Income inequality and economic growth panel data in south east asia

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM HO CHI MINH CITY INSTITUTE OF SOCIAL STUDIES VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS INCOME INEQUALITY

UNIVERSITY OF ECONOMICS ERASMUS UNVERSITY ROTTERDAM

HO CHI MINH CITY INSTITUTE OF SOCIAL STUDIES

VIETNAM – THE NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

INCOME INEQUALITY AND ECONOMIC GROWTH:

PANEL DATA IN SOUTH EAST ASIA

BY

NGUYEN LE PHUONG LINH

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, November 2016

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

VIETNAM THE NETHERLANDS

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

INCOME INEQUALITY AND ECONOMIC GROWTH:

PANEL DATA IN SOUTH EAST ASIA

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

NGUYEN LE PHUONG LINH

Academic Supervisor:

PROFESSOR TRAN TIEN KHAI

HO CHI MINH CITY, November 2016

ACKNOWLEDGEMENT

Foremost, I would like to express my sincere gratitude toward my thesis advisor, Professor Tran Tien Khai for his valuable guidance and encouragement for my thesis His comments helped me for the writing of this thesis

I would also like to thank Professor Truong Dang Thuy who gave me necessary guidance on data processing techniques and was willing to give me a hand whenever I have difficulties from my thesis I am gratefully indebted for his worthy advises

Furthermore, I would also like to thank all lecturers, staff and my VNP 20 classmates at the Vietnam Netherlands Program

Finally, I would like to send my deepest thanks toward my family members, my parents,

my older sister and younger brother who have always been by my side, and encouraged me throughout my studying years and through the time of writing this thesis This achievement would not have been possible without them

ABSTRACT

This study has examined the impact of different levels of economic growth toward income inequality, applying the Kuznets reverted U-shape model and the Tribble S-curve model by employing the annual panel dataset from 1990 to 2012 at the country-level in the Southeast Asia New econometric technique Driscoll and Kraay is handled with the panel regression technique to correct for the autocorellation, heteroskedasticity and cross-section dependence The analysis finding reveals the significant relationship between income inequality and economic growth Furthermore, the empirical evidence also provides support for the existence of Kuznets inverted-U as well as inverted S-shaped curve in eight countries

in Southeast Asia

Key words: Income inequality, economic growth, Kuznets curve, S-curve, Southeast Asia,

Panel data

TABLE OF CONTENT

Acknowledgement ii

Abstract iii

Table of content iv

List of table v

List of figures vi

Chapter 1 Introduction 1

1.1 Problem statement 1

1.2 Research objective 3

1.3 Research question 3

1.4 Research scope 3

1.5 Thesis structure 3

Chapter 2 Literature review 5

2.1 Arguments on the concept of Inequality 5

2.2 Theoretical literature 6

2.2.1 The Kuznets theory 7

2.2.2 Evolution derived from Kuznets theory 8

2.2.2.1 Kuznets theory in term of technology process 8

2.2.2.2 Kuznets theory in term of financial development 10

2.2.3 The S-curve 10

2.3 Empirical literature 13

2.3.1 Empirical papers support the growth-inequality relationship 13

2.3.2 Empirical papers against the growth-inequality relationship 14

2.3.3 Determinants of inequality 15

2.3.3.1 Structural shifts 15

2.3.3.2 Education and labor skills 16

2.3.3.3 Population growth 17

2.3.3.4 Institution 18

Chapter 3 Research methodology 19

3.1 Data source 19

3.2 Research methodology 22

3.2.1 Model specification 22

3.2.2 Measure of income inequality 23

3.2.2.1 The Lorenz curve 23

3.2.2.2 The Standardized World Income Inequality Database 24

3.2.3 Estimation strategy and correction model 26

3.2.3.1 The model of Pooled regression 26

3.2.3.2 The model of fixed effects estimation 27

3.2.3.3 The model of random effects estimation 28

3.2.3.4 The choice of panel regression model 29

3.2.3.5 Driscoll and Kraay standards and correction model 30

Chapter 4 Empirical results 31

4.1 Overview of the income inequality in Southeast Asia 31

4.1.1 Current status 31

4.1.2 Pattern of change 32

4.2 The data descriptions 33

4.2.1 The descriptive statistic 33

4.2.2 The possible relationship by scatter 36

4.3 Panel data regression 37

4.3.1 Diagnosis analysis 37

4.3.2 Empirical results and analysis 39

Chapter 5 Conclusion and policy implication 42

5.1 Concluding remarks 42

5.2 Policy implication 43

5.3 Limitations and directions for further research 44

Reference 45

Appendix A: Regression results 49

Appendix B: Results of Breusch-Pagan LM test 55

Appendix C: Results of Hausman test 56

Appendix D: Modified Wald Test for groupwise heteroskedasticity 58

Appendix E: Wooldridge test for autocorrelation in panel data 59

Appendix F: Pasaran test for Cross-section dependence 60

LIST OF TABLES

Table 3.1: Variable description 21

Table 4.1: Current level of income inequality in Southeast Asia 32

Table 4.2: Summary statistics 33

Table 4.3: Measures of GDP per capita and income inequality in Southeast Asia over 5-year period from 1990-2012 35

