 129 107 154 132 126 110 AR(1) test 0.83 0.35 0.74 0.24 0.63 0.47 0.45 0.32 AR(2) test 0.05 0.05 0.1 0.77 0.07 0.14 0.09 0.89 P-Value Hansen test 0.42 0.25 1 1 0.13 0.58 0.97 0.94 Number of Countries 63 64 55 56 66 65 57 57 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM Note: Diagnostic tests (Hansen and first and second-order autocorrelations) reveal no evidence against the validity of the instruments used by the SGMM estimator. We use only lag 1 of explanatory variables as instruments for all the specifications. 13 Median of level of governance=-0.507 ; median of level of control of corruption=-0.36 41

VI. Check Panel Smooth Transition Regression: Robustness In this section, we introduce and estimate the Panel Smooth Transition Regression (PSTR) model, so as to accommodate other issues that have arisen in the literature on the relationship between poverty reduction, economic growth and good governance, and to test the robustness of our results. We will (i) examine the impact of good governance on the income of the poor as a function of development level and (ii) we will investigate the growth effects on the income of the poor through the level of governance quality. The PSTR developed by González et al. (2005), is a generalization of the Hansen (1999) Panel Threshold Regression (PTR) model. The PSTR considers the speed of transition from one regime to the other. The passage from a regime to another is gradual. Hence, this method may be more accurate than the one used in the previous sub-section. 6.1. Estimation of Model (i) For the first step (i), the PSTR model is defined as follow: (3) Where llllyyyy iiii denotes the logarithm of income of the 20 percent poorest, GGGGGG iiii represents a vector of governance indicators; llllll iiii is the logarithm of the GDP per capita, uu ii is an individual fixed effect, and εε iiii stands for the idiosyncratic error. Moreover, the transition function is given by a logistic function: (4) Where gg(llllyy iiii, γγ, δδ) is a continuous function and it is bounded between [0,1]. It depend on the transition function i.e. the logarithm of GDP per capita (llllyy iiii ), a smooth parameter γγ, and a threshold parameter δδ. The advantage of this method compared to SGMM is the fact that it incorporates the change effect of individual heterogeneity in the same country over time. This approach introduces the concept of heterogeneity in time and space. Besides, the PSTR allows the effect of good 42

governance on poverty reduction to vary with the level of economic development. Accordingly, the marginal impact of the governance indicator depending on the economic development is given by: (5) The properties of the transition function involve: When estimating the parameters of the PSTR model, the individual effects uu ii are removed by eliminating individual-specific means and thus it is a transformed model by nonlinear least squares that one estimates (González et al. (2005)). The testing procedure of González et al. (2005) consists: first to test the linearity against the PSTR model, and second to determine the number rr of transition function. Considering equation (3), the linearity check consists in testing:. Then three standard tests can be applied using these statistics: Lagrange Multiplier of Fisher (LLLL FF ), Wald test (LLLL), and Pseudo Likelihood-ratio (LLLLLL). The results presented in Table 15, suggest no evidence of non-linearity regarding the effects of governance on the income of the poor as a function of the level of development. These findings are contrasting with our previous results, which show that the impact of good governance on the income of the poor is greater in countries with high development levels. 43

Table 15: Parameter estimates for the PSTR model (i) Threshold variable Level of Development N of transition function (r*) 1 (HH 0 : rr = 0 vvvv HH 1 : rr = 1) Fisher Test of linearity (LLLL FF ) 0.018 (0.982) Wald Test (LLLL) 0.056 (0.972) LRT 14 Test 0.056 (0.945) (HH 0 : rr = 1 vvvv HH 1 : rr = 2) Fisher test of no remaining nonlinearity (LLLL FF ) 0.402 (0.670) Wald Test (LLLL) 1.246 (0.536) LRT Test 1.249 (0.536) Number of Observations 1120 15 Note: The test of linearity has an asymptotic FF(1, TTTT NN 1) distribution under HH 0 and FF(1, TTTT NN 2) for the no remaining nonlinearity test with NN the number of individuals and TT the number of periods. For statistics, the p-values are in parentheses. For parameters, ββ 0 and ββ 1, the standard errors are parentheses and are corrected for heteroskedascity. 6.2. Estimation of Model (ii) For the second step (ii), the PSTR model is written as follow: 16 (6) where the logistic transition function is: (7) The marginal effect of income growth depending on governance quality is given by: (8) The properties remain the same as in the first step. 14 Likelihood Ratio Test 15 10 periods and 112 countries 16 With two explanatory variables : log of GDP per capita and control of corruption 44

