The role of growth and income distribution in poverty reduction the case of the south eastern asian countries

Additionally, inequality as well as the level of initial inequality is found to affect the impact of income growth on poverty reduction significantly.. Higher inequality as well as highe

Ho Chi Minh City, Vietnam

Erasmus University of Rotterdam,

Under the Supervision of Assoc Prof Dr Nguyễn Trọng Hoài

Vietnam – Netherlands Programme, August 2013

Ho Chi Minh City, Vietnam

Erasmus University of Rotterdam,

Under the Supervision of Assoc Prof Dr Nguyễn Trọng Hoài

Vietnam – Netherlands Programme, August 2013

This is to certify that thesis entitled “The Role of Growth and Income Distribution in Poverty Reduction: The Case of South-Eastern Asian Countries”, which is submitted

by me in fulfillment of the requirement for the degree of Master of Art in Development Economic to Vietnam-The Netherlands Programme The thesis comprises only my original work and due supervision and acknowledgement have been made in the text to all other material used

Hồ Quốc Tuấn

I would like to acknowledge all the lecturers at the Vietnam – Netherlands Programme for their knowledge of all the courses, during the time I studied at the program Besides, I would like to thank all the academic and technical staffs of the Vietnam – Netherlands Programme for supporting me during that time

I would like to express my gratitude to my family, especially my parents I would not complete this thesis, as well as study in this program, without their scarification, encouragement and important support Additionally, I am also grateful to my better half for her personal support and patience at all times, which accelerate my thesis writing process For the love and expectation of my family, which motivate my effort

to complete this master degree, my mere expresses of gratitude here have never been sufficient

Last, but not least, I thank my friends and people, who have any help and support for

my thesis but are not above-mentioned

ABBREVIATIONS

GDP Gross Domestic Product

Gini Gini coefficient

GLS Generalized Least Squares

GNP Gross National Product

MDG Millennium Development Goal

OLS Ordinary Least Squares

P3 Poverty squared gap index

ABSTRACT

This thesis research contributes to exploring the impact of economic growth and inequality on poverty In detail, this research aims to identify the effect of income growth on poverty and the impact of inequality on the rate at which growth reduces poverty This research use unbalanced-panel data for South-Eastern Asian countries in the period from 1981 to 2010 The research argues that for all three measures of poverty, i.e headcount ratio (P0), poverty gap (P0) and poverty squared gap (P2) income growth has a significant and negative impact on poverty The estimated growth elasticity of poverty for the whole region is -3.25 for P0, -4.00 for P1 and -4.47 for P2 Additionally, inequality as well as the level of initial inequality is found to affect the impact of income growth on poverty reduction significantly Higher inequality as well as higher initial level of inequality decelerates the growth elasticity

of poverty in absolute value Finally, the impacts of income growth and inequality are

larger on the more sensitive measures of poverty

TABLES OF CONTENTS

LIST OF TABLES xi

LIST OF FIGURES xii

Chapter 1 INTRODUCTION 1

1.1 Problem Statement 1

1.2 Research Objectives 2

1.3 Research Questions 2

1.4 Research Scope and Data 3

1.5 Organization of the Thesis 3

Chapter 2 LITERATURE REVIEW 5

2.1 Definitions 5

2.1.1 Poverty 5

2.1.2 Economic Growth 6

2.1.3 Inequality 6

2.2 The Important Role of Economic Growth in Poverty Reduction 7

2.3 Growth and Redistribution Components of Poverty Reduction 10

2.4 The Impact of Inequality on the Responsiveness of Poverty to Growth 12

2.5 Analytical Framework 15

2.6 Chapter Remark 16

Chapter 3 RESEARCH METHODOLOGY 18

3.1 Model Specification 18

3.2 Research Hypothesis 21

3.3 Measurements of Variables 22

3.3.1 Poverty 22

3.3.2 Inequality 23

3.3.3 Income 25

3.3.4 Initial Inequality and Initial Income or Consumption 25

3.3.5 Growth Rate of Poverty, Income and Inequality 25

3.4 Data Collection 26

3.5 Estimation Strategy 27

3.5.1 The Constant Coefficients Model 27

3.5.2 Random Effects Regression Model 28

3.5.3 Fixed Effects Regression Model 28

3.5.4 Choosing between the FE Model and RE Model: The Hausman Specification Test 29

3.6 Chapter Remark 30

Chapter 4 DATA ANALYSIS 31

4.1 South-Eastern Asian Countries: An Overview on Poverty, Income and Inequality 31

4.2 Poverty, Income Growth and Inequality: The Interrelationship 32

4.3 Empirical Results 35

4.3.1 Estimation Process 35

4.3.2 Choosing the Most Appropriate Model 36

4.3.3 Discussions on the Research Results 37

4.4 Estimating the Growth Elasticity of Poverty 41

4.5 Chapter Remark 43

Chapter 5 CONCLUSION AND POLICY IMPLICATION 44

5.1 Conclusion 44

5.2 Policy Implication 45

5.3 Limitation and Suggestion of Further Research 46

REFERENCES 48

APPENDIX A: DESRRIPTION OF THE DATASET 53

APPENDIX B: DESCRIPTIVE STATISTICS OF VARIABLES 59

APPENDIX C: RESULTS OF HAUSMAN TEST 62

APPENDIX D: REGRESSION RESULTS 65

APPENDIX E: ESTIMATION OF GROWTH ELASTICITY OF POVERTY 68

LIST OF TABLES

Table 3.1 Summary of expected sign of coefficients 22

Table 4.1 Summary of Hausman Test Results 36

Table 4.2 Choice between FE model and RE model 37

Table 4.3 Summarized Estimation Results 38

Table 4.5 Estimation of Growth Elasticity of Poverty 42

LIST OF FIGURES

Figure 2.1 Analytical Framework 15

Figure 3.1 Lorenz Curve 24

Figure 4.1 Income growth and poverty 33

Figure 4.2 Inequality and poverty 33

Figure 4.3 Income and inequality 34

Chapter 1 INTRODUCTION

This chapter begins the thesis research by introducing the research topic and states the research problems in terms of social and scientific point of view, which motivate this thesis project This chapter also includes the research objectives and research questions that will be solved by the thesis In addition, this chapter provides the research scope of this thesis It will end in bringing out the organization of the thesis

