Faculty of Economics and Business
The effect of income inequality on economic
growth in countries with different levels of
incomes
Bachelor Thesis
Laura Bouwman (11049103)
Economics and Finance
26-06-2018
Statement of Originality
This document is written by Laura Bouwman who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
The existing literature shows that income inequality plays an important role in the process of economic growth. However, the direction of the effect of income inequality on growth is still controversial. This research aims to examine the effects of income inequality on economic growth across countries with different levels of incomes by using panel data for a sample of 48 countries in the period 1980-2012. The yearly Gini coefficient has been used as inequality measure. Fixed effects and random effects estimation techniques are employed, followed by a generalized method of moments (GMM) technique in order to solve for endogeneity problems. Negative and statistically significant estimates are obtained for the effect of income inequality on economic growth independent of countries’ initial levels of income classification.
Keywords: income inequality, economic growth, Gini coefficient, Kuznets curve JEL Codes: O1, O4, I3
Table of Contents
1. Introduction ... 5 2. Literature review ... 6 2.1 Income inequality ... 6 2.2 Theoretical framework ... 9 2.2.1 Positive effect ... 9 2.2.2 Negative effect ... 10 2.2.3 Non-linear effect... 11 2.3 Empirical framework ... 13 2.4 Overview ... 14 3. Methodology... 16 3.1 Data description ... 16 3.2 The model ... 17 4. Results... 19 4.1 Empirical results ... 19 4.2 Interpretation of results ... 23 4.2.1 Control variables ... 23 4.2.2 Explanatory variables ... 245. Conclusion and discussion ... 25
6. References ... 26
7. Appendix ... 29
7.1 Descriptive statistics ... 29
7.2 Hausman specification tests ... 31
1. Introduction
In recent decades, income inequality has been rising globally. The richest 1% of the global population acquired 22% of the total global income growth since 1980, which is twice as much as the bottom 50% individuals who only received 10% of total income growth
(Alvaredo, Chancel, Piketty, Saez & Zucman, 2018). Furthermore, approximately 82% of all wealth created last year went to the richest 1% of the global population, while the bottom 50% saw no wealth increase at all (Pimentel, Aymar, Lawson, 2018). Not only has income inequality been rising globally, income inequality has also increased within many countries over the past few decades. Piketty (2015) argues that in these countries the rate of return on capital is greater than the economic growth over the long-term. The relatively higher capital income enables capital owners to accumulate more wealth which will likely result in an even greater unequal distribution of wealth in the future.
The increase in income inequality naturally gives rise to concerns about the macroeconomic impact of the increase in income disparities. The relationship between income inequality and economic growth has received a considerable amount of attention by economists over the past decades. The extensive literature on income inequality and economic growth demonstrates that the effect of income inequality on economic growth is still
controversial. Forbes (2000) states that the level of income inequality has a significant positive relationship with economic growth. Nevertheless, other economic theories suggest that income inequality is detrimental to an economy (Persson & Tabellini, 1994; Perotti, 1996; Sukiassyan, 2007).
Gaining insights into the direction and magnitude of this effect could benefit policy makers since they will be able to conduct a more effective fiscal policy in order to decrease income inequality and simultaneously stimulate economic growth. Therefore, the aim of this thesis is to attempt to clarify the extent to which income inequality is related to economic growth in countries with different levels of incomes. Barro (2000) has already examined the relationship between income inequality and economic growth rates for a broad panel of countries that varies by level of economic development between 1965 and 1995. However, over the past 20 years a lot has changed due to the ever-increasing globalization, the financial crisis, and even further increases in global income inequality. As a consequence, there may be more recent and relevant conclusions to be drawn with regards to the influence of income inequality on economic growth across countries with different levels of economic
income inequality and economic growth for different levels of countries’ initial incomes during 1980 and 2012, controlling for a set of relevant covariates.
This thesis is structured as follows. Section 2 consists of a literature review giving an overview of income inequality, as well as a review of earlier theoretical and empirical research on the effects of income inequality on economic growth. In section 3, the data sample and econometric models used in the quantitative research are described. The results of the econometric methods performed are presented and discussed in section 4. In the final section conclusions are drawn and potential further research is discussed.
2. Literature review
This section briefly reviews the concept of income inequality by discussing methods to measure income inequality as well as the development of income inequality. Subsequently, theoretical arguments aiming to assess how income inequality affects the rate of economic growth are provided, followed by an appraisal of existing empirical research.
2.1 Income inequality
A study on the relationship between inequality in income and economic growth cannot be conducted without consensus on the appropriate way to interpret and measure income inequality. This section briefly discusses the concept of income inequality, several common measures of income inequality and the evolution of income inequality within countries.