Table 16 below presents the results of the second step using equation (6). Depending on the transition function 17, the effects of income growth on the income of the poor are positive and significant. Also, we find evidence of non-linearity. The effects of growth of per capita income in the income of the poor have been found to decrease with the control of corruption. Graph 4 illustrates these findings. The relationship between corruption and economic growth may be difficult to establish: the control of corruption can be detrimental to growth, after reaching some thresholds. This theory can hold for low governance countries (as shown in Governance specifications in Table 14 section 5.4), control of corruption seems lead to greater impacts of per capita income growth on the income of the poor in these countries than in countries with higher levels of governance. Table 16: Parameter estimates for the PSTR model (ii) Threshold variable CCCCCCCCCCCC N of transition function (rr ) 1 (HH 0 : rr = 0 vvvv HH 1 : rr = 1) Fisher Test of linearity 3.024 (0.051) Wald Test 8.613 (0.013) LRT Test 8.727 (0.000) (HH 0 : rr = 1 vvvv HH 1 : rr = 2) Fisher test of no remaining nonlinearity 0.193 (0.825) Wald Test 0.573 (0.751) LRT Test 0.574 (0.751) Parameter ββ 0 0.378 (0.024) Parameter ββ 1-0.057 (0.02) Parameter 0-0.071 (0.07) Parameter 1 0.111 (0.07) Location parameter δδ 0.29 Smooth parameter γγ 2.66 Number of Observations 918 18 Note: The test of linearity has an asymptotic F(1, TN N 1) distribution under H 0 and F(1, TN N 2) for the no remaining nonlinearity test with N the number of individuals and T the number of periods. For statistics, the p-values are in parentheses. For parameters, β 0 and β 1, the standard errors are parentheses and are corrected for heteroskedascity. 17 The transition function depends upon the governance indicator : control of corruption here. 18 The first period and 10 countries were dropped from the model (2) regression. 45

Graph 4: Marginal impact of income growth on the income of the poor 46

VII. Conclusion and Discussion This paper examines first how pro-poor and inclusive growth is by assessing respectively the impacts of income growth on poverty reduction and on the bottom share of the income distribution using a sample of 113 countries over the period 1975-2012. Second, it investigates the effects of good governance in reducing poverty and attaining inclusive growth. Finally, this paper assesses what factors have been driving these outcomes. It comes not surprisingly that growth is in general pro-poor. Incomes of the poorest 20 percent rise while poverty headcount ratio at $2 decreases with mean per capita incomes as economic growth proceeds. But inequality negates this effect. The regressions corroborate many of the stylized facts while displaying differences across regions in how pro-poor growth has been. A second important finding is that the combination of political, economic and institutional features of good governance improves the income of the poor and decreases poverty. Especially, the control of corruption, government effectiveness and regulatory quality, have the most positive impact on the income of the poor. Our third result is that globally growth has not been inclusive. But, once we use the SGMM approach, findings from the benchmark specification show that the bottom share of the income distribution rises faster than GDP per capita (average income). Still, inequality dampens this effect. Good governance also appears to increase the bottom quintile share of the income distribution. Regarding inclusive regressions, some regions do better; growth in Europe and Central Asia as well as in the MENA and Sub-Saharan Africa has been inclusive when considering the simple relationship between the logarithm of the bottom share in the income distribution and the logarithm of the GDP per capita and also when controlling for inequality. These outcomes are quizzical, and are likely driven by the recent high economic growth in these regions. 47