Table 4.4: Variance inflation factor (VIF) 37

Table 4.5: Model comparison 38

Table 4.6: Diagnostic problem 38

Table 4.7: Panel regression 41

LIST OF FIGURES

Figure 2.1: The Kuznets’ model 8

Figure 2.2: Lewis’ dual sector model 16

Figure 3.1: The Lorenz’ curve 23

Figure 4.1: Gini coefficient index in Southeast Asian from 1980 to 2012 33

Figure 4.2: Relationship between income inequality and economic development 36

in Asia, which indicated an increase in inequality (worsening in Gini index) during the 1990s and the 2000s, including PRC, India and Indonesia, where have the highest population of the region Besides that, many of the Asian countries also experienced the rapid spread in expenditure share of top 5% and 1% income classes, implying that the advantage groups in the society is getting richer in a much faster way Eventually, in the universal point of view, although gini coefficient in developing Asia is still in an average level 28-51, compared to the income inequality in Latin America the Caribbean and Sub-Saharan Africa, 45-60 and 30-66 respectively, the wider gap of income inequality in those Asian countries is such a considerable issue while the inequality in some Latin America has been remedied over recent years

There are several reasons why economic growth should be concerned with inequality

On the economic point of view, the inequality is not purely a political object but also considered as an important factor, which matters for growth and poverty in long run Some evidences show that countries with higher level of inequality will hinder the economic growth, thus leading to the increasing level of poverty as well as contributing to the higher inequality degree of education, health and so on Furthermore, inequality is an important factor leading to crime, social unrest or violent conflict There is empirical evidence that income inequality weakened the investment and growth by filling up social discontent, increased political instability (Alesina and Perotti,1996)

In fact, the most concerning object in all economic debates is the development of the economy as a whole or the economic growth particularly On the other hand, inequality or income distribution can negatively affect the physical and human capital accumulation, which is the driving force to growth Therefore, it is obvious that poor income distribution

may impede the economic growth in the long run Eventually in order to improve the living standard and optimize social welfare as a whole, it is important to look into the effects of economic growth on income distribution A desirable economic growth will reduce the income disparity as well as lower poverty rate leading to a stable society

From the current situation of the economy and society, inequality is one of the most worthy issues in economics, yet very few studies are paid certain concern about the effect of economic growth on income inequality, especially for a particular region such as South East Asia, where the inequality is increasing recently, which may worsen economic growth in the long run

There have been some empirical studies regarding to the relationship between growth and income inequality It is generally believed that the level of the economic growth is cited

as a remedy for improving the poor situation (Adam, 2004; Dollar &Kraay, 2002) However, given evidence that current pattern of growth and globalization are raising income distribution gap, which hinders the poverty reduction rate (Ravallion, 2001) According to Kuznets (1955), growth plays an important growth to inequality by the concept of reverted U-shaped relationship In the initial phase of development, increasing economic growth will lead to the increase in inequality, but in latter stage, the inequality will reduce with the upward speed of growth

For those reasons listed, this research will go further to examine the relationship between economic growth and inequality nexus focusing in Southeast Asian region combining of 8 countries from 1990 to 2012, the period of new technologies and institutional transition associated with rapid economic growth The result is able to suggest some policy implications for the decision-making process

Even though there have been a number for publications in public domain, this study is different with other previous studies conducted in the following key areas: (i) explore the effect of economic growth on income inequality in a new context of sub-region - Southeast Asia - where encountering the increasing income inequality while achieving a faster and higher level of economic growth; (ii) because of the shortage of income inequality data over

a long time period, most of empirical studies about growth and inequality have relied on cross-section regression as a framework for their analysis, and have not had further researches for particular regions, this study will apply the Driscoll and Kraay estimation in

the panel regression to capture the systematic differences in behavior over time and space; (iii) examine the Kuznet reverted U-shaped hypothesis with two economic sectors and the S-curve hypothesis with the addition of service sector

1.2 Research Objectives

The purpose of this study is to examine the influence of economic growth toward the income inequality in the pattern of the panel data Accordingly, the two main objectives have been considered as following:

(i) Analyzing the potential impact of economic growth on inequality in Southeast

Asia within the context of the Kuznet reverted-U-shape hypothesis

(ii) Examining this relationship with higher level of growth, testing the S-curve

(ii) Is the relationship of income inequality and economic growth consistent with

the hypothesis of Kuznets reverted U-shape?

(iii) In the higher level of economic growth, does the reverted U-shape still hold?

1.4 The Research scope

This study has analyzed the research objectives by using the 8 countries in Southeast Asian region in the period 1990-2012, where encounters a higher income inequality with a higher level of economic growth Data are collected at the countries level

1.5 The research structure

The composition of this thesis includes five critical chapters as following:

Chapter 2 presents the necessary theories and the empirical researches involving with the theories

Chapter 3 explores the research methodology, building the models, explain the data measurement, and introduce the applied econometric technique to answer the research question

Chapter 4 analyzes the results of the research from the econometric technique used, as well as explains the findings in the comparison with other empirical papers and reports Chapter 5 summarizes discussions, proposes possible policy implications, indicates the limitations and suggests directions for further studies

Chapter 2

Literature Review

This chapter will present the relevant theories and the related empirical studies, as well

as analyze different levels of growth on inequality the “reverted U” shape and the “S-shape” The structure of this chapter will be determined as following:

i The theoretical literature capturing the crucial theories in the study

ii The empirical researches providing the empirical evidence for the related theory; iii The hypothesis construction answering for research questions

2.1 Arguments on the concept of Inequality

Whenever refers to the concept of inequality the comparability between elements is always the first perspective mentioned The comparison rests on particular measurable aspects using equal indexes or indicators According to Cowell (1995), a straightforward concept of inequality “obviously” suggests a departure from simple idea of equality, meaning that two or more quantities are the same size

Nonetheless, the disputations often occur whenever confining the concept of inequality into the relation with economic context and social pattern In this situation, inequality is frequently related to the discrepancy of income, consumption or wealth, which demonstrates some sort of perception about fairness, so it eventually raises different ethical viewpoints and arguments It is also controversial when mentioning the measurement concern Sen (1973) generally categorized the inequality into two kinds of measures: statistical measurement and ethical valuation The statistical measurement aims to obtain the extension of inequality from

an objective viewpoint, using numerical measurement of comparable alterations of income, while ethical valuation tries to measure inequality in term of some normative notion of social welfare Still, it is problematic that there is not a clear definition specifying the two approaches

