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 pro-

poor 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 pro-

poor 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 outliers2 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 GDP3 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 Lorenz4 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 official5

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 Effects6 models, the following equation depending on the

different effects the paper is investigating.

(1)

where 𝑌𝑌𝑖𝑖𝑖𝑖 is a vector of three variables depending on what the paper is investigating. This vector

includes first 𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 - the income of the poorest 20 percent in the income distribution in country 𝑖𝑖 at

time 𝑡𝑡, second 𝑃𝑃𝑖𝑖𝑖𝑖 – poverty headcount ratio at $2 a day (PPP) in percentage of population in country

𝑖𝑖 at time 𝑡𝑡, and finally a measure of inclusiveness 𝑙𝑙𝑖𝑖𝑖𝑖 that denotes the bottom quintile share (the

income share of the poorest 20 percent) of the income distribution in country 𝑖𝑖 at time 𝑡𝑡.

𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 is the logarithm of GDP per capita, 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 denotes the logarithm of the Gini index. 𝑙𝑙𝐺𝐺𝐺𝐺

denotes a set of the six governance indicators plus the aggregated indicator of governance that we

create using the Principal Component Analysis (PCA7), and 𝑋𝑋 represents the set of control variables.

Finally, ∝𝑖𝑖 indicates a country-specific effect, 𝜇𝜇𝑖𝑖 specifies a time-specific effect, and 𝜀𝜀𝑖𝑖𝑖𝑖 is the time-

varying 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 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 and 𝑙𝑙𝑙𝑙𝑃𝑃𝑖𝑖𝑖𝑖 as dependent variables with only 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 and

𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 as explanatory variables i.e. with 𝛽𝛽𝐺𝐺𝐺𝐺𝐺𝐺 = 0 and 𝛽𝛽𝑋𝑋 = 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 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖- 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 tests8 (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 (𝛽𝛽𝑋𝑋 = 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 69: 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

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.

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

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 governance10 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 Median11 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 governance12. 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 Median13 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. Panel Smooth Transition Regression: Robustness Check 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 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 denotes the logarithm of income of the 20 percent poorest, 𝑙𝑙𝐺𝐺𝐺𝐺𝑖𝑖𝑖𝑖 represents a

vector of governance indicators; 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖 is the logarithm of the GDP per capita, 𝑢𝑢𝑖𝑖 is an

individual fixed effect, and 𝜀𝜀𝑖𝑖𝑖𝑖 stands for the idiosyncratic error.

Moreover, the transition function is given by a logistic function:

(4)

Where 𝑔𝑔(𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖, 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 (𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖), 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 𝑢𝑢𝑖𝑖 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 (𝐿𝐿𝐿𝐿𝐹𝐹), 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 (𝐻𝐻0: 𝑟𝑟 = 0 𝐺𝐺𝑣𝑣 𝐻𝐻1: 𝑟𝑟 = 1) Fisher Test of linearity (𝐿𝐿𝐿𝐿𝐹𝐹) 0.018 (0.982) Wald Test (𝐿𝐿𝐿𝐿) 0.056 (0.972) LRT14 Test 0.056 (0.945) (𝐻𝐻0: 𝑟𝑟 = 1 𝐺𝐺𝑣𝑣 𝐻𝐻1: 𝑟𝑟 = 2) Fisher test of no remaining nonlinearity (𝐿𝐿𝐿𝐿𝐹𝐹) 0.402 (0.670) Wald Test (𝐿𝐿𝐿𝐿) 1.246 (0.536) LRT Test 1.249 (0.536) Number of Observations 112015

Note: The test of linearity has an asymptotic 𝐹𝐹(1,𝐿𝐿𝑙𝑙 − 𝑙𝑙 − 1) distribution under 𝐻𝐻0 and 𝐹𝐹(1,𝐿𝐿𝑙𝑙 −𝑙𝑙 − 2) for the no remaining nonlinearity test with 𝑙𝑙 the number of individuals and 𝐿𝐿 the number of periods. For statistics, the p-values are in parentheses. For parameters, 𝛽𝛽0 and 𝛽𝛽1, 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 function17, 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 𝐶𝐶𝐺𝐺𝑟𝑟𝑟𝑟𝑢𝑢𝑙𝑙 N° of transition function (𝑟𝑟∗) 1 (𝐻𝐻0: 𝑟𝑟 = 0 𝐺𝐺𝑣𝑣 𝐻𝐻1: 𝑟𝑟 = 1) Fisher Test of linearity 3.024 (0.051) Wald Test 8.613 (0.013) LRT Test 8.727 (0.000) (𝐻𝐻0: 𝑟𝑟 = 1 𝐺𝐺𝑣𝑣 𝐻𝐻1: 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 𝛽𝛽0 0.378 (0.024) Parameter 𝛽𝛽1 -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 91818

Note: The test of linearity has an asymptotic F(1, TN − N − 1) distribution under H0 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, β0 and β1, 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 list19

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

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.