Fourth, when studying what determines how pro-poor and inclusive growth is, the results suggest that expansion of human capacities through health and education spending, infrastructure enhancement, and financial development (M2 as percentage of GDP and financial openness) are the main factors positively influencing poverty reduction. The results also suggest that programs such as fighting infant mortality and HIV/AIDS are pro-poor. Also, using multivariate specifications of structural factors, results show that growth is inclusive, stipulating ways to conduct inclusive policies. Finally, using the PSTR approach, we find evidence of non-linear relationship of the impacts of income growth on the income of the poor as a decreasing function of the control of corruption. The countries in which political corruption dominates may drive these outcomes. It may be understood from this paper that an important element of pro-poor and inclusive strategies is to continue efforts to strengthen governance, make government and policies more transparent and effective, control corruption, and promote economic and social fairness. In addition, the policies for attaining pro-poor and inclusive growth need to be more broad-based by focusing on social development but also on financial inclusion. The financial inclusion policies must be closely monitored especially in achieving inclusive growth. Financial inclusion is considered to be an essential element for social inclusion of the poorest and the most vulnerable in the society. Many economic models predict that financial development can reduce both poverty and inequality directly by allowing larger investments, lowering credit constraints on the poor, and indirectly by enhancing capital allocation and growth (Banerjee and Newman, 1993; Rangarajan, 2008). Beyond the expected effects of financial inclusion on inclusive growth, it is interesting to examine how vulnerabilities and crises in the financial system that could follow buoyant economic activity could adversely impact the poor and dampen inclusive growth; and how good governance may interact with these effects. 48

Appendix Appendix A Table A1: Country list 19 Country Code Region 1.Albania ALB ECA 2.Algeria DZA MENA 3.Argentina ARG LAC 4.Armenia ARM ECA 5.Australia AUS HIINC 6.Austria AUT HIINC 7.Azerbaijan AZE ECA 8.Bangladesh BGD SA 9.Belarus BLR ECA 10.Belgium BEL HIINC 11.Belize BLZ LAC 12.Bhutan BTN SA 13.Bolivia BOL LAC 14.Bosnia and Herzegovina BIH ECA 15.Botswana BWA SSA 16.Brazil BRA LAC 17.Bulgaria BGR ECA 18.Burkina Faso BFA SSA 19.Burundi BDI SSA 20.Cambodia KHM EAP 21.Cameroon CMR SSA 22.Canada CAN HIINC 23.Central African Republic CAF SSA 24.Chile CHL LAC 25.China CHN EAP 26.Colombia COL LAC 27.Costa Rica CRI LAC 28.Cote d Ivoire CIV SSA 29.Croatia HRV HIINC 30.Czech Republic CZE HIINC 31.Denmark DNK HIINC 32.Dominican Republic DOM LAC 33.Ecuador ECU LAC 19 Since Armenia is an outlier, it was dropped from the estimation. 49

34.Egypt, Arab Republic EGY MENA 35.El Salvador SLV LAC 36.Estonia EST HIINC 37.Ethiopia ETH SSA 38.Fiji FJI EAP 39.Finland FIN HIINC 40.France FRA HIINC 41.Gambia, The Republic GMB SSA 42.Georgia GEO ECA 43.Germany DEU HIINC 44.Ghana GHA SSA 45.Greece GRC HIINC 46.Guatemala GTM LAC 47.Guinea GIN SSA 48.Guinea-Bissau GNB SSA 49.Guyana GUY LAC 50.Honduras HND LAC 51.Hungary HUN HIINC 52.India IND SA 53.Indonesia IDN EAP 54.Iran, Islamic Republic IRN MENA 55.Ireland IRL HIINC 56.Israel ISR HIINC 57.Italy ITA HIINC 58.Jamaica JAM LAC 59.Jordan JOR MENA 60.Kazakhstan KAZ ECA 61.Kenya KEN SSA 62.Kyrgyz Republic KGZ ECA 63.Lao PDR LAO EAP 64.Latvia LVA ECA 65.Lesotho LSO SSA 66.Lithuania LTU ECA 67.Luxembourg LUX HIINC 68. Macedonia. FYR MKD ECA 69.Madagascar MDG SSA 70.Malawi MWI SSA 71.Malaysia MYS EAP 72.Maldives MDV SA 73.Mali MLI SSA 74.Mauritania MRT SSA 75.Mexico MEX LAC 50