Moreover, the further controversy even goes beyond this point It is earlier mentioned that inequality is about the comparison of the living standard of each individual or household

in a particular society The problem is that there have not been agreements among scientists

so far about the exact definitions of the standard of living of an individual and the way to measure it Thus, there has not been a single index which is able to fully describe the living standard To overcome this issue, some researchers commonly consider income, consumption and wealth as proxies for living standard in the literature of inequality but none

of them fully cover the health, happiness or even freedom aspect Therefore, we can see that the living standard of an individual or household is, in fact, a multidimensional concept involving, in principal, every aspects of direct consumption and non-consumption activities and services, Gallo (2002)

As discussed above, inequality is a multidimensional concept, then there are additional social indicators are exploited to support for the lack of income or expenditure The Equalities Review (2007d: 18) identifies ten dimensions of inequality, including longevity, health, education, physical security, standard of living, individual, family and social life However, when using a broad category of indicators to describe inequality as the different variables may result in ambiguous answers Hence, in order to make the comparison possible

in all diversified population, it is really worthy to have a standard indicator in a uniform scale This is why most of the researches so far mainly focus on income as key measure of inequality because of its sufficient of source of data and it can somehow assess the inequality

in a broader aspect of society Therefore, the discussion given in this paper will focus on the income inequality

Finally, there is also slightly different between two terms “inequality” and “inequity” Both of them all refers to the unequal condition but inequity is usually considered under qualitative circumstances, while inequality is generally related to things that can be asserted

in numbers Thus it is probably stated that inequality is a necessary but not a sufficient condition for inequity to exist For these two reasons, this paper will be conducted mainly with income inequality

2.2 Theoretical literature

The concept of inequality is broad and there have been different aspects about the theory of the impact of growth on inequality, so this paper will focus on the theory of the most well-known and commonly believable, the Kuznets hypothesis His theory, then, is developed further by many researchers even in theory and empirical studies

2.2.1The Kuznets Hypothesis

The Kuznets (1955) curve plays the initial and then becomes the main theoretical approach to assess the determinants of inequality and development level It is well-known for the “inverted U” shape inferring to the relationship between income and inequality according

to which the level of inequality would first increase and then decrease with economic growth The most basic theory for the systematic relationship between inequality and income was the migration-based model provided by Kuznets (1955) The theory was assumed that there are only two sectors in the economy: agriculture and industry Initially the agricultural and rural sector contributes a large portion in the economy, characterizing by low income and comparably low degree of inequality Whereas, the industrial and urban sector is proportionally small at the beginning, have high income per capita and, high inequality rate within the sector

The economic growth is triggered by the movement of persons and resources from agriculture to industry The persons who move obtain a higher salary, thus raising the degree

of inequality The increase of labor forces in agricultural sectors changing to industry makes the size of the small rich group in the industrial sector bigger and bigger Therefore, at the very early stage of the development, the economic growth rate and level of inequality have positive trend

As the diminishing of agricultural sector and continuing urbanization, more of the poor agricultural workers are now enable to join up with those in the relatively rich industrial group At the same time, within the same sector many workers starting at the bottom of industrial sector tend to climb up with those of the higher income Apart from that process, the size of the agricultural labor force gradually diminishes, which leads to the increase in relative wage in agriculture and rural sector These movements eventually reduce the degree

of inequality Hence, at the later stage of development, the growth rate of income tends to be negative with the level of inequality

Figure 2.1: The Kuznets’ model

The weakness of the model is that there is no mentioning about the productivity level in two sectors as well as there is no explanation about the human capital accumulation that would affect wages, which leads to the migration decisions of the individual

Then his idea was developed further by Robinson (1976), doing main research about the migration from agriculture to industry Following the Initial Theory of Kuznets, large number of empirical papers strongly support for the Kuznets hypothesis using cross-section evidence such as Ahluwalia (1976), Papanek and Kyn(1986), Bourguignon (1994), Milanovic(1995) and Jha(1996) and also got refuted by others

2.2.2 Evolution of inequality derived from Kuznets Theory

2.2.2.1 Kuznets Theory in term of Technological process

The relation between the level of output per capita and inequality can be explained in another approach in term of technology evolution The poor sector employs an old and outdated technology; meanwhile, the modern and rich sector own more advanced techniques (Aghion and Howitt, 1997; Galor and Tsiddon, 1997; Helpman, 1997) Galor and Tsiddon (1997) developed a novel endogenous mechanism that generated the inverted-U relationship between income inequality and output per capita

The mechanism developed in this study was confined to a small open economy with perfect capital mobility This theory supposes that the economy operates in a competitive market, which is characterized by perfect information, homogeneous good and efficiency

labor in the production process The level of human capital is accumulated from generation

to generation which will advance the degree of production technology The main force triggers the development process is the initial distribution of human capital At the early stage of development, the inequality in human capital plays the main role to enable the highly educated class of society overcome a low stable equilibrium , while the equality may trap the whole society at a low level of investment in human capital Thus, inequality essentially increases the human capital aggregation and total output in the early stage of growth The higher segment of the society move to the higher-investment stationary equilibrium, leading

to its increase of income, then widens the inequality in income between the higher and lower human capital groups Furthermore, because the average level of human capital as a whole increases, the output increases This pattern follows the first stage of Kuznets curve, the growth rate increases along with the increase of income inequality