1.1 Problem Statement

Reducing poverty is the first and conceivably the most important goal of the Millennium Development Goals (MDGs) In detail, the target of the MDG1 is to halve “the proportion of people whose income is less than $1 a day” (United Nations,

2010, p 6) during the period 1990-2015 For the South-Eastern Asian countries, in terms of poverty reduction, the poverty rate (measured by headcount ratio) falls from 39% in 1990 to 19% in 2005 (United Nations, 2010) This is a great achievement that the South-Eastern Asian countries have early met the target of MDG1 in 2005, ten years before the MDGs deadline Moreover, although poverty in this region was reduced sharply (more than a half during this period), there stills a large proportion of people in these countries living on less than the poverty line According to United Nations (2013), the poverty rate of the South-Eastern Asia region, measured by headcount index, is 14% in 2010 Therefore, this successful poverty alleviation progress must be maintained For these reasons, analyzing the sources of the successful poverty alleviation progress of these countries is of interest in both policy and scientific perspectives

Economists have debated about how economic growth and declines in income inequality can reduce poverty In the first side of the debate, researches emphasize that economic growth is crucially important for poverty reduction, and they neglect the role of inequality For example, change in poverty is attributed to the major role of growth by Squire (1993), Bruno, Ravallion, and Squire (1998), and Dollar and Kraay (2002) In the second line of researches, economic growth and income inequality have been argued influence poverty together For instance, Datt and Ravallion (1992) and

Kakwani (1993) decompose changes in poverty into income (growth) and distribution (inequality) effects The final group of thoughts argues that beside the important role played by growth in poverty reduction, income inequality also matters because it affects the rate at which income growth can reduce poverty Some remarkable researches in this group are Ravallion (1997), Collier and Dollar (2001), Ravallion (2001), Adams (2004), Fosu (2008) and Heltberg (2002), who discuss that the growth elasticity of poverty strongly depends on the level of inequality

In light of these conflicting points of view, this thesis is intending to identify the role

of economic growth and inequality in terms of poverty reduction in the case of Eastern Asian countries This thesis pays attention on the South-Eastern Asian countries because of their successful progress of fighting poverty

South-1.2 Research Objectives

The thesis aims to identify the role played relatively by growth and inequality in poverty reduction In other words, it intends to analyze the impact of growth and inequality in poverty In particular, this thesis has two main research objectives: (i) To analyze the extent to which growth can affect poverty by using of elasticity approach

(ii) To identify how inequality can affect the relationship between growth and poverty

by considering a function of growth elasticity of poverty

For the first objective, this research aims to estimate the growth elasticity of poverty, which shows how much poverty will decrease if income increases by 1 per cent The second purpose of this thesis is to identify the impact of inequality on the responsiveness of poverty to income growth This indicates how the growth elasticity

of poverty can be affected by changes in inequality

1.3 Research Questions

This thesis aims to answer two questions:

(i) How much can economic growth reduce poverty?

(ii) How can inequality affect the ability of growth to reduce poverty?

1.4 Research Scope and Data

This thesis concentrates on the South-Eastern Asian countries because of two reasons Firstly, the region is more successful than some other regions of the world in poverty reduction from 1990 to 2010 For instance, according to United Nations (2013), the Sub-Saharan Africa reduces poverty rate from 56% to 48%, Southern Asia from 51%

to 30%, Latin America from 12% to 6%, and all developing countries from 45% to 22% These amounts of reductions (in relative means) are smaller in comparison with the South-Eastern Asia region, which achieves the reduction from 45% to 14% over the same period The second reason is that there still being almost one fifth of the population of this region still living below $1.25 a day in 2010 (United Nations, 2013) Hence, the root cause of poverty reduction in this region must be considered in order to perform effective policy in favor of maintaining poverty alleviation and improving well-being of the poor

The empirical analysis section of the thesis will be implemented by employing unbalanced panel data from 8 countries in the South-Eastern Asia region over the period from 1981 to 2010 The data was collected from ‘PovcalNet: the online tool for poverty measurement developed by the Development Research Group of the World Bank’

1.5 Organization of the Thesis

This thesis includes five chapters, which are organized as follow:

Followed by this introductory chapter, chapter 2 will survey the literatures in the debate about the role of growth and inequality in poverty reduction As mentioned above, there are three sides in the debate In terms of poverty reduction, the first side emphasizes the role of growth, the second side equalizes the role of growth and inequality by decomposing poverty reduction into growth and distribution components, whereas the final side clearly states that inequality affects the ability of growth in poverty alleviation Literatures will be reviewed as they belong to one of these groups depending on their findings Chapter 2 will end with an analytical framework

The third chapter will present the research methodology of the thesis At the

beginning, this chapter will introduce an empirical model that will be used in the

thesis Then, it will provide research hypotheses that will be tested by using the data

analysis in the following chapter This chapter will also bring out the measures of

variables and data collection method It will finally state the estimation strategy which

is the methods of panel data regression

Chapter 4 will bring out the findings of this thesis research that obtained from the

estimation results This chapter also analyzes how the research results can answer the

research questions arisen in the introduction chapter and how the research hypotheses

in chapter 3 will be tested

The final chapter is chapter 5, which will discuss about some concluding remarks

Based on the findings from chapter four, this chapter will also suggest some policy

recommendations for poverty reduction strategy in South-Eastern Asian countries

Chapter 2 LITERATURE REVIEW

This chapter will survey the literatures on the interrelationship between growth, inequality and poverty It will firstly provide definitions of the related terms and key concepts This chapter will then review literatures based on their findings about the growth-poverty-inequality interrelationship As mentioned in the previous chapter, there has been a debate between three sides of researches and three corresponding arguments about the role of growth and changes inequality in poverty alleviation Three main sections in this chapter would be associated with three arguments of the sides in this debate Each of these sections will survey the studies that support the argument of the respective side in the debate The present chapter will end with an analytical framework based on which this thesis perform its empirical study