76.Moldova MDA ECA 78.Montenegro MNE ECA 79.Mozambique MOZ SSA 80.Namibia NAM SSA 81.Nepal NPL SA 82.Netherlands NLD HIINC 83.Nicaragua NIC LAC 84.Niger NER SSA 85.Nigeria NGA SSA 86.Norway NOR HIINC 87.Pakistan PAK SA 88.Panama PAN LAC 89.Paraguay PRY LAC 90.Peru PER LAC 91.Philippines PHL EAP 92.Poland POL HIINC 93.Romania ROM ECA 94.South Africa ZAF SSA 95.Swaziland SWZ SSA 96.Sweden SWE HIINC 97.Switzerland CHE HIINC 98.Tajikistan TJK ECA 99.Tanzania TZA SSA 100.Thailand THA EAP 101.Timor-Leste TMP EAP 102.Trinidad and Tobago TTO HIINC 103.Tunisia TUN MENA 104.Turkey TUR ECA 105.Turkmenistan TKM ECA 106.Uganda UGA SSA 107.Ukraine UKR ECA 108.United Kingdom GBR HIINC 109.United States USA HIINC 110.Uruguay URY LAC 111.Venezuela, RB VEN LAC 112.Vietnam VNM EAP 113. Yemen, Republic YEM MENA 114.Zambia ZMB SSA 51

Table A2: Explanation of variables Variable Source Description/Definition Survey means POVCALNET, POVCALNET measures welfare by income or consumption as LIS determined in the surveys. For LIS, DKK calculate survey means of disposable income directly from the micro survey data on household level Lnyp POVCALNET, Logarithm of Income of the poorest 20 percent of the income LIS distribution lnq POVCALNET, Logarithm of the share of the Income of the 20 percent poorest LIS of the income distribution- bottom quintile share lnp WDI Poverty headcount ratio at $2 a day (PPP) in percentage of population lny WDI Logarithm of GDP per capita based on purchasing power parity (PPP constant 2005 international dollar) lngini WDI Logarithm of GINI index. Gini index measures the extent to which the distribution of income or consumption expenditure among individuals or households within an economy deviates from a perfectly equal distribution. Governance variables Voice and accountability WGI It reflects perceptions of the extent to which a country s citizens are able to participate in selecting their government, as well as freedom of association and media. Estimate of governance range from approximately from -2.5 to 2.5 (strong performance) Control of corruption WGI It defines perceptions of the extent to which public power is exercised for private gain and captures the state by elites and private interests. Government Effectiveness WGI It describes perceptions of the quality of public and civil services, the quality of policy design and implementation, and the reliability of the government s duty to such policies. Regulatory quality WGI It defines perceptions of the capability of the government to formulate and realize sound policies and regulations 52