In this thesis, income inequality refers to the extent to which individual’s or
household’s disposable incomes are distributed in an unequal manner in a particular year and a given country.
The method used to measure income inequality is essential since small differences in the method employed can result in large disparities in the estimated effect of inequality on economic growth (Panizza, 2002). In order to further understand the effects of within-country income inequality, several ways in which income inequality can be measured are discussed. Charles-Coll (2011) reviews some methods which are widely employed in economic studies.
The simplest method to measure income inequality is by comparing different income groups that are ranked according to income quantiles. The most commonly used inter-decile ratio is the 20/20 ratio which compares the amount of income of the top 20% of individuals or households to the bottom 20% (Charles-Coll, 2011). However, according to Palma (2011) an income that lies between 40th and 90th decile stays relative stable and changes in the
incomes in both the highest and lowest segments. Hence, the Palma ratio is a preferred inter-decile ratio which compares the income of the top 10% of people to the income of the bottom 40% in a country (Cobham & Sumner, 2013). Significant shortcomings of the inter-decile ratio method are that a change in the amount of national income in the middle class is not reflected into the inter-decile ratio and that the measure does not fall into an absolute scale1.
Another (more complicated) method is the Theil index which uses the entropy measures based on information theory (Charles-Coll, 2011). More inability to distinguish a group of income earners from each other based on their resources means higher equality. The formula for calculating the Theil index is the following:
𝑇𝑇 = 𝑇𝑎=1=𝑁1∑ (𝑦𝑖
𝑦̅ 𝑁
𝑖=1 ∗ 𝑙𝑛𝑦𝑦̅𝑖)
Where yi is the income of individual i.
The Theil index ranges from 0 to 1, where a value of 0 reflects maximum inequality and a value of 1 perfect equality. Disadvantages of the Theil index include the conceptually and mathematically complexity as well as the lack of an intuitive graph in order to represent the inequality index (Charles-Coll, 2011).
The most widely used measure of income inequality is the Gini coefficient which is developed by statistician Corrardo Gini and based on the Lorenz curve. The Lorenz curve represents the relation between the cumulative proportions of the population of a country (x-axis) and the cumulative proportions of their income (y-(x-axis). Figure 1 (Mirzaei, Borzadaran, Amini & Jabbari, 2017, p.207) shows the Lorenz curve, the line of perfect equality as well as areas A and B. The Gini coefficient reflects the difference between perfect equality and the actual cumulative distribution of income divided over the total area under the perfect equality line. It is computed with the following formula (Charles-Coll, 2011):
𝐺𝑖𝑛𝑖𝑖,𝑡 = 𝐴 𝐴 + 𝐵
The value of the Gini index ranges between 0 and 1. A value of 0 means perfect equality, in which case area A disappears and the Lorenz curve coincides with the line of equality. A value of 1 indicates perfect inequality, in which case area A will cover the whole lower triangle. Because data on the Gini coefficient is widely available and relatively easy to understand, a lot of empirical researches use the Gini coefficient as a means to assess income inequality (Charles-Coll, 2011). Since differences in the method used could result in large disparities in the estimated effect of inequality, it is easier to compare this research with other
empirical researches if the Gini coefficient is used. Hence, the Gini index is used to indicate income inequality in a particular country at a particular time in this thesis.
Figure 1: the Lorenz curve and the line of perfect equality
Now that the concept and methods of measuring income inequality are explained, the development of income inequality in different levels of economic development is addressed. Kuznets (1955) suggests the Kuznets’ curve, which is illustrated by an inverted U-shape and denotes that as a country develops from a primarily agricultural society to an industrialized economy, inequality in the distribution of incomes first increases and later decreases. In developing countries, the vast majority of the workforce works in agriculture or other low-productivity sectors and hence income inequality is low. If a country transforms from an agricultural society to a high-productivity sector, income differences increase because jobs in the industrial sector generally generate higher incomes than those in the agricultural sector. As industrialization proceeds, the share of labour force in the agricultural sector tends to diminish and inequality begins to decrease. Eventually, a point is reached where almost the entire population of a country is included in the industrial sector, accompanied by high incomes, and income inequality tends to fall to a relatively low level (Kuznets, 1955).
Acemoglu and Robinson (2002) provide a political economic theory on the Kuznets curve. They argue that increasing development and simultaneously increasing income inequality may lead to political instability, which potentially enforces institutional changes that encourage income distribution and decline income inequality. Hence, development does
not necessarily follow the path of the Kuznets’ curve but may as well follow a democratic path, depending on the change of income inequality and the degree of political mobilization.