According to Asian Development Bank (1999, p 5):

Poverty is a deprivation of essential assets and opportunities to which every human is entitled Everyone should have access to basic education and primary health services Poor households have the right to sustain themselves

by their labor and be reasonably rewarded, as well as having some protection from external shocks Beyond income and basic services, individuals and societies are also poor - and tend to remain so - if they are not empowered to participate in making the decisions that shape their lives

This thesis uses the income aspect of poverty This implies that it does not concern with other dimensions of welfare that go along with the definition of poverty, such as health care services, education, etc Because of this problem and the limitation of data

in other aspects of standard of living, the monetary indicators of poverty will be employed in this thesis

2.1.2 Economic Growth

Economic growth or growth, in short, is a widely popular concept defined as an expansion in the economy of a country (Soubbotina & Sheram, 2000) Growth usually refers to an increase in income or per capita income of an economy In particular, economic growth can be conceptualized as an increase in GDP or GNP (as well as their per capita measures) that is calculated in relative or percentage means In addition, growth rate of an economy was calculated as a percentage increase in the income of a country Thus, the higher growth rate is, the larger the increase in income

of a country is

2.1.3 Inequality

According to Litchfield (1999, p 1), inequality is defined as “a dispersion of a distribution, whether that be income, consumption or some other welfare indicators or attributes of a population”

As discussed by World Bank (2013), in analysis, inequality refers to “relative” poverty because it mentions about “having a little in a specific dimensions compared

to other members of society”

In other words, inequality concerns how total welfare can be distributed to individuals and households in a society As mentioned above, social welfare is a multi-

dimensional concept that includes wealth, health, life expectancy, etc However, except for income and expenditure, other dimensions of welfare could not be measured easily As a result, like many other researches, this thesis treats inequality as

it concerns about income inequality that is how total income of an economy is distributed to its households and individuals If a country experiences a large difference between income of the poorest group and the richest group of people, it can

be said to face high inequality, and vice versa

The definitions and concepts have been explained in this section The next three sections will survey literatures mentioned about the poverty-growth-inequality interrelation

2.2 The Important Role of Economic Growth in Poverty Reduction

In the first line of researches on the debate about the poverty-economic inequality interrelationship, researchers emphasized the role of economic growth, rather than changes in income distribution, in poverty alleviation In terms of poverty reduction, many authors argue that economic growth is the main source and the important tool for reducing the incidence of poverty In this section, some empirical researches of authors who stressed the role of growth, called “growth optimists” (Heltberg, 2002), for example Fields (1989), Squire (1993), Ravallion (1995), (Ravallion & Chen, 1997) Bruno et al (1998) and Dollar and Kraay (2002), will be discussed

growth-Fields (1989) aimed to explore the poverty-growth-inequality interrelation by empirically testing his seven hypotheses about the interaction among these concepts with the data from “a research project on the political economy of poverty, inequality and growth” (Fields, 1989, p 167) by World Bank in 1985 The evidence showed no systematic tendency for inequality to increase or decrease with income growth Another result of Fields (1989) is that there is no evidence supporting for the argument that inequality tends to increase more rapidly in the early stages of development than in the later stages The above findings are considered the rejection

of Kuznets hypothesis, which stated that income inequality will increase in the early stage of growth and then it will start to go down, or the relationship between inequality and growth is represented by the inverted U-shape (Kuznets, 1955) In

addition, Fields (1989) found supporting evidences for the reduction in poverty associated with income growth Growth was treated as the major explanation for the decrease in poverty, as in his writing “economic growth nearly always is associated with a reduction in absolute poverty” (Fields, 1989, p 177)

Similarly rejecting the Kuznets hypothesis and assuming that income distribution is relatively stable over time, Squire (1993) performed an empirical research argued that changes in poverty were mainly affected by change in mean income or consumption Using a data set that involves 21 developing countries, this paper obtained the growth elasticity of poverty about -0.24, which can be interpreted as if growth in mean expenditure increases by 1 percentage point, the growth rate of poverty (measured by headcount index) will decrease by 0.24 percentage points Squire (1993) also stated that countries with the growth rate excess of 3% per year faced sharp declines in poverty measured by headcount index whereas in countries of which the growth rate

is less than 1%, the decrease is modest and in countries with negative growth rate, poverty increase Nevertheless, this research experienced the two main drawbacks The first one is the lack of some alternative measures of poverty, i.e poverty gap index and squared gap index Secondly, due to the assumption of stable income distribution, it neglects the role of inequality in poverty reduction

An empirical research that aims to test the impact of growth on poverty and inequality was done by Ravallion (1995) Using data from developing countries in the 1980s, this study obtained a strong negative correlation between poverty and average living standards or average income This paper also rejected the Kuznets hypothesis as it found that there was no tendency for inequality to increase or decrease with economic growth within developing countries (Ravallion, 1995) However, this research faces some problems Firstly, based on the regressions results obtained in this paper, there stills a noticeable variance in poverty that cannot be explained in the regressions Secondly, because of the employed cross-countries data, this paper did not consider the trend of increasing average income, which should be considered in order to explain variations in poverty

Ravallion and Chen (1997) tested the claim that the incidence of poverty was more serious in the 1990s despite average income of countries increased at this time The data used is the household surveys data from 67 developing countries over the world