The increase in the level of human capital generates the technological process, increasing the wage of each efficiency labor in the economy as a whole Lowered human capital group will take this benefit to increase their investment in human capital Owing to diminishing return to the family specific external effects as well as human capital, the lower segment will have higher rate of investment Hence, the two groups move toward the same dynamic system and converge to the same higher steady state where aggregate output is higher and wage income is identical among groups Thus, compliance with Kuznets “reverted U” shaped, in the first period of development the increase of output level will accompany with the increase of inequality, then in the later stages of development the economic growth will associate with more equal distribution of income and human capital

In this context, the introduction of technological innovation, such as electrical power, computers and internet has the tendency to increase the inequality at the early stage At the beginning, the number of people in the relatively high income technology advanced sector is very few, then more and more people gradually transfer toward this newly flourish sector to get higher income, the inequality tends to rise with the expending output product The inequality will decrease and equalization point reach when relatively few people left in the backward sector and those newly move to more advanced sector are able to meet up with who is already in this setor At the end, the relative rage of those staying in the backward sector increase because of the decrease of the supply factors in this sector

2.2.2.2 Kuznets Theory in term of financial development

Many recent models have generated the Kuznets curve within the explanation between income distribution and growth under the development of the financial environment (Greenwood and Jovanoic, 1989) The dynamic progress under the financial system from a financially uncomplicated environment to a modern financial structure corresponds with the shift of people and resources from the agricultural sector to the industrial sector In the early stages of development, the financial markets in the economy are basically absent and gradually start their first step to grow Until the intermediate stage of the development process, the picture of the financial framework is initially shaped, which characterizes by the increase of both economic growth rate and saving rates accompanying with the widen gap between the rich and the poor When the development progress reaches the mature stage, the financial intermediation has rapidly extended its structure in the economy Finally, the economic growth rate jumps to a higher pace than that in its infancy period, along with the stabilization of income distribution and the fall of saving rate

2.2.3 The S-curve

The S-curve is obviously an approximate hypothesis as Kuznets’ inverted U-shaped Actually an economy in the reality does not strictly have two sectors of agriculture and industry, even non-socialist or non-OPEC countries are not rigorously dualistic Besides that, the income inequality is affected by several other elements than the degree of economic growth Therefore, there have been, recently, several empirical evidences suggesting that the extension in the later stage of the Kuznets’ U-curve is actually the S-curve (List and Gallet, 1999; Tribble, 1996, 1999) The S-curve hypothesis supposes a multiple structural turning points in the economic growth process Hence, even the inverted U-shape curve hypothesis is

an illustration of the secular nature of economic growth It may be concluded that the association between income inequality and economic growth is likely different to a critical level as economic activity continues

It is controversial that Kuznets’ inverted U shape is unfinished illustration of the economic economic process In fact, the previous studies were mostly limited by cross-sectional database at a particular time period or panel dataset with limited time span, which mainly defined in the agricultural to manufacturing structural transition (Kuznets, 1955, 1963; Paukert, 1973; Ahluwalia,1976; Jha, 1996) while Kuznets (1955) definitely indicated that the

economic growth problem is the lack of time-series within the context of income distribution and economic growth rate relationship (Kuznets, 1955) Fortunately, the more recent studies have generally exploited a longer time period, thus allow the possibility to observe a greater likeliness of the evolution of the manufacturing to services among different countries with high-level of economic growth (Ram, 1997)

Tribble (1999) provides a basis theory to explain the S-curve hypothesis as a feasible extension of Kuznets’ hypothesis as well as suggests two crucial turning points of the relation

of the economic growth and income inequality According to Tribble (1999), because of the nature of the economic growth cycle, the S-curve will illustrate a structure of multiple turning points, whether it is applied within a strictly times-series, cross-sectional or panel database He proposes that the key force generates the turning points is the critical shifts in the share of income of upper and middle classes relative to the income share of lower classes

In particular, in the early period, under the S-curve regime, the relative income share of the middle and upper groups increase accompanying with the decline of the share of lower income groups The first structural turning point occurs with the following decline in the comparative share in the upper and middle classes income and the expansion of the share of the lower income class, which is happened as the “trickle-down” effect This process characterizes for the transition from the agricultural to manufacturing sector The second turning point occurs after that when the economy starts with the transition of manufacturing

to service sector under the same mechanism with the prior one This process is highlighted by the resource reallocation of upper and middle groups from the manufacturing to newly developed service sectors The critical driver of this structural movement is the trend of savings and investments of the upper and middle income groups in the society

The economic structural shift associated with the turning point in the S-curve hypothesis is, on the other hand, explained by the change in intrasectoral and intersectoral inequality (among agriculture, manufacture and service sector) Prior to the first critical turning point presented with agricultural to manufacturing sector shift, the inequality rate in manufacture exceeds that of in agriculture Moreover, growth rate of income per capita in the manufacturing sector accelerates the rate of such growth in the agricultural sector (Kuznets, 1955) Besides that, Lewis (1954) suggested that agricultural productivity are lower than manufacturing productivity as well as agricultural prices is decreasing relative to

manufacturing prices both in intrasector and intersector Generally the favorable trade of manufacturing over agriculture and the movement of resources and human from agricultural sector to industrial sector continue until the surplus of agricultural labor is fully absorbed into the manufacturing sector After the completion of labor absorption, the growth rate and inequality achieve at the highest level of the critical turning point, the wages in the agricultural sector increases when inequality begins a period of decline (Lewis, 1954)

The explanation for the second critical turning point suggests that the growth in per capita income of service sector exceeds that of manufactures as well as the the level of inequality within the newly emergent service sector is higher than that of in manufacturing sector Therefore, terms of trade in service sector improves with respect to both manufacturing and agriculture (Tribble, 1996) This discrepancy will lead to the movement

of labor from manufacturing to service sector and the process only finishes when the labor in the manufacturing sector is fully absorbed into the service sector Because the industrializing production has been globalized, the labor absorption in this case will be widespread in a greater manner than in the situation of agriculture during the agricultural to manufacturing structural change