Nevertheless, income inequality in developed countries has increased since the 1980s and Kuznets theory became controversial. Piketty (2014) expressed criticism on Kuznets’ views and argued that government interventions are necessary to create income equality in developed countries. In his book “Capital in the 21st century” he proposes a progressive
taxation of labour income and inherited wealth as well as a rise of social transfers. Before implementing Piketty’s (2014) proposal to decrease income inequality, it is essential to gain insights into the effects of increasing income inequality on a national level on economic growth.
2.2 Theoretical framework
In this paragraph, several theoretical arguments on how national income inequality affect the rate of economic growth are analysed. Classical approaches predict that income inequality enhances economic development. However, in the last decades modern approaches have advocated a negative relationship between income inequality and economic growth.
2.2.1 Positive effect
Traditionally, there are two main approaches as to why income inequality has a positive effect on economic growth. According to the first approach, income inequality provides incentives to work. The second approach argues that inequality increases the aggregate savings rate which stimulates investments.
Firstly, the possibility to obtain higher incomes provides incentives for people to work harder and invest (Lazear & Rosen, 1981). Assuming that highly educated people are more productive and thus get higher incomes than average, more people may be encouraged to pursue higher education in order to become more productive, which increases aggregate productivity and thus economic growth (Cingano, 2014). Besides, income redistribution, to achieve a more equal distribution of income, is typically accompanied by a progressive income tax structure. This negatively affects human incentives, which is likely to result in reduced work effort and investments (Knowles, 2005; Biswas, Chakraborty & Hai, 2017).
Another explanation is derived from Keynes (1937) and was further developed by Kaldor (1955). According to this approach, the marginal propensity to save is an increasing function of wealth. High-income households tend to have a higher marginal propensity to save than income households. Redistribution income policies intended to favor the
low-income households reduce the aggregate savings rate in an economy and hence investments which is harmful for economic growth (Cingano, 2014). However, a plausible argument that mitigate this effect is also derived from the standard Keynesian model which demonstrates that low-income households tend to have a higher marginal propensity to consume than high-income households (Keynes, 1937). Income redistribution therefore increases the aggregate consumption in a country and thus economic growth. As a consequence, the total effect of redistribution of incomes on aggregate demand is debatable.
Even though the effect of income redistribution on aggregate demand is doubtful, redistribution of income changes the type of financing investments which affects economic growth. Bhattacharya (1997) developed a neoclassical growth model in which investments must be external or internal financed. However, external finance is provided by workers who are paid for labor and thus external finance is subjected to the standard costly state
verification. Redistributing income away from the rich harms the ability of the rich to internally finance their own investments. As a consequence, more monitoring costs appear due to the presence of the costly state verification and credit market conditions tend to
worsen. This effect declines capital accumulation, which is detrimental for economic growth.
2.2.2 Negative effect
Nowadays four principallines of thinking are being followed to provide an explanation as to how income inequality might negatively affect economic growth. These channels being endogenous fiscal policy, political instability, the presence of imperfect credit markets for financing investments in education and endogenous fertility.
The first channel is through endogenous fiscal policy. According to the median voter theorem, the tax rate selected by the government is the one preferred by the median voter since the probability to be re-elected is the highest. Greater income inequality and thus a relatively capital-poor median voter might cause a majority of voters to insist on income redistribution through a higher equilibrium level of capital taxation and regulation on capital. Governments tend to succumb under this pressure and produce policies that will rise taxes in investments and growth-promoting activities (Bertola, 1993; Persson & Tabellini, 1994. This reduces incentives to invest of foreign investors which slows down economic growth.
The second channel is through political instability. Even if the government does not give in to the demand for fiscal redistribution financed by an increase in the equilibrium level of capital taxation as mentioned above, economic growth could still decline. A plausible explanation is that inequality may lead to social unrest and political instability, which could
lead to a decrease in investments and thus to a decline in economic growth (Alesina & Perotti, 1996).
The third channel is through the presence of imperfect credit markets for financing investments in education. Galor and Zeira (1993) developed an equilibrium model of open economies. This model consists of two periods. In the first period, individuals may either invest in human capital and acquire education or work while being unskilled. In the second period, individuals form part of the skilled or unskilled workforce depending on their educational attainment (Galor & Zeira, 1993). Since individuals are under loan constraints, the inheritance of each individual determines whether an individual is able to invest in human capital. As a result, individuals with small bequests have less easier access to investment in human capital and may choose to become low-skilled even though the rate of return to the individual and the society is high (Cingano, 2014). Therefore, the greater the degree of income inequality, the lower the share of human capital in an economy. This reduces
aggregate output and therefore harms aggregate economic activity. Note that the effect of this channel is weaker if education is being financed by the state, which is partly the case in most developed countries (Odedokun & Round, 2004).