This study found no support for the argument that distribution tends to worsen as average living standard grows in developing countries Consequently, it rejects the Kuznets hypothesis Moreover, the evidences of this paper also showed that there was neither a correlation between growth in average consumption and inequality nor an underlying trend of inequality Because inequality was stable in 1990s, Ravallion and Chen (1997, p 380) concluded that “there is a strong association between the rate of growth in average living standards and the rate at which absolute poverty fell”

The Kuznets hypothesis was also tested and rejected by an empirical study of Bruno

et al (1998) with various type of data including cross-country, time series and panel data In the first part of the testing process, the authors used cross-countries data set conveys 44 developing countries With this data, they found no evidence of an inverted U-shape relation between income inequality and growth In the next stages of the testing, Bruno et al (1998) employed panel data from Deininger, Squire, and Zhang (1995) and time-series data from Ravallion and Datt (1996) The results show that “92 per cent of the variation in Gini indices by country and date is accounted for

by cross-countries variation whereas only 7 per cent is accounted for by variation over time” (Bruno et al., 1998, p 5) In short, neither type of data indicates the inequality-growth relationship suggested by Kuznets (1955) The rejection of the Kuznets hypothesis leads to an important conclusion that there does not exist any systematic trend for income inequality to increase or decrease with growth (Bruno et al., 1998) Hence, on average, absolute poverty will fall with increase in income In other words, economic growth was considered the main tool for fighting poverty

Dollar and Kraay (2002) shared the same view with the authors mentioned earlier of Kuznets hypothesis They rejected Kuznets hypothesis with the argument that they found no evidence for that the relationship between growth and inequality is statistically and significantly different in developed and developing countries Moreover, the results of Dollar and Kraay (2002) shows that the share of income of the poor does not change systematically with changes in income Strictly speaking, one can say that the gains of the poor and non-poor from increase in income were held

to be constant over time, based on the results of Dollar and Kraay (2002) Consequently, income of the poor increases as average income of a country grows, provided the poverty line does not change Therefore, as Dollar and Kraay (2002) emphasized, growth is important for poverty reduction

In summary, many authors have paid their attention on the Kuznets hypothesis when trying to explore the interrelationship between poverty, growth and inequality In this section, some researchers, for instance Fields (1989), Dollar and Kraay (2002), Squire (1993), Ravallion (1995), Ravallion and Chen (1997) and Bruno et al (1998), have employed various data in order to test the Kuznets hypothesis They have obtained the same result, which is the rejection of the hypothesis Based on the rejection, they argued that instead of changing systematically with growth, inequality is rather relatively stable over time The rejection of the Kuznets hypothesis allows researchers

to conclude that the poor actually gains from the increase in aggregate income proportionately Based on this conclusion, researchers have emphasized that economic growth is an important source of poverty reduction The next section will discuss about the second side in the debate that have concentrated on the two components of changes in poverty, growth component and distribution component

2.3 Growth and Redistribution Components of Poverty Reduction

In the second side of the debate on the impacts of growth and inequality on poverty reduction, researchers have decomposed changes in poverty into income (growth) and distribution (inequality) components The quantification of poverty reduction has been developed by Datt and Ravallion (1992), Kakwani (1993) and White and Anderson (2001) As will be discussed below, these authors have applied the similar methodology of the decomposition However, the results have been quietly different when considering whether growth or inequality played the major role in poverty alleviation

A theoretically quantitative method of decomposing changes in poverty into growth and redistribution components was developed by Datt and Ravallion (1992) This study defined the “growth component” as the change in poverty due to a change in income holding distribution constant, and the “redistribution component” as the change in poverty due to a change in distribution holding income constant (Datt & Ravallion, 1992) The decomposition allows the authors to measure thoroughly the contribution of distributional changes and growth to poverty reduction In the article, Datt and Ravallion (1992) then applied the decomposition method to analyze poverty reduction in Brazil and India Comparing changes in poverty in the two nations in the 1980s, this research indicated that poverty reduction in India was mainly contributed

by decrease in inequality By contrast, in Brazil within the same period, poverty reduction was largely due to economic growth However, the method of decomposition may face a problem resulted in by defining the growth and redistribution components in condition that distribution and income are stable relatively Thus, the method cannot explain such a case of growth with falling inequality

Using the same measures of poverty, inequality and growth, Kakwani (1993) constructed a similar theoretical methodology to decompose a change in poverty into

a change in average income and a change in income inequality, called growth component and distribution component The impacts of each component on poverty reduction was quantified separately in this method in order to explore whether growth

or better distribution (in the view of the poor) is the primary cause of the reduction in poverty Relying on the breakdown of change in poverty, Kakwani (1993) developed the formulas for calculating the growth elasticity and inequality elasticity of poverty The method was then applied in exploring the poverty reduction of Cote d’Ivoire in

1985 This paper calculated the growth elasticity and inequality as follow: with the 91,394CFAF (the Cote d’Ivoire national currency unit) poverty line, the growth elasticity was -2.87 and the inequality elasticity was 7.86; with the 162,613CFAF poverty line, they are -1.54 and 1.70 relatively (Kakwani, 1993) These results were calculated in case of the headcount ratio measures of poverty In case of other measures, the author obtained similar results, means higher inequality elasticity in comparison with growth elasticity (in absolute value) In short, despite the levels of poverty line and the poverty measures, it is undoubted that poverty in Cote d’Ivoire in 1980s is more sensitive to changes in income distribution than to economic growth From a different point of view, White and Anderson (2001) saw poverty as in a general picture about the pattern of growth This paper supported the “growth versus distribution” dichotomy For poverty reduction analysis, White and Anderson (2001) concentrated on the growth of the income of the poor and constructed a decomposition analysis based on the growth function of the poor’s income To examine the discrimination between growth and distribution in terms of poverty reduction, the income growth of the poor was broken down into the impacts of overall growth and of changes in income distribution The growth regressions for each quintile of income distribution were then estimated in order to achieve the

contribution of each impact to the increase in income of the poor Although in more than a quarter of cases changes in inequality played a more important role, White and Anderson (2001) indicated, in a total of 143-countries cases, the growth impact dominated the distribution impact in contributing to reduce poverty This finding conflicts with the result obtained by Kakwani (1993), which stated that the more important role in poverty reduction was played by the distribution component The confliction between these results can be explained by the difference in data used, Cote d’Ivoire in Kakwani (1993) and global data in White and Anderson (2001)