The formulation of the S-curve hypothesis could be presented as follow:

GINI=β0+ β1 PGDP+ β2PGNP2+ β3PGNP3+μ With β1 > 0, β2 < 0, β3 > 0 and |β1|>| β2|>| β3|

2.3 Empirical literature

2.3.1 The empirical papers support the growth- inequality relationship

There have been many controversies among researchers about the relationship between the economic growth and income distribution Kuznets (1955)’s first idea about the relationship between economic development and income inequality is a reverted U-shape curve This hypothesis was strongly proved by the empirical research of time series date in three countries of England, the United States and Germany He concluded that this relation is positive at the early of the development and then negative in the latter stage of this process

As more and more workers shift from the agricultural sector to the industrial sector, the agricultural labor supply is decreased thus driving up wages in this sector Whereas persons move to the industrial sector later obtaining the income of the richer workers

Finally, wages in both traditional and modern sectors increase, which leads to the decrease of inequality (Kuznets, 1955)

After Kuznets there are many empirical papers to contribute to his hypothesis Ahluwalia (1976) conducted across-section analysis on the relation between inequality and economic growth with a sample of 60 countries including 40 developing countries, 14 developed countries and 6 socialist countries, the result gives out a same pattern with the Kuznets “reverted U” hypothesis The research also provides the same result with expended regression model including the additional explanatory variables reflecting different aspects of economic growth, which seems to explain well some of improvements in income inequality observed in later stage of economic growth However, the cross section results are not strong enough to support the view that a faster growth is systematically associated with higher inequality in any given development level

Latter, Barro (2000) have suggested that international data show a clear empirical phenomenon of the Kuznets curve His study indicated that income inequality first increases then consequently declines along with the economic growth The result of the study is illustrated significantly by the majority of panel data over time and cross-section The result

of Kuznets curve is reasonably stable from the 1960s through the 2000s Similar study was conducted by Lessmann (2013) using a panel dataset to explore the relationship between spatial inequalities and economic growth in 56 countries in period 1980-2009 by apply parametric and semiparametric technique The result strongly supports for the reversal U-curve of Kuznets (1955)

2.3.2 Empirical papers against the growth- inequality relationship

Papanek and Kyn (1986) concluded there is not a significant relationship of economic growth and income distribution, even if it held the Kuznets’ hypothesis, the result would weaken over time To remedy the weakness of cross-sectional and time-series data, some recent studies have exploited the panel data Anald and Kanbur (1993) have tested the robustness of inequality and economic growth relationship based on the estimates of Ahluwalia (1976) and Ahluwalia et al (1979) with analyzed to differences in functional forms and data collection They found out that with different forms, imply with different conclusions for the income inequality and economic growth relationship Thus, the result of the U-shape might be very sensitive to the choice of the functional forms, some equations

support the reverted-U curve hypothesis but not for all Besides, they also built a consistent data set, but results are reversal with the conventional “reverted U” shape

Besides that Ravallion & Chen (1997) have ascertained that economic growth can reduce or increase the income inequality equally Additionally, Dollar and Kraay (2002) states that the poor will benefit proportionally with the economic growth rate of average incomes, which promotes the idea that growth may not have any significant impact in the income distribution Furthermore, Deininger and Squire (1998) found that only 5 countries (10% of samples) has inverted U-shape, the rest has the U-shape Most of the case

is contrast with Kuznets Moreover, there is no evidence of Kuznets’ reverted U-shape in the results of fixed-effects in the observed model The result is also same with Perera and Lee (2013) On the other hand, Lyn and Yeh (2009) have found out growth has positive effects on income inequality, and vice versa

Although there have been endless studies examining the linkage between growth and inequality but most of them employed the overall worldwide dataset and very few focusing

on the particular effect of homogenous Southeast Asian countries Hence, it needs a more researches examining the effect of growth on income inequality, especially in Asian region

2.3.3 Determinants of income inequality

Ahliwalia (1976) analyzed the specific mechanism of economic growth factors to affect the degree inequality from availability of explanatory variables

2.2.1.1 Structural shifts

Intersectoral shifts accompanied with economic growth have long been recognized as a possible factor explaining for impact of the development or economic growth on inequality Kuznets (1963) argued that the development process involves with the rapid growth in the high income of modern sector, which slowly absorbs labors from low income of traditional sector This process of development will lead to an increase in relative income distribution in the early stage of development then reduce in the later stage, which make the precise U-shape result

Ahluwalia (1976) indicate that the average incomes ratio between sectors may also follow a U-shaped pattern, which income disparity between two sectors widen in the early stages, then narrow in the later stage of development It is possible that in the early stage, the urban sector is in shortage of capital and resources leading to the low productivity and low

income level in the rural sector, thus leading movement of population from rural to urban The urban sector, then, gradually expand along with the lower pressure on population than that of rural sector Consequently, the capital becomes less scare, and more resources improve the productivity in the low income sectors in the later period of development and reduce the income differentials between sectors

Because of the absence of data on inequality within each sector which are the mean of income and population shares, Ahluwalia (1976) has used two explanatory variables to capture the impact of intersectoral shifts on the inequality These are the agriculture share in GDP and the urban population ratio The agricultural share in GDP is expected to fall and urban population ratio is expected to increase with the economic development as the modern sector grows at rapid rate and the fully absorbs population from traditional sector