The final channel is through endogenous fertility. Morand‘s (1999) theoretical
framework demonstrates that parents only invest in the education of their children if their own human capital is above a certain threshold that is derived from the costs of raising a child and the return of investments in human capital. The positive relationship between parental human capital and the level of investments in children’s human capital may very well lead to a path of persistent economic growth. However, if the human capital level of parents is below the threshold, parents prefer to have a greater amount of children over investing in the future of a few, and the economy is locked in a poverty trap with low initial capital and high fertility. Greater income inequality means an increasing number of people living below the threshold. This would increase the fertility rate of a country instead of human capital and thereby reduce economic growth.
2.2.3 Non-linear effect
Several theories predict that income inequality can affect economic growth in either a positive or negative direction. However, given the complex relationship between income inequality and economic growth, there is no explanation that is fitting for all stages of economic development. Additionally, it is important to note that the effects of income inequality through some channels depends on the stage of economic development of a country. Galor
(2000) argues that the positive effects of income inequality may offset the negative effects in developing countries, while the positive effects of income inequality are dominated by the negative effects in developed countries.
The positive effects of income inequality on economic growth through an increasing aggregate savings rate may be particularly noticeable in early stages of economic
development. This can be explained by the fact that physical capital is scarce in low-income countries, causing a higher rate of return of physical capital than rate of return of human capital (Galor, 2000). As a consequence, physical capital accumulation tends to stimulate economic growth in low-income countries. Income inequality transfers resources towards individuals with a higher marginal propensity to save and therefore physical capital accumulation increases enhancing the process of economic development (Galor & Moav, 2004).
However, as a country develops, the rate of returns on human capital rises due to capital-skill complementary. Hence, human capital accumulation is a prime source for
economic growth (Galor & Moav, 2004). Since individual’s investments in human capital are subjected to diminishing marginal returns, it is optimal for a country’s growth to equalize the marginal returns of human capital across individuals because that maximizes then aggregate returns to investment in human capital (Galor, 2000). In the presence of imperfect credit markets for financing investment in education, redistribution of income (e.g. greater equality) therefore positively affects aggregate level of human capital and economic growth. Besides, the wage differences across individuals within developed countries decrease, resulting in lower differences in the marginal propensity to save across individuals and a decline in the positive effect of income inequality on growth through greater aggregate savings rate (Galor, 2000).
The findings of Galor (2000) are consistent with the Kuznets’ curve which implies that income inequality rises in early stages of economic development and declines in later phases of development. Nevertheless, it is important to note that in Morand’s framework (1999) the decision to have children is not based on parental altruism but on the old-age-support motive, whereby children are seen as the only asset for parents to secure an income for their old-age. The old-age motive is only likely to apply commonly in developing countries. Greater income inequality in developing countries therefore increases the fertility rate of the countries and slows down economic growth. Hence, income inequality has a negative effect on economic growth in developing countries, which is inconsistent with Kuznets’ curve. Whether the effect
of an increasing aggregate savings rate or the effect of increasing fertility rate supersedes the other is not clear.
2.3 Empirical framework
Despite the extensive theories that have been developed as to why income inequality is negatively, positively or non-linearly related to economic growth, there remains considerable disagreement on the strength of each effect. In addition, empirical research attempts to
examine the direction and significance of the effect of income inequality on economic growth. This paragraph briefly discusses several findings of empirical research.
Persson and Tabellini (1994) define a model that relates economic growth to income inequality and democracy using the income share of the third quintile as a measurement for income equality. A large variety of countries that covers the 1960-1985 period is included in the sample and they examine the effect using Ordinary Least Squares (OLS) and 2 Stages Least Squares (2SLS) as econometric estimation techniques. Results show that income inequality in general has a negative and significant effect on economic growth caused by redistributive fiscal policy. When taking into account whether a country is democratic or not, a negative but non-significant effect is estimated for non-democracies and a negative but significant effect is measured for democracies. Democratic countries are also often developed countries and therefore one can conclude that a negative and significant effect can be
measured for developed countries.
However, as this research is extended and other explanatory variables are included, the evidence for the effect of income inequality on growth caused by fiscal redistributive policies becomes less convincing (Perotti, 1996). The effects of the other channels are estimated using panel data of 64 countries observed for 1960-1985 and a 2SLS technique. The data reveals that income inequality causes political instability and unequal societies tend to have higher fertility rates and lower rates of investment in education. These factors lower rates of investment and therefore growth. Empirical research of Odedokun and Round (2004) in African countries identifies the same channels through which inequality negatively affects economic growth.