This section can be summarized as a “growth versus distribution dichotomy” in the explanations of poverty reduction The authors in this line of researches have decomposed changes in poverty into growth and distribution effects Separately, the researchers have found different major sources of poverty alleviation due to the different data Kakwani (1993) argued that the role of decrease in inequality dominated, whereas the growth effect was found to dominate by White and Anderson (2001) Similarly, Datt and Ravallion (1992) found that in Brazil the major role in poverty reduction was played by growth and in India, it was played by distribution The final section of the literature review will deal with the final side in the poverty-reduction debate, which aims to oppose the growth versus distribution dichotomy

2.4 The Impact of Inequality on the Responsiveness of Poverty to Growth

In the third side of the debate over growth, inequality and poverty, researchers made

an agreement with the first side mentioned in the first section, that income growth played a stronger role in poverty alleviation However, what makes the difference is that researchers, for instance Ravallion (1997), Ravallion (2001), Heltberg (2002), Adams (2004) and Fosu (2008, 2009, 2010a, 2010b, 2010c, 2011), argued that inequality also matters because it rather affects the extent to which growth can reduce poverty than impact directly on poverty

Ravallion (1997, p 3) introduced a regression function concerns with the argument that initial inequality can affect the rate of poverty reduction, called “growth elasticity argument” This paper tested the hypothesis that the responsiveness of the rate of poverty reduction to growth would become smaller as inequality increased and poverty would be unresponsive to growth if inequality were sufficiently high Using

data from 23 developing countries, the evidence indicated that the tested hypothesis could not be rejected Hence it is reasonable to state that “higher inequality tends to entail a lower rate of poverty reduction at any given rate of positive growth” Ravallion (1997, p 7) Moreover, the article found that the growth elasticity of poverty decreases sharply as inequality rises Consequently, initial inequality does influence how much will the poor gain from growth, or by how much will income of the poor increase as aggregate income increases The higher initial inequality is, the less strong the impact of growth on poverty reduction is

The question of how inequality matters the growth-poverty relationship was fundamentally analyzed by Ravallion (2001) Employing data from household surveys, Ravallion (2001) also declared that the poor actually share in the benefits of increasing average income of a country However, the differences in initial inequality, both global and national, resulted in differences in how much the poor gain from the income expansion (Ravallion, 2001) In this study, the effect of inequality on the responsiveness of poverty to growth was explored as twofold Firstly, the evidence tends to support the argument that inequality is harmful for growth Thus, lower inequality is expected to encourage economic growth and hence to reduce poverty due

to the increase in income of the poor gained from aggregate income growth Secondly, based on the estimated growth elasticity of poverty, the paper found that the elasticity decreases considerably as the extent of initial inequality increases As a result, at any given rate of income growth, countries with rising inequality experience the lower rate of poverty reduction, whereas in countries with more equal distribution

of income, poverty declines at a higher rate This result is similar to that obtained by Ravallion (1997)

Heltberg (2002) theoretically analyzed the poverty-growth-inequality interrelationship

by using a different methodology - synthesizing studies in this field Similar to many other researches, this paper traced economic growth as the main engine for poverty reduction The research also rejected the decomposition of changes in poverty into growth and distribution components, or the growth versus distribution dichotomy Despite growth is considered a necessary condition for poverty alleviation, the level

of inequality and its changes also matter (Heltberg, 2002) However, the impact of inequality cannot be in the same position with economic growth This is due to inequality strongly effects the growth elasticity of poverty, and initial inequality

contracts economic growth, which is the major source of poverty reduction These results are nearly close to that found by Ravallion (2001)

The impact of inequality on how poverty is responsive to income growth was also found by Adams (2004) In this paper, under the assumption of relatively constant inequality, the author aimed to estimate the elasticity of poverty respect to growth and

to identify the sensitivity of the results to different measures of economic growth What makes this study interesting is that Adams (2004) tends to relax the assumption

of stable income distribution by dividing the full sample of data into two sub-samples – low inequality countries (Gini index less than 0.4) and high inequality countries (Gini index more than 0.6) – then estimate growth elasticity for each sub-sample The results showed that the estimated growth elasticity of poverty is larger in the low inequality countries and smaller in higher inequality countries This finding does not change with different measures of growth Accordingly, Adams (2004) concluded that with a given rate of growth measured by either changes in mean survey incomes or changes in GDP per capita, low inequality countries would be more efficient in poverty alleviation than high inequality countries

In his studies, Fosu (2008, 2009, 2010a, 2010b, 2010c, 2011) constructed the closely similar specifications for analyzing the interrelation between poverty, growth and inequality Based on these specifications, the papers derived the formulas to calculate the growth elasticity of poverty Different data sets were applied for each of these papers – for example Sub-Saharan African countries in Fosu (2008), comparative global data in Fosu (2010c), African economies in Fosu (2010a), etc – for testing “the effect of income distribution on the ability of growth to reduce poverty” Fosu (2010b,

p 1) The same result was obtained from these papers: the responsiveness of poverty

to income growth is a decreasing function of inequality or initial level of inequality Thus, the policy implication for poverty reduction in almost cases is not just the concentration on economic growth because the more equal income distribution can accelerate the poverty-reduction ability of growth