Rule of law WGI It reflects insights of the extent to which agents have confidence in and accept the rules of society (property rights, the police, the courts, the quality of contract implementation) Political Stability and no violence WGI It defines perceptions of the likelihood that the government will be destabilized or defeated by unconstitutional or violent processes, including terrorism and politically-motivated violence Structural Factors Hegdp WDI Public health expenditure (% GDP). It consists of recurrent and capital spending from government (central and local) budgets, external borrowings and grants (including donations from international agencies and nongovernmental organizations), and social (or compulsory) health insurance funds. Mortality5 WDI Under-five mortality rate per 1,000 live births is the probability per 1,000 that a newborn baby will die before reaching age five, if subject to current age-specific mortality rates. pvih WDI Prevalence of HIV refers to the percentage of people ages 15-49 who are infected with HIV. SpendingEdu WDI Public expenditure on education as % of GDP is the total public expenditure on education expressed as a percentage of the Gross Domestic Product (GDP) in a given year. Public expenditure on education includes government spending on educational institutions (both public and private), education administration, and transfers/subsidies for private entities (students/households and other privates entities). SchoolSec WDI Gross enrolment ratio. Secondary. All programs. Total is the total enrollment in secondary education, regardless of age, expressed as a percentage of the population of official secondary education age. GER can exceed 100% due to the inclusion of over-aged and under-aged students because of early or late school entrance and grade repetition. GenderParity UNdata Gender Parity Index for adult literacy rate. It is equal to the ratio of female adult literacy rate to the male adult literacy rate. Water WDI Improved water (% of population with access) - Access to an improved water source refers to the percentage of the population 53

using an improved drinking water source. The improved drinking water source includes piped water on premises and other improved drinking water sources Sanitation WDI Improved Sanitation (% of population with access). Access to an improved sanitation structure refers to the percentage of the population using an improved sanitation structure. Inflation WDI inflation, consumer prices (annual %) Inflation as measured by the consumer price index reflects the annual percentage change in the cost to the average consumer of acquiring a basket of goods and services that may be fixed or changed at specified intervals, such as yearly. M2 WDI Money and quasi-money (M2) as % of GDP Openness WDI Trade openness is the sum of exports and imports of goods and services measured as a share of GDP EmploymentA WDI Employment in agriculture (% of total employment) EmploymentI WDI Employment in industry (% of total employment) EmploymentS WDI Employment in services (% of total employment) Investment WEO Defined as a percentage of GDP FinOpenness The Chinn-Ito index (KAOPEN) is an index measuring a country's degree of capital 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. Source: http://web.pdx.edu/~ito/kaopen_chinn-ito_hi0523.pdf 54

Table A3: Descriptive Statistics of Main Variables Variable Mean Standard Max M in Number of Number of Deviation Observations Countries Between Within Between Within Between Within Log of income of 6.71 1.46 0.27 9.89 7.79 4.3 5.67 531 112 the poor Log of poverty 2.75 1.73 0.63 4.55 5.39-2.78-1.49 434 92 Headcount ratio Log of GDP per capita 8.47 1.17 0.25 10.79 9.79 6.21 7.15 944 112 Log of Gini index 3.68 0.23 0.09 4.23 4.09 3.16 3.31 456 111 Control of corruption -0.05 0.97 0.17 2.43 0.7-1.18-0.69 557 112 Government Effectiv. 0.01 0.93 0.15 2.12 0.79-1.44-0.57 557 112 Political Stability -0.16 0.84 0.27 1.48 1.07-1.94-1.39 557 112 Regulatory quality 0.06 0.87 0.18 1.81 1.06-1.97-0.55 557 112 Rule of law -0.1 0.94 0.15 1.93 0.54-1.47-0.83 557 112 Voice and account. -0.01 0.89 0.17 1.6 0.49-1.9-1.09 557 112 Note: One observes that within transformed variables vary more than between transformed variables. Graph 2: Global Evolution of the income share of the poor Globally, the income share of the poor increased over the period 1975-2010. Lower and upper bounds are anomalous for year 1980, because of the few observations available in 1980 are either greater or smaller than observations for the other years. 55

Figure 2: Growth and poverty headcount ratio at $2 (a) Between transformed variables (b) Within transformed variables 56

Figure 3: Correlation Matrices (a) Between transformed variables (b) Within transformed variables 57

Note: Armenia is an outlier. 58