Forbes (2000) criticizes the results of Perotti (1996) and states that country-specific and time-invariant omitted variables result in a negative bias in the estimated inequality coefficient. She uses different estimation techniques such as fixed effects, random effects, Chamberlain's π-matrix technique and the generalized method of moment (GMM) in order to measure the effect of income inequality on GDP per capita growth in 45 different countries
between 1965 and 1995. By focusing on GMM, Forbes (2000) corrects for the negative bias and suggests a positive effect of income inequality on economic growth within countries in the short- and medium term. Additionally, she also obtains positive and similar values of the estimated coefficient in the other used estimation methods. These findings are supported by Li and Zou (1998) who perform cross-sectional regressions with democratic dummy variables and time-specific dummy variables. If the Gini coefficient increases by one standard deviation for the fixed effects and random effects model, the rate of economic growth tends to increase by 0.45-0.48% and 0.33-0.35% respectively.
However, it is controversial whether the found empirical results indicate towards an overall effect of national inequality on growth that is positive or negative. Growing empirical literature concludes that the relationship between inequality and growth can better be
characterized as a non-linear relationship. Barro (2000) researched a broad panel of countries and showed a roughly zero relationship between income inequality and growth. His research has been extended by focusing on countries at vastly different levels of economic
development. The empirical analysis suggests that higher inequality tends to delay growth in developing countries and encourages growth in developed countries. For countries with a GDP per capita below $2070 (1985 U.S. dollars) the effect of income inequality on economic growth is negative while the inverse is true for countries with a GDP per capita over $2070. The results of this analysis are inconsistent with the findings of Galor, (2000) who argues that income inequality is positively related to growth in developing countries and negatively related in developed countries. Barro’s (2000) outcome is explained by the fact that individuals in low-income countries are more exposed to credit-market constraints and the growth-promoting factors of inequality may dominate in high income countries. Lin, Huang, Kim and Yeh (2009) implement an instrumental variables regression approach of Caner and Hansen (2004) and provide empirical support for the findings of Barro (2000). Lin et al. (2009) indicate that there indeed exists a non-linear relationship between inequality and growth. Specifically, greater inequality deteriorates economic growth in low-income countries and enhance growth in high-income countries. The findings reveal that a trade-off between income inequality and growth exists in high-income countries.
2.4 Overview
Many theories exist for assessing the effect of within-country income inequality on economic growth. Income inequality provokes work effort incentives and a greater aggregate savings rate which enhance economic growth. However, inequality may negatively affect the rate of
growth through a decrease of aggregate consumptions, redistributive fiscal policy, political instability, the presence of imperfect credit markets and higher fertility rates. These different theories tend to have offsetting effects and the net effect of within-country income inequality on the growth rate in several countries is therefore ambiguous (Galor, 2000; Barro, 2000).
The ambiguous effect in theoretical research also appears in existing empirical findings. Some researchers have reported a positive relationship between inequality and growth (Forbes, 2000; Li & Zou, 1998). Nevertheless, other empirical analyses indicate an inversed effect of income inequality on growth (Persson & Tabellini 1994; Peroti, 1996; Odedokun & Round, 2004). Whether the effect is positive or negative could depend on country-specific characteristics. Barro (2000) and Lin et al. (2009) already have examined the net effects of income inequality on growth across different countries with diverging per capita GDP’s and found that greater inequality deteriorates economic growth in low-income
countries and enhances growth in high-income countries which is in contrast to the theoretical framework of Galor (2000).
The discussion above shows that the found relationship between income inequality and growth depends on methods of measuring and the selected period of time that is assessed. One big limitation that affects the studies above is the aged data. Most data analysed in all researches is collected before the 21st century and the most recent data, analysed by Lin et al.
(2009), is already one decade old. Over the last decade a lot might have changed due to the ever-increasing globalization, the financial crisis, and even further increases in global income inequality. As a consequence, there may be more recent and relevant conclusions to be drawn about the direction and magnitude of income inequality on economic growth across countries. Moreover, previous research has only distinguished the effect of income inequality on growth for developing and developed countries (Galor, 2000; Barro; 2000, Lin et al., 2009). Since sustainable growth in large middle-income countries is considered to provide positive spillovers to the rest of the world (Worldbank, 2018), an appraisal of the effect of income inequality on their economic growth is meaningful. This paper will complement the previous literature by studying the effects of income inequality on economic growth in countries with different income levels using the most recent available data and different econometric methods.
3. Methodology
3.1 Data description
This research aims to examine the effects of income inequality on economic growth using panel data for a sample of 48 countries in the period 1980-2012. Cross-section time-series estimation makes it possible to control for time-invariant country-specific fixed effects, thereby cancelling out potential omitted variable biases that will affect the coefficients of the estimates. The sample includes countries at vastly different levels of economic development and yearly data in order to estimate the effect in the short-run. Table 3 and table 4 of the Appendix show the selection of countries as well as their descriptive statistics.