In summary, the growth versus distribution dichotomy has been rejected in this line of researches The papers surveyed in this section have emphasized the influential role of growth in poverty reduction On the other hand, the important thing is that they have focused on how inequality can affect the growth-poverty relationship In general,

these authors obtained the same result, which is the negative correlation between inequality, as well as its initial level, and the growth elasticity of poverty (in absolute value) Hence, they have suggested that the policies in favor of poverty reduction should concern about economic growth and inequality together

This section brings out a debate in which researchers argued about the role of growth and variations in inequality in poverty reduction There are three arguments and three corresponding group of researches discussed in this section Relying on the arguments

of this section, the next section will construct an analytical framework based on which the empirical study will be implemented

2.5 Analytical Framework

The interrelation between poverty, growth and inequality can be graphically shown in the following figure

Figure 2.1 Analytical Framework

Source: Author’s analysis

This analytical framework includes two channels as follows:

Chanel (1) is based on Fields (1989), Squire (1993),Ravallion and Chen (1997), Bruno et al (1998) and Dollar and Kraay (2002)

Chanel (2) is based on Ravallion (1997, 2001), Heltberg (2002), Adams (2004) and Fosu (2008, 2009, 2010a, 2010b, 2010c)

As obtained by many researchers, growth can be expected to reduce poverty However, as argued by (Ravallion, 1997, 2001), Heltberg (2002) and Fosu (2008,

2009, 2010a, 2010b, 2010c, 2011), the ability of growth to reduce poverty depends

strongly on the level of inequality In thesis, economic growth will be expected to reduce poverty, based on the negative estimated growth elasticity of poverty On the other hand, inequality will be anticipated to decelerate the poverty-reduction ability of growth, relying on its negative impact on the poverty elasticity with respect to income

2.6 Chapter Remark

Researchers have not made agreement about how poverty can be reduced This chapter represents a debate among economists about how economic growth and inequality can affect poverty There tends to be three main groups of researches surveyed in this chapter The first group of researches has argued that poverty can be reduced almost entirely by economic growth This result was obtained from the rejection of the Kuznets hypothesis, which discusses that there is an inverted U-shape relation between economic growth and income inequality Then, because inequality does not change with changes in income, economic growth can raise income of the poor In other words, one can say that growth can reduce poverty, providing that inequality is relatively stable over time Authors in the second group have claimed that changes in poverty are due to growth as well as inequality independently These researchers have constructed a theoretical method to decompose the variance in poverty into income effect and distribution effect Their view can be known as

“growth versus distribution dichotomy” The empirical studies in this group have found different results about whether growth or better distribution (lower inequality) impact more strongly on poverty reductions Finally, some recent papers in the third group have indicated that the extent to which growth can reduce poverty depends heavily on the level of income inequality This final argument has debated about the assumption of constant income inequality suggested by the first side and the growth versus distribution dichotomy suggested by the second side They have argued that better income distribution, in the view of the poor, can magnify the absolute value of the growth elasticity of poverty This means that inequality also matters because in affect the ability of growth in poverty reduction, not because of its direct impact on poverty The empirical evidences have indicated that there is a negative correlation between the growth elasticity, in absolute value, and the level of inequality

The next chapter will present the methodology of the thesis This will explain how this thesis can use collected data to analyze empirically the roots of poverty reduction

in the South-Eastern Asian countries

Chapter 3 RESEARCH METHODOLOGY

This chapter will concern with research methodology of the thesis Firstly, it will present the empirical models that will be employed to answer the research questions mentioned in the first chapter The second section of this chapter will serve three research hypotheses that are going to be tested in the thesis Next, the third section will provide the variables included in the empirical models as well as the way that they can be measured To implement the empirical model, the penultimate section of this chapter will bring out the data collection method, the type of data and the data sources The final section will discuss about the estimation method This section will also consider the choice between the alternative regression models

3.1 Model Specification

At beginning, it should be better to define clearly the meaning of growth elasticity of poverty In this research, growth elasticity of poverty is the elasticity of poverty with respect to income, which measures how much will poverty change in percentage terms when income increases by 1% Formally, it can be calculated as:

P

Y Y

P Y

Fosu (2009) constructed a model that is interested in the growth rate of poverty, income and inequality, as below:

g Y a y G a g a y a a

p 0 1  2  3 I  4 I (3.1.1),

where p is growth rate of poverty measures, y is growth rate income, g is growth rate

of Gini coefficient, and G I and Y I are relatively the initial level of Gini coefficient and

income In this equation, a 0 is intercept; a 1 and a 2 are the relatively independent effect

of y and g on p; a 3 is the interactive effect of y with G I on p; and a 4 is the interactive

effect of g with Y I on p

The first research objective of this thesis is to identify the extent to which growth can affect poverty This means to estimate the growth elasticity of poverty Based on equation (3.1.1), the growth elasticity of poverty can be calculated as:

I

E  1 3 (3.1.2)

To estimate the growth elasticity of poverty, this formula is applied, relying on the

estimated coefficients a 1 and a 3 , and G I obtained from sample mean Because an

increase in income would reduce poverty, the coefficient a 1, which is the independent

impact of y on p, is expected to be negative The value of the estimated elasticity is

also expected to be negative due to this reason

The second research objective of the present thesis is to analyze how inequality can influence the poverty-reduction ability of growth Equation (3.1.2) allows this thesis

to consider the first argument about this relationship According to Ravallion (1997,

2001) and Fosu (2009, 2011), the initial level of inequality (G I) can influence the growth elasticity of poverty Equation (3.1.2) indicates the dependence of the growth elasticity of poverty on initial inequality Because higher initial inequality can reduce

the elasticity in absolute value (Ravallion, 1997), the coefficient a 3 is expected to be positive