This structural change could be expressed in the Lewis’ dual Sector model as follow: The model assumes that economies have dual sector: large agricultural sector characterizes as inefficient labor employment, low productivity, low income and low savings and small industrial sector identifies as high productivity, high income and high saving The higher wage in industrial sector attracts the workers to migrate into the cities until there is no income difference in these two sectors

Figure 2.2: Lewis’ dual Sector model

Source: IB-Economics

2.2.1.2 Education and labor skills

Another mechanism through which the economic growth affects the inequality is the improvement in the educational characteristic and skill endowments of the labor force, which will reduce the income equality rate in the long run It is supposed that the improvement of high skilled labor will make a lift of low paid and unskilled employment to high paid and skilled employment under the condition of the technology development combining with the traditional marginal productivity theory This change will generate higher labor incomes, narrow skill gap, as well as increasing the share of wages in total output Apart from that, unlike physical capital the expansion of human capital accumulation in the economy importantly produces a diffusion effect to a wiser generation of population Therefore, in term of technological assumptions about the factor productivity, it is disputed that the pattern

of skill intensive development cannot lead to the concentration of income like that of capital intensive development, as human capital cannot be accumulated in a single person, and at any rate it cannot be bestows across generations in the same way as physical capital works Ahluwalia (1976) examines the education-equality relationship using three explanatory variables, the literacy rate, the primary school enrollment ratio and the secondary school enrollment ratio The results prove that education positively affects the relative equality, meaning that higher education can lower the income gap The results are also widely in same patterns with the pervious cross-section empirical evidence of Adelman and Morris (1973)

2.2.1.3 Population growth

It is assumed that the rapid increase of population growth rate will accelerate the income inequality The relationship between population growth and the degree of inequality has been received relatively less attention The theory of this mechanism is built under the empirical observation of cross country experience The regime implies that within the same level of economic growth, the income inequality will worsen off in a faster population growth economy when it reaches a given level of per capita income compared with a slower growth of population The literature suggests two mechanisms It is possible that different distribution in income groups characterizes with different growth rate, same as with the high population rate characterizes with divergence in population across groups, which finally leads to greater inequality For instance, the poorer income group will have lower income growth rate as well as higher speed rate population than that of richer group, which make the

income distribution gap wider and wider It is well-known that the process of economic growth generates a “demographic transition”, meaning that the declines in the growth rate of total population will occur with economic growth, and this process will narrow the intergroup differentials in population growth Another possibility to explain the relationship between population growth and income distribution is that higher growth rates of population imply greater pressure of labor supply on other productive factors causing the decline in the share

of a person in total output For instance, in the case of fixed factor such as land - a faster population growth generating a higher population density has burden higher rental share on

population, thus make the inequality increase

2.2.1.4 Institution and inequality

Chong and Gradstein (2007) by applying GMM system methodology in a cross-section panel have proved that there is a significant causal linkage between institutions and inequality, meaning that institution causes inequality and vice versa, which is consistent with the theory Furthermore, they also analyze the size effect of the contribution of each type of causality by employing a simple VAR method The result suggested that the direction impact

of inequality to institutions appears to have stronger influence than the reverse causality and this outcome is robust in different institutional measures, sample size and inequality measurement Gupta, Davoodi, and Alonso-Terme (2002) also reinforce the effect of institution on inequality that reduction in corruption will lead to the improvement of inequality In order to avoid the causal impact of inequality on the corruption, the study has employed the instrument variable of democracy index of countries, and this result is hold under sensitivity analysis

indicate that the influence of growth on inequality is positive by a pooled dataset from 83 countries observed from 1965 to 2003 Therefore, it is reasonable for expecting that the development level will have positive relationship with the income inequality in Southeast Asian countries

H2: The income inequality first increases then decreased along with the level of economic development

Kuznets (1955) stated that within the economy with two sectors, the income inequality first increase and then decrease with the economic growth This process is illustrated by the inverted U-shape Kuznets’ hypothesis has been confirmed in a number of empirical studies with different econometric methods, for instance Ahluwalia (1976), Papanek and Kyn(1986), Bourguignonand Morrison (1990), Bourguignon (1994), Jha(1996), Barro (2000) and Desborders and Verardi (2012) Recently, the paper of Desborders and Verardi (2012) has contributed to the reversed U-shape hypotheisi by employing Baltagi and

Li (2002)’s semi-parametric fixed effects regression estimator For the listing researched papers, income inequality expected to increases and then decrease with the development level of economy

H3: The economic development and income inequality is a quadratic equation, first increases then decreases and finally increasing again along with the economic growth

Actually an economy in the reality does not strictly have two sectors of agriculture and industry, Tribble (1996, 1999) proposed an S-curve hypothesis as an expansion of the Kuznets U-curve identifying two critical turning points in the inequality and economic growth relationship, from agriculture to manufacture and from manufacture to service sectors Sinha (2005) proved the relationship between economic growth and inequality in India during the period 1980-81 through 1997-98 is in fact displays S-curve Tribble (1996) has examined the inequality and economic growth relationship by applying the time series data for the economy of the United States and the findings is possible to explain by the S-curve hypothesis rather than Kuznets’ inverted U-curve Hence, it is highly believed that with the extension of the service sector, the economic growth and income inequality nexus in Southeast Asia is the S-curve shape with two critical turning points