In order to obtain a maximum comparability with the existing literature, the Gini coefficient has been chosen as inequality measurer. However, there are several data sources for the Gini coefficient. Solt (2016) has developed the Standardized World Income Inequality Database (SWIID), with the purpose of increasing comparability by standardizing Gini
observations from different databases. The data collected from the Luxembourg Income Study (LIS) is used as reference data to harmonize several data bases into one database which lists country’s income inequality by year. Inequality in disposable income is used since it accounts for redistributive policies implemented by a country which better reflects the real income inequality. Given the availability of the yearly Gini coefficient, a selection of 48 countries between 1980 and 2012 has been created. This time range is used because the SWIID only covers data for a small number of countries before 1980 and after 2012.
The Gross Domestic Product (GDP) is the most widely accepted and used form in order to determine economic performance of a country. The GDP per capita in current US dollars of country i at time t is obtained from a dataset of the World Bank. GDP per capita in current US dollars is used in order to eliminate local currency issues, such as exchange rate fluctuations and inflation.
Additionally, other variables are included in the model to strengthen the results. Firstly, the lagged per capita GDP is used to control for convergence. Secondly, the Penn World Tables 9.0 contains data about the population growth and human capital of a country. The index of human capital is based on the average years of schooling from the Barro and Lee (2013) dataset and assumes a rate of return to education based on Mincer equation estimates (Psacharopoulos, 1994). Finally, dummy variables are included to distinguish the effects of income disparities on economic growth across countries with different levels of incomes. The World Bank currently classifies countries into four income groups: low, lower-middle,
upper-middle and high-income countries. GNI per capita, converted from local currency in US dollars using the World Bank Atlas method, is used to divide economies. These countries might be classified differently in ensuing years. One limitation of this dataset is that the availability is only limited to 1987-2016. Since this research examines the effects of income inequality on growth between 1980 and 2012, the assumption that the classification of countries into an income group before 1987 is equal to the median classification during the five years after 1987 is made.
3.2 The model
The aim of this thesis is to analyse the impact of income inequality on economic growth in countries with different income levels. The model to estimate this effect has been developed on the basis of the existing literature and complements the existing literature by adding dummy variables for low-income (L), low-middle income (LM), upper-middle income (UM) and high-income (H) countries. The equation to be estimated is the following:
𝑌𝑖,𝑡 = 𝛽0+ 𝛽1ln (𝐺𝐷𝑃𝑖,𝑡−1) + 𝛽2𝐻𝐶𝑖,𝑡−1+ 𝛽3𝑃𝑂𝑃𝑖,𝑡+ 𝛽4𝐺𝐼𝑁𝐼𝑖,𝑡−1𝐿 + 𝛽5𝐺𝐼𝑁𝐼𝑖,𝑡−1𝐿𝑀 + 𝛽6𝐺𝐼𝑁𝐼𝑖,𝑡−1𝑈𝑀 + 𝛽7𝐺𝐼𝑁𝐼𝑖,𝑡−1𝐻 + 𝜆𝑡+ 𝜇𝑖,𝑡 (1)
𝑌𝑖,𝑡 denotes the annual per capita GDP growth, 𝐺𝐷𝑃𝑖,𝑡−1 the per capita GDP, 𝐻𝐶𝑖,𝑡−1 the human capital, 𝑃𝑂𝑃𝑖,𝑡 the annual population growth, 𝐺𝐼𝑁𝐼𝑖,𝑡−1 the income inequality, 𝜆𝑡 country-specific fixed effects and 𝜇𝑖,𝑡 the error-term. The interaction term of the Gini coefficient with the dummy variables differentiates the effect of income inequality on economic growth for countries with different levels of incomes. The annual per capita GDP growth rate and the population growth rate are computed according to the following
equations:
𝑌𝑖,𝑡 = ln(𝐺𝐷𝑃𝑖,𝑡) − ln (𝐺𝐷𝑃𝑖,𝑡−1) 𝑃𝑂𝑃𝑖,𝑡 = ln(𝑃𝑂𝑃𝑖,𝑡) − ln(𝑃𝑂𝑃𝑖,𝑡−1)
The variables described above are included in the model in order to control for the steady state growth of a country and convergence. Population growth and human capital determine the steady state growth level of income per capita of a particular country. The higher the rate of population growth and the lower the rate of human capital, the lower the steady state growth of a country (Mankiw, Romer & Weil, 1992). The lagged value of per capita GDP is included because the theory of convergence states that countries with a lower GDP per capita will tend to grow at faster rates than economies with a higher GDP per capita which indicates
that 𝛽1 should be negative. The lagged value of GDP per capita makes the equation above a dynamic panel model.