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

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

Table A4: Correlations between explanatory variables and lagged and differenced instruments used in the SGMM estimation.20

Variable Log of GDP per capita Log of Gini Lag_1(Log of GDP per capita) Lag_1(Log of Gini) Diff (Log of GDP per capita) Diff (Log of Gini) Log of GDP per capita 1 Log of Gini 0.06 1 Lag_1 (Log of GDP per capita) 0.986 0.071 1 Lag_1 (Log of Gini) 0.038 0.892 0.025 1 Diff (Log of GDP per capita) 0.089 -0.057 -0.076 0.078 1 Diff (Log of Gini ) 0.039 0.037 0.087 -0.417 -0.289 1

Let’s note if correlations are small, instruments are weakly correlated with the offending explanatory variable, thus instruments are poor predictors of the endogeneous predictor. In this case, lagged variables of log of GDP per capita and of log of Gini are good instruments, but variables in difference are a little weak because correlations are smaller than 0.1.

20 Table A3 presents correlations for Table 2 (Bechmarck results) in which we use only lag (1 1) as instruments.

59

Appendix B

Table 5: Governance indicators and Pro-poor Growth Regressions (2) (1) (2) (3) (4) (5) (6) (7) Variables lnP lnP lnP lnP lnP lnP lnP Log of GDP per capita -1.29*** -1.76*** -1.58*** -1.51*** -1.17*** -1.81*** -1.55*** (0.27) (0.29) (0.29) (0.18) (0.25) (0.27) (0.2) Log of Gini Index 2.99*** 2.63*** 3.06*** 2.28** 3.59*** 2.66*** 2.75*** (0.78) (0.71) (0.7) (0.7) (0.68) (0.64) (0.68) Governance -0.3* (0.16) Control of Corrup -0.1 (0.35) Gov Effectiveness -0.007 (0.35) Political Stability -0.31** (0.15) Regulatory quality -0.44* (0.25) Rule of law 0.14 (0.32) Voice and account -0.4* (0.23) Constant 2.15 7.46* 4.45 6.60* -0.9 7.9** 5.22* (0.58) (3.84) (3.55) (3.17) (2.85) (3.17) (3.13) Observations 284 284 284 284 284 284 284 AR(1) test 0.1 0.24 0.23 0.2 0.22 0.3 0.12 AR(2) test 0.03 0.01 0.019 0.04 0.06 0.02 0.01 P-Value Hansen test 0.22 0.15 0.21 0.104 0.08 0.38 0.24 Number of countries 90 90 90 90 90 90 90 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 3 of explanatory variables as instruments for (1), lag 1 and lag 2 for (2) (3) (5) (6) only lag 1 for (4) (7).

60

Table 11: Structural determinants of How Pro-poor Growth Is (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) Variables lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP lnP Log of GDP per capita -1.56*** -1.03** -1.11*** -1.86*** -1.59*** -1.31*** -0.90** -1.52*** - 1.77*** 0.72*** 0.88*** 0.79*** 0.92*** 0.89*** 0.71*** 0.55*** (0.19) (0.25) (0.25) (0.28) (0.32) (2.98) (0.38) (0.21) (0.26) (0.13) (0.09) (0.14) (0.1) (0.14) (0.08) (0.11) Log of Gini Index 2.76*** 1.95** 2.03*** 2.98*** 1.8 2.78*** 1.94*** -1.54*** -1.89*** -1.61*** -1.51*** -1.54*** -2.08*** -1.87*** (0.55) (1.01) (0.64) (0.97) (1.16) (0.35) (0.65) (0.31) (0.37) (0.34) (0.23) (0.32) (0.27) (0.26) Control of Corrup 0.16 -0.04*** 0.25 0.55 -0.02*** -0.56 - 0.88** -0.1 0.07 0.39*** 0.22** 0.28*** 0.32*** 0.23** 0.44*** 0.4*** (0.28) (0.34) (0.16) (0.42) (0.27) (0.37) (0.43) (0.29) (0.42) (0.13) (0.11) (0.1) (0.12) (0.11) (0.1) (0.12) Hegdp -0.23** (0.10) Mortality5 0.009** (0.004) pVIH 0.02 (0.02) SpendingEdu -0.26** (0.09) SchoolSec -0.007 (0.01) GenderParity 0.12 (0.46) Sanitation -0.03** (0.016) Water 0.001 (0.01) Inflation -0 (0.001) M2 0.001 (0.002) Openness -0.002 (0.002) EmploymentA -0.004 (0.007) EmploymentI -0.03*** (0.01) EmploymentS 0.001 (0.007) Investment -0.01*** (0.006) FinOpenness 0.09 (0.06)

61

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 except (5) and (8) in which we add lag 2.