Additionally, the sign of coefficient a2 is anticipated to be positive because higher

inequality would decelerate poverty reduction; and the expected sign of a 4 is also positive, as argued by Fosu (2009, p 729), “at a higher initial income, acceleration in inequality would map a larger proportion of the income distribution into poverty increase”

This thesis also employs another model to operate the empirical research Fosu (2010b) developed a different model in this field, which employs the logarithms of poverty, income and inequality, as below:

G Y b G b Y b b

In this equation, P is the poverty measure, Y is the level of income, and G is the level

of Gini coefficient The coefficients in this equation are interpreted as b 0 is intercept;

b 1 and b 2 are relatively the separate effects of lnY and lnG on lnP; and b 3 stands for

the interactive impact of lnY and lnG on lnP It is clear that the expected sign of b 2 is positive because higher inequality can increase poverty when income growth is zero

To achieve the first research objective, the growth elasticity of poverty can be derived from equation (3.1.3) as in the following formula:

G b b

E P  1 3ln (3.1.4) Based on equation (3.1.4), the growth elasticity of poverty can be predicted with the

estimated coefficients b 1 and b 3 , and lnG is the logarithm of the sample mean value of

G Again, because growth is anticipated to reduce poverty, the coefficient b 1 is expected to be negative and the value of the estimated elasticity is expected also to be negative

For the second research objective, the present research intends to analyze how inequality can influence the poverty-reduction ability of growth To achieve this goal, this thesis aims to explore the impact of inequality on the growth elasticity of poverty The argument is that the growth elasticity of poverty depends significantly on inequality, relying on Heltberg (2002), Adams (2004) and Fosu (2008, 2010a, 2010b, 2010c) Because higher inequality can entails a lower growth elasticity of poverty (in

absolute value), the expected sign of coefficient b 3 is positive

This section is constructed based on the linkage between literatures on the impact of growth and inequality on poverty reduction and the analytical framework constructed

in the previous chapter As emphasized, growth undoubtedly does reduce poverty by increasing income of the poor Almost researches has agreed about this argument, for example Fields (1989), Kakwani (1993), Ravallion (1995), Bruno et al (1998), Adams (2004), etc However, there are two arguments about the way to analyze the impact of inequality on poverty-reduction ability of growth On one hand, higher initial level of inequality can decelerate the impact of growth on poverty, for example Fosu (2009, 2011), Ravallion (1997) and Ravallion (2001) This argument will be tested by using model (3.1.1) On the other hand, inequality can reduce the growth elasticity of poverty, for instance Fosu (2008, 2010a, 2010b, 2010c) This argument will be tested by employing the model (3.1.3) Strongly based on this analysis, the next section will state the research hypotheses

3.2 Research Hypothesis

As discussed in the previous section, income growth by itself can reduce poverty This is due to it increases income of the poor at the same time with income of other members in the society This argument, along history, was supported by many researches in this field, for instance Fields (1989), Squire (1993), Ravallion (1997) , Adams (2004), etc Therefore, this thesis makes a hypothesis, which states that poverty measures and income growth are negative correlated

The argument about the impact of inequality on the responsiveness of poverty to income growth is mentioned as twofold in the previous section Firstly, based on Fosu (2009) and Ravallion (1997), the initial inequality does reduce the absolute value of the growth elasticity of poverty Hence, the second research hypothesis of the thesis is that the initial level of inequality negatively influences the growth elasticity of poverty Secondly, inequality can also affect the growth elasticity of poverty; higher inequality may entail the lower growth elasticity of poverty in absolute value This discussion was supported by, Heltberg (2002), Fosu (2010b), etc As a result, the final research hypothesis of this section mention about the negative correlation between inequality and the growth elasticity of poverty

In summary, there are three research hypotheses that will be tested in the thesis:

Hypothesis 1: There is a negative correlation between growth of income and changes

in poverty

Hypothesis 2: There is a negative correlation between initial level of inequality and

the growth elasticity of poverty in absolute value

Hypothesis 3: There is a negative correlation between inequality and the growth

elasticity of poverty in absolute value

The summary of expected signs of coefficients, which closely related to the research

hypotheses, are listed in the following table

Table 3.1 Summary of expected sign of coefficients

Specification (3.1.1)

Specification (3.1.3)

Source: Author’s analysis

In this section, the research hypotheses are stated clearly based on the model specifications The next section will consider the measures of variables that appear in the empirical models

Headcount index (P0) measures the number people whose income is less than a

standard amount of money called poverty line Formally,

)(

11

N N

N P

N

i i



, (3.3.1)

where N P is the number of poor individuals of the total population (N), Z is the poverty line and I is an indicator function that equals 1 if y i < Z, and 0 otherwise.

Poverty gap index (P1) measures “the extent to which individuals on average fall below the poverty line” (Haughton & Khandker, 2009, p 70) In other words, it measures how far the poor are from the poverty line In practice, the poverty gap index is usually expressed in percentage term The formula of poverty gap index is

G N

P

1 1

1

(3.3.2)

where G I Zy i I y i Z : the poverty gap

Squared poverty gap index, or poverty severity index, is mathematically simply the

weighted sum of poverty gaps with the weights are the relative gaps themselves This measure is calculated as:

1

1),(



From this essential formula, it can be easy to derive the three popular poverty measures If α equals 0, the poverty measure becomes P0, which is simply headcount index Similarly, if α equals 1, the poverty measures becomes P1 that is poverty gap index, and the squared poverty gap index, P2, is associated with α equals 2

3.3.2 Inequality

This thesis measures inequality by using Gini coefficient According to Haughton and Khandker (2009), there exists many measures of inequality, for instance Gini coefficient, Atkinson index, Theil index, etc Nevertheless, this thesis chooses the Gini coefficient because of two reasons Firstly, it is the most popular measure of

inequality Secondly, as Haughton and Khandker (2009) argued, Gini coefficient satisfies many criteria of a good inequality measures It does, with four criteria: mean independence, population size independence, symmetry, and Pigou-Dalton transfer sensitivity