One problem with the estimated equation described above is that, as explained in the second section, there is a causal relationship which runs in the opposite direction, i.e. the GDP per capita affects the income inequality level in a country, which is illustrated by Kuznets’ (1955) inverted U-curve. First, this relationship is estimated by the following equation:
𝐺𝐼𝑁𝐼𝑖,𝑡 = 𝛼0+ 𝛼1(𝑙𝑛𝐺𝐷𝑃𝑖,𝑡) + 𝛼2(𝑙𝑛𝐺𝐷𝑃𝑖,𝑡)2+ 𝜆
𝑡+ 𝜀𝑖,𝑡 (2)
Before estimating this equation, a Hausman specification test is performed in order to analyse whether random or fixed effects are preferred in the model (Hausman, 1978). This is followed by an OLS estimation of model 2 to explore whether there is a significant relationship
between the economic development level and the level of inequality as proposed by Kuznets (1955). Significance of the independent variables reveals a reverse causality problem. As a result, the explanatory income inequality variables in model 1 are endogenous which means that these variables are correlated with the error-term. Estimating the parameters by using OLS technique would therefore result in a bias of the estimated coefficients. This econometric problem can be solved in several ways.
Firstly, several researchers have used lagged values of the Gini coefficient under the hypothesis of error independence, which means that 𝐺𝐼𝑁𝐼𝑖,𝑡−1 is correlated with 𝜇𝑖,𝑡−1 but not with 𝜇𝑖,𝑡, in order to avoid reverse causation and to justify the chosen OLS method (Persson & Tabellini, 1994; Naguib, 2015). Nonetheless, Bellamare, Masaki and Pepinsky (2015) allege that this solution is an illusion and provide analytical results on the bias resulting from lagged variables in dynamic panels in an OLS regression framework. Secondly, the bias caused by reverse causality could be solved by using exogenous or predetermined instruments. A valid instrumental variable must satisfy two conditions, known as the instrument relevance condition and the instrument exogeneity condition (Stock & Watson, 2015). An instrument is relevant if it is correlated with the explanatory variable and an instrument is exogenous if it is not correlated with the error-term. However, there are merely exogenous variables that are time variant and therefore this model needs to rely on
predetermined instruments, namely the lagged values of the Gini coefficient, by using a system generalized method of moments (GMM) technique which is especially designed for endogenous variables that are correlated with the past (Roodman, 2006). The GMM method is based on the assumption that the instruments are strictly exogenous. A Sargan-test is used to identify if the instruments are indeed exogenous and thus uncorrelated to the error-terms.
Nevertheless, the system GMM technique presents two possible problems that need to be taken into account. Firstly, there is a trade-off between the length of the lags and the sample since observations for which lagged observations are unavailable are scrapped from the estimation (Roodman, 2006). The second problem is that GMM is chiefly invented for
samples with a “small T and large N”. Since the second and further lagged values of variables are used as instruments, the number of instruments increases automatically as the number of years in the panel increases which dissatisfies the exogeneity condition (Grijalva, 2011). In order to solve this, a restricted number of instruments are used for the generalized method of moments approach. Since first differences would make weak instruments, only third, fourth and fifth lagged instruments of the Gini coefficient are used. Besides, the instruments are ‘collapsed’ to limit the number of instruments (Roodman, 2006).
Section 2 has shown that the diversity in measures and methods used causes
divergence between the results and makes it hard to compare them. Even though the criticism that OLS regression techniques, which are using lagged variables for the endogenous
variables, could still result in biased estimated coefficients, this research performs also OLS estimations under the hypothesis of error independence in order to compare these results with the findings of the system GMM method.
In summary, first an OLS estimation is made to test the inverted U-curved relationship between the GDP per capita and the income inequality level in a country. Next, OLS
estimations with random effects and fixed effects are performed to estimate the effects of income inequality on economic growth in countries with different levels of incomes. This is followed by a system GMM technique to get unbiased coefficients of model 1 described above. Finally, the results from these performed estimation methods are used to interpret the coefficients.
4. Results
4.1 Empirical results
In this section, the results of the performed analyses are reported. Figure 2 illustrates the relation between the natural logarithm of GDP per capita and the Gini coefficient for 48 countries between 1980 and 2012. The quadratic fitted line plot shows indeed the inverted U-curve initiated by Kuznets. One can remark that the beginning of the U-curve, e.g. data from the primarily agricultural societies, is missing. Table 5 of the Appendix shows the percentages of the observed income levels in the sample. Low-income countries are underrepresented (13.194%) and high-income countries are overrepresented (44.949%) in the sample, which
possibly explains the missing part of the U-curve. The threshold where GDP per capita becomes negatively related to the Gini coefficient is approximately at a GDP per capita in current US dollars of $2200.
Figure 2: relationship between per capita GDP and the Gini coefficient
The results of the performed regression that corresponds with figure 2 are reported in table 1. The Hausman specification test has been performed and suggests that fixed effects are preferred in model 2. The results of the Hausman specification test can be found in table 5 of the Appendix. The results of table 1 are consistent with Kuznets’ theory. Both GDP per capita and squared GDP per capita are individually and jointly statistically significant from zero at 1%. If the GDP per capita in current US dollars is approximatively equal to 0, then the estimated Gini coefficient is equal to 0.208. As a country develops, the Gini coefficient first increases and falls when the per capita GDP passes $2200.
As described in section 2, inequality affects the investment ratio and thereby affects growth. Kaldor (1955) states that redistribution of income reduces the aggregate savings rate and hence investments. However, higher income inequality could lead to social unrest and a taxation of capital in favour of the capital-poor median voter which both decrease investments (Bertola, 1993; Persson & Tabellini, 1994; Alesina & Perotti, 1996). In other theories,
inequality affects growth indirectly by first influencing the accumulation of human capital (Galor & Zeira, 1993; Morand 1990). The analysis in table 1 shows that the investment rate
does not significantly affect he Gini coefficient. Other regressions in which the investment rate is the dependent variable verifies that greater inequality does not necessarily affect the investment rate. A possible explanation is that the positive effect and negative effect of income inequality on the investment rate may offset each other.
Table 1: Results from the OLS regression
Gini Fixed effects Fixed effects
Log of per capita GDP 0.0310*** (0.0059)
0.0388*** (0.0065) Squared log of per capita
GDP -0.0012*** (0.0003) -0.0013*** (0.0004) Investment rate 0.0005 (0.0043) Human capital -0.0180*** (0.0047) Number of observations 1582 1582 Number of groups 48 48
*,**,*** denote that the null-hypothesis is rejected at 10%, 5%, and 1 % respectively. Standard errors are represented in the parentheses.
Table 2 presents estimates of the OLS regression method with random effects, the OLS regression method with fixed effects and the system GMM where lagged per capita GDP, lagged human capital, population growth and the lagged Gini coefficients for several levels of incomes are on the right-hand side of equation.
The first column suggests that income inequality has a negative significant effect on economic growth in low-income countries. An increase of the Gini coefficient of 0.01 is followed by a decrease of 0.2133% in economic growth. As a country develops and is marked as low-middle income country, then the negative effect is not significant anymore. If a
country develops further and is marked as upper-middle income country or high-income country, there seems to be a positive direction in the effect of income inequality on growth. However, since these effects are not significant, income inequality appears to have less of an influence for upper-middle income and high-income countries. The directions of the estimates using both random effects and fixed effects models suggest a non-linear effect whereby higher inequality tends to delay growth in developing countries and encourages growth in
developed countries, which is consistent with the findings of Barro (2000) and Lin et al. (2009). However, the p-values of the test are very high, well above the usual threshold of 5%, which indicates that there is not enough evidence that income inequality significantly affect economic growth.
The third column shows the estimated coefficients when the estimation is performed with GMM. It appears that the per capita GDP, human capital and population growth do not have a significant effect on economic growth, while income inequality in countries with different levels of incomes does. Specifically, income inequality has a negative effect on economic growth which becomes larger as a country develops. If the Gini coefficient increases by 0.01, the short-run decrease of economic growth in a country ranges between 1.0441% and 1.7549%.
Table 2: results from random effects OLS, fixed effects OLS and system GMM Growth per capita GDP Random effects
(1)
Fixed effects (2)
System-GMM (3)
Log of per capita GDP -0.04010*** (0.0070) -0.1182*** (0.0121) -0.0126 (0.0481) Human capital 0.0471*** (0.0103) 0.3045*** (0.0306) 0.0302 (0.1524) Population growth -0.9366** (0.4758) 1.4071* (0.8148) -8.7217 (8.7324) Gini low-income countries -0.2113*** (0.0748) -0.1946 (0.1866) -1.0441* (0.5823) Gini low-middle income
countries -0.0599 (0.0558) -0.0038 (0.1682) -1.0273* (0.5565) Gini upper-middle income
countries 0.0083 (0.0546) 0.0718 (0.1713) -1.0672* (0.5820) Gini high income
countries 0.0609 (0.0766) 0.0105 (0.1865) -1.7549** (0.8709) Number of observations 1533 1533 1533 Number of groups 𝑅2 48 0.0312 48 0.0155 48 N/A
*,**,*** denote that the null-hypothesis is rejected at 10%, 5%, and 1 % respectively. Standard errors are represented in the parentheses.
Since the estimates differ significantly based on the utilized technique (e.g. OLS or GMM), it