Constant 6.16** 10.75*** 4.96** 19.45*** 8.90 2.32 5.35 4.82*** 10.16*** 5.95*** 6.14*** 5.83** 5.02*** 4.45*** 8.55*** 8.65*** (2.64) (2.37) (2.34) (2.66) (2.78) (4.83) (4.53) (3.09) (3.38) (1.97) (1.69) (2.39) (1.19) (1.85) (1.24) (1.54) Observations 284 283 215 240 246 129 281 278 274 284 286 222 222 222 275 283 AR(1) test 0.48 0.11 0.15 0.16 0.21 0.46 0.04 0.18 0.4 0.43 0.33 0.87 0.86 0.93 0.36 0.53 AR(2) test 0.1 0.23 0.22 0.75 0.08 - 0.05 0.045 0.09 0.12 0.11 0.08 0.08 0.06 0.08 0.06 P-value Hansen test 0.26 0.04 0.48 0.45 0.34 0.55 0.053 0.16 0.009 0.06 0.03 0.06 0.26 0.03 0.012 0.008 Number of clusters 90 90 70 84 85 74 90 89 87 107 107 94 94 94 103 105 Model SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM SGMM

62

References: Acemoğlu D., Johnson S., and Robinson J., 2004 “Institutions as the Fundamental Cause of Long-

Run Growth” NBER, Working Paper No. 10481.

Acemoğlu D., and Robinson J., “Why Nations Fail: The origins of power, prosperity, and poverty”

Alesina, A., and Rodrik D., 1991 “Distributive Politics and Economic Growth”, NBER Working

Paper No. 3668.

Alesina, A., and Rodrik D., 1994 “Distributive Politics and Economic Growth”, Quarterly Journal

of Economics, 465-490.

Arrelano, M., and Bover, O., 1995, “Another look at the instrumental variable estimation of error-

components models”, Journal of Econometrics, Vol. 68, Issue 1, Pages 29-51.

Balakrishman, R., Steinberg, C., and Syed, M., 2013, “The Elusive Quest for Inclusive Growth:

Growth, Poverty, and Inequality in Aisa”, IMF Working Paper, WP/13/152/

Banerjee, A., and Newman, A., 1993 “Occupational Choice and the Process of Development”,

Journal of Political Economy, Vol. 101, No.2, pp. 274-98.

Bourguignon, F. and Verdier, T., 2000 “Oligarchy, democracy, inequality and growth”, Journal of

Development Economics, Vol. 62 (2000) 285-313

Deininger, K., and Squire, L., 1998 “A New Data Set Measuring Income Inequality”, The World

Bank Economic Review, 10(3): 565-91.

Dollar, D. and Kraay, A., 2002 “Growth is Good for the Poor”, Journal of Economic Growth, 7,

195-225.

Dollar, D., Kleineberg T., and Kraay, A., 2013 “Growth Still Is Good for the Poor”, Policy

Research Working Paper, WPS6568.

Forbes, K. J., 2000 “A Reassessment of the Relationship Between Inequality and Growth”, The

American Economic Review.

63

Fouquau, J. Destais, G., Hurlin, C., 2009, “Energy demand models: a threshold panel specification

of the Kuznets curve”, Applied Economics Letters 16 (12), 1241-44.

Fouquau, J. Hurlin, C., Rabaud, I., 2008, “The Feldstein-Horioka puzzle: A panel smooth transition

regression approach”, Economic Modelling 25 (2), 284-299.

Gonzalez, A. Tersvirta, T., Dijk, D. v., 2005, “Panel smooth transition re- gression models”,

Stockholm School of Economics, Working Paper Series in Economics and Finance, No. 604.

Goudie, A., and Ladd. P., 1999 “Economic growth, Poverty and Inequality”, Journal of

International Development, Vo. 11, Issue 2, pages 177-195.

Hansen, B. E., 1999. “Threshold effects in non-dynamic panels: Estimation, testing and inference”

Journal of Econometrics 93 (2), 345-368.

Kaufmann, D., and Kraay, A., 2002 “Growth Without Governance”, World Bank- Development

Research Group, World Bank Policy Research Working Paper No. 2928.

Kaufmann D., Kraay, A., and Zodio-Lobaton, P., 1999 “Governance matters”, The World Bank,

Policy Research Working Paper Series 2196.

Kraay, A. 2004 “When Is Growth Pro-Poor? Cross-Country Evidence”, The World Bank-

Development Research Group (DECRG), IMF Working Paper No. 04/47

Kuznets, S., 1955, “Economic Growth And Income Inequality”, American Economic Review.

Lopez, H., 2004 “Pro-growth, pro-poor: is there a tradeoff?, The World Bank, Policy Research

Working Paper Series 3378.

Organisation for Economic Co-operation and Development, 2012, “Income Inequality and Growth:

The Role of Taxes and Transfers”, OECD Economics Department Policy Notes, No. 9 January.

Rauniyar, G., and Kanbur, R., 2010, “Inclusive Development: Two Papers on Conceptualization,

Application, and the ADB Perspective”, Manila: Asian Development Bank.

64

Ravallion, M., and Chen S., 1997, “What Can New Survey Data Tell Us about Recent Changes in

Distribution and Poverty?”, World Bank Economic Review, Vol. 11, No. 2, pp 357-82.

Ravallion, M., and Chen S., 2003, “Measuring Pro-Poor Growth”, World Bank Policy Research

Working Paper No. 2666.

65