Chapter 3

This chapter will mention the research methodology in following aspects:

i The review of the data collection for examining the economic growth and income

inequality relationship

ii The model specification

iii The research methodology specifies for the applied econometric techniques then

latter answers for the research objectives

3.1 Data Source

This study employs an annual panel dataset from 1990 to 2012 using country-level data

of the Southeast Asia sample The original sample consisted of eleven countries, however, the observation from Brunei Darussalam, Myanmar, and Timor-Leste have been removed due to the lack of the dependent variable Gini index and explainable variable the GDP per capita, respectively Therefore, the final sample covers 8 countries in the region (Kingdom of Cambodia, Republic of Indonesia, Lao People’s Democratic Republic, Malaysia, Republic of Philippines, Republic of Singapore, Kingdom of Thailand, and Socialist Republic of Vietnam) The SEA region is considered to be a suitable choice of sample because many countries in this area have been achieving the high growth rate accompanying with high inequality level Furthermore, it is easier to observe the variation of income per capita along with the increase of economic growth of from high to low income countries by capturing the homogenous group and same characteristic countries

The mainly observed data Gini index comes from SWIID 2014 of Solft (2009) Data on GDP per capita, PPP (current international dollars) comes from World Bank, Civil liberty comes from Freedom House and the rest of essential data sources are collected from World Bank and United Nations The details of all analyzed variables being used in the estimation are illustrated in Table 3.1

Table 3.1 Variable description

Gini Index Index Income Inequality measurement, 0-100, higher value represents higher inequality SWIID 2014

Unemployment Percentage Unemployment refers to the share of the labor force that is without work but

available for and seeking employment

Inflation Percentage The annual percentage change in the cost to the average consumer WB 15 +

Civil Liberty Index Civil liberties index, 1-7 where 1 is extreme civil liberties and 7 is extreme no civil Freedom House +

Liberties Education index Index Education index is an average of mean years of schooling (of adults) and expected

years of schooling (of children), both expressed as an index obtained by scaling with the corresponding maxima

Population 15_64 Percentage Ratio of people between the ages 15 and 64 to total population UN + Population density People/sq

km of land area

Labor force ratio (F/M) Ratio Ratio of female to male labor force participation rate (%) (modeled ILO estimate) UN -

Details of sources: WB 15, World Bank 2015; UN, United Nations; SWIID 2014, Standardized World Income Inequality Database.

3.2 Research methodology

Many previous studies that have examined the relationship between inequality and economic growth commonly criticized from three following aspects: the parametric form, the cross-sectional or panel of the test and the measurement of income inequality This section will discuss what the previous studies was conducted; then suggest a possible model and dataset to have a closer look on the inequality-economic growth relationship

3.2.1 Model specification

Because of the limitation of dataset in time series, most of the previous researchers used the cross-section data to capture the relationship between the income inequality and growth such as Ahluwalia, 1976; Anand and Kanbur, 1993 and JHA, 1996 However, cross-section analysis has a certain limitation because it is likely to suffer from omitted variables bias or the endogeneity problem Cross-section cannot capture the change of economic growth and income inequality over time Consequently, recent studies, for instance, Baro, 2000; Chamber, 2007; Desbordes, 2012, are able to exploit the full information of panel dataset even in cross-section and time dimension The panel data will give more information, more variability, less collinearity among variables, more degree of freedom and more efficiency Another advantage of panel datasets is their ability to control for individual heterogeneity Panel dataset are also better able to identify and estimate those effects that pure cross-section

or our time-series data cannot detect

Hence, this study will employ the panel date set to answer the research questions, the following model is considered in this empirical study:

Income inequality it = β 0 + β 1 Ln(GDP per capita) it + β 2 Ln 2 (GDP per capita)+ β 3 Ln 3 (GDP

per capita) + β 4 (Control variables) it + 𝜀 it

Where:

(i) Dependence variable: Gini coefficient index

(ii) Level of economic growth: GDP per capita - the natural log of GDP per capita to the

power of two and three

(iii) Control variables: Agriculture share, Industry value added, unemployment, Trade,

Inflation, civil liberty, Education index, Gross capital formation, log life expectancy, pop 15_64, pop density, and labor force ratio (female/male)

3.2.2 Measurement of income inequality

3.2.2.1 The Lorenz Curve

Figure 3.1: The Lorenz Curve

The Lorenz curve is the upright visual perception of the Gini coefficient, which is illustrated by the thick curve in the above figure The vertical axis illustrates the accumulative proportion of income, which matches with the unit on the horizontal axis The Gini index ratio is visually the division of the percentage of the area between the Lorenz curve and a ideal dash line and the maximum area under the dash line The proportion of the total population from the poorest to the richest is measured by the horizontal axis This it is obvious that the Gini coefficient get zero value when the income of the population is equally distributed and vice versa, in case the aggregate income gathers to only one person, leaving the rest with no income at all, then the Gini coefficient receives the value of 1, or 100% Basically, gini index is the most broadly index to express the inequality level

In the condition of perfectly equal income distribution as the each person or household

in the population receives the same level of income, the Lorenz curve becomes the dashed 45-degree line However, when income is distributed unevenly within the population; the income distribution of the poor is relatively lower than that of the rich population, the Lorenz curve is the thick curve below the 45-degree line When the income distribution level increases and increased, the thick curve will gradually concaves towards the bottom right-hand corner of the graph

3.2.2.2 The Standardized World Income Inequality Database

Data on income inequality of this study comes from the extensive compilation for a large panel of countries in Soft (2014) Studies regarding to the inequality have been encountered great barrier on data issues Hence, the study of Soft (2014) introduces the Standardized World Income Inequality Database (SWIID), which is a composition of comparable Gini indices including of gross and net income inequalities for 153 countries from 1960 to 2013

Researchers have spent many resources to compile the datasets on income inequality for a half century Among them, the two most influential projects are the Luxembourg Income Study (LIS) and the assembled dataset of Deininger and Squire (1996) but they have both advantages and disadvantages The LIS has introduced the most comparable data but covered relatively few countries and years, present data available for only 30 countries dating from 1993, almost of which are the world richest countries; whereas the Deininger and Squire dataset and its successors can generate more observations yet loss of comparability, as they are based on different income definitions such as gross or net and different reference units namely households or individuals

Because of the relative full observations, many researchers prefer to apply the Deininger and Squire data set, which is used by hundreds of cross-sectional studies The World Income Inequality Database (WIID), which inherits the Deininger and Squire (1996) dataset, is created by the World Institute for Development Economics Research of the United Nations University in 2008 Although the WIID have certain improvement to broaden the number of gini indices to 5314 observations in 160 countries, the tradeoff between comparability and coverage still remains Babones and Alvarez-Rivadulla (2007) has

comparability and coverage They calculated the average differences between various income definitions and reference units based from the original data of WIID to make a constant adjustment, which results in a unified data set representing household per capita gross income inequality Unfortunately, as this kind of constant adjustment is unable to capture the change of taxes and redistribution income in the society of the government, as well as the variation across countries and over time in the differences of income definition or reference unit, it may lead to the overestimation or underestimation of income inequality value in some cases Consequently, the analysis of the SIDD is still a cost of reliability and comparability Solf (2014) has further improvement of inequality measurement framework by constructing a Standardized World Income Inequality Database (SWIID) aiming to offer a broader statistic research by maximizing the comparability of gini index but still fully cover across countries and over time SWIDD was based on WIID data version 2.0c and constructed for following process: first, eliminate those observations that unlikely provide full data of a country's population; second, classify the data according to their reference unit and income definition, which finally results into a twenty-one category dataset of country-year observations Each pair of the available observations will generate a ρab ratio, which functionally provides fully set of data on all countries and years from the incomplete inequality variables For instance, missing observations in category A could be simply fulfilled by multiplying available data in B by ρab Nevertheless, this ratio ρab will change across country and over time as this adjustment is also change according to government policy, tax regulation as well as the trend in consumption and savings and other factors However, its reliability and validity have to be gone through assessments First, Solf (2014) examined the uncertainty in the SWIID estimates by the standard of errors and obtain fairly good results with about 30% of the observations with 01 or lesser point of standard errors, more than 60% and 85% of the standard errors with smaller than 2 points and 3 points, respectively Those statistical numbers suggest that the SWIID is not perfect but it is quite good Another assessment of the SWIID is a comparison of the relationship of gross and net income inequality between the countries of the developing world and the advanced industrial countries In fact, in most of the developing countries the tax is relatively low and the income redistribution policy is slightly effective The correlation between gross and net income inequality in developing, therefore, will higher than that of advanced countries This is true in

developing countries is 0.967 and 0.749 in advanced countries Lastly, Solf also take the assessment of SWIID by examine its relationships with various social indicators, such as life expectancy at birth and infant mortality rates

In general, SWIID represents a prior choice in the balance between comparability and coverage; hence, this paper will take the employment of the Gini index from SWIID as the most reliable dataset at current time

3.2.3 Estimation strategy and correction model

Distinguishing from other previous studies, which mostly exploit the cross-section data, this research will take the advantage of panel data to explore the relationship between the level of economic growth and income inequality Generally, the Pooled OLS, Fixed effects models (FEM) and Random effects model (REM) are common methods used in panel data

3.2.3.1The model of pooled regression

The common or the most restrictive model of panel data is described as follow:

Yit= α + X’it β + µit i= 1, , N t=1,…,T Where: i denoting individuals or household, and t denoting time

Panel models mainly differ on their assumption on µit In the pooled OLS, the error term

µit is assumed to identically and independently distributed across i and t, E(µ)=0 and var(µ)=σ2 It is also assumed that coefficient across time and cross-section remains constant, and efficiently estimated by least squared (OLS) Therefore, the pooled OLS method is also known as the constant coefficient model Specially, the pooled OLS has some important assumption as follow:

- The explanatory variables are non-stochastic If they are stochastic, they are uncorrelated with the error term In other word, the error is uncorrelated with the individual specific effects The variables and error term are assumed to be strict exogenous, meaning it does not depend on current, past and future values of the error term µit

- The error variance is homoscedasticity and no serial correlation

However, the assumption of standard error in this method is strict and sometime not realistic The possible problem is that there may have unobserved individuality of each

subject and it is included in the disturbance term µit As a result, the error term has the tendency to correlate with some of the explanatory variables included in the model In this case, the estimated coefficient in the model will be biased as well as inconsistent

3.2.3.2 The model of fixed effects estimation

The fixed effect model is as follow:

Yit= α + X’it β + µit i= 1, , N t=1,…,T

µit = µi+ υitYit= (α+ µi) + X’it β + υit

with the identifying condition ∑𝑁𝑖=1𝜇𝑖 = 0, and the individual effects are assumed as unobserved constant (parameters), is the Fixed effects (FE) regression model (υit) fulfills the usual condition on error: E(υit)= 0, varυit = σ2

Fixed effects model allow the intercept may differ across subject (here the eight countries), but it does not vary over time, or time-invariant, and the coefficients of the explanatory variables do not vary across individual or over time In this case, it allows the time invariant individual effect correlate with X’it, as the same time assume that X’it is not correlated with error term Hence, the fixed effect can control for omitted variables bias and remove the characteristics of time-invariant from the independent variables

To allow the intercept vary among of each individuals, it is commonly used the differential intercept dummy technique, which measure the characteristics of each countries through dummy variables Hence, fixed effects model also known as least squared dummy variables model (LSDV) With this technique, we will have different intercept for each country, also called one-way fixed effect If we allow for both individual and time effects into the mode, it will become two-way fixed effect model

However, it should be cautious with fixed effect model, because if too many dummy variables are introduced to the model, the degrees of freedom problem as well as the possibility of multicollinearity will occur In case the number of parameters goes to infinity, LSDV estimators will be inconsistent