The Gini coefficient bases on Lorenz curve Hence, to calculate Gini coefficient, one must have a Lorenz curve Lorenz curve is a cumulative frequency curve that shows the distribution of income, in comparison with the uniform distribution that presents perfect equality An illustrative Lorenz curve, which based on data for Vietnam, is shown in the following figure

Figure 3.1 Lorenz Curve

Source: Haughton and Khandker (2009)

Based on the Lorenz curve, Gini coefficient is calculated asGiniA/AB Two extreme cases arise If A equals zero, Gini coefficient becomes zero, which implies perfect equality On the other hand, if B equals zero, Gini coefficient becomes 1, which implies complete inequality Haughton and Khandker (2009) also presented a formula to calculate Gini coefficient formally, which is:



 N x i x i y i y i Gini

1

1 1

where x i and y i are two points on the x-axis and the y-axis relatively, and N stands for the number of intervals on the axis If N equal intervals are equal to each other, the

formula (3.3.6) becomes:

 

  



 N y i y i N

3.3.3 Income

To measure income, this thesis uses the survey mean income, or consumption, from household surveys Because some countries survey income while others survey consumption in their household surveys and because income of the poor can be only enough for attaining a minimal living standard, economists usually make an assumption that poor households use all their income for consuming Therefore, either survey mean income or consumption can be used as a measure of income in this thesis The use of survey mean income or consumption to measure income is similar

to other researches, e.g Adams (2004) Thus, economists usually use the word

“survey mean” to mention about either the mean income or consumption Despite it does not capture other aspects of living standard, the following two reasons can explain the use of survey mean in studies on the field of poverty Firstly, as mentioned

in the previous chapter, this thesis concentrates on monetary aspect of poverty As a result, in poverty analysis, it is reasonable to employ survey mean as a proxy of income Secondly, in almost researches about poverty, scientists popularly use survey mean as a measure of income For example, survey mean were employed in Adams (2004), Ravallion (1997, 2001) and Fosu (2008, 2009, 2010a, 2010b, 2010c, 2011)

3.3.4 Initial Inequality and Initial Income or Consumption

This thesis measures initial inequality and initial income as follows The initial level

of inequality is simply the first observation for inequality Similarly, the initial level

of income is the first observation for income

3.3.5 Growth Rate of Poverty, Income and Inequality

It is also noticeable that the growth rate of poverty, income and inequality are needed

to apply the model (3.1.1) Therefore, this thesis aims to obtain these numbers by

treating them as the growth rates of each variable in relative countries In detail, the formulas of the growth of income, inequality and poverty are as follows:

%1000

Y

Y Y growthY

%1000

G

G G growthG

g

%1000

P

P P growthP

p

where Y it , G it and P it are respectively income, Gini coefficient and poverty measures

for country i in year t; and Y i0 , G i0 and P i0 are respectively income, Gini coefficient

and poverty measures for country i in the first observed year

The measures of variables are already discussed clearly in this section In summary, this thesis uses Gini coefficient to measure inequality, survey mean (income or consumption) to measure growth, and three most popular indices including headcount ratio, poverty gap and squared poverty gap to measure poverty The next section will mention about the method of collecting data that includes data source and countries and years of surveyed data

3.4 Data Collection

The dataset is collected from the source of ‘PovcalNet: the online tool for poverty measurement developed by the Development Research Group of the World Bank’ The URL for this source is http://iresearch.worldbank.org/PovcalNet/index.htm?2

From this source, the downloaded data includes the countries in the South-Eastern Asia region of which the household surveys are available However, because different countries have implemented their surveys in different years, this thesis actually obtains unbalanced panel data Additionally, in the data source, the most recent surveys of the countries were performed in 2010 Consequently, the dataset of this research contains 58 observations from 8 South-Eastern Asian countries in different

years within the period 1981-2010 The list of observed countries and the dataset are detailed in Appendix A1 and A2

3.5 Estimation Strategy

This section provides the estimation methods related to the model specifications and collected data discussed in the previous sections It will consider alternative regression models associated with collected panel data of this thesis, including the constant coefficients model (Pooled model), the fixed effects regression model (FE model) and the random effects regression model (RE model) In addition, this section will implement a test for choosing the most appropriate regression model that will be applied in order to obtain the research results

3.5.1 The Constant Coefficients Model

According to Johnston and DiNardo (1997), constant coefficients model, or pooled model is the simplest panel data model that goes on by essentially disregarding the structure of panel data The model is given by:

it it

Beside the advantages, the pooled model has its own drawbacks Although makes the model simple and easy to perform, the assumption of independently and identically distributed error term seriously ignores the panel structure of data Hence, in spite of its simplicity, pooled model is often inappropriate for analyzing panel data

3.5.2 Random Effects Regression Model

Different from the pooled model where T observations on n individuals are treated as

nT observations on different individuals, the random effects (RE) model does consider the panel structure of data The structure of RE model is as follows:

It is crucial to emphasize that in the RE model, the time-invariant individual effect, α i,

is assumed to be uncorrelated with x it, cov(i,x it)0 According to Johnston and DiNardo (1997), the OLS regression will not be efficient although it yields consistent

estimates of β This is due to it underestimates the standard errors

Therefore, to deal with this problem, the method of feasible generalized least squares (FGLS) is required in order to produce the consistent and efficient estimates of coefficients The solution involves two steps The first step is to derive an estimator of the covariance of matrix of error term The second step is to use this covariance

matrix in the estimate of β

3.5.3 Fixed Effects Regression Model

Similar to the RE model, the fixed effects (FE) model is structured as:

it it

y   , where it i v it What makes the FE model be distinguished from the RE model is that in the FE

model, the time-invariant individual effect, α i , is assumed to be correlated with x it,

0

)

,

cov(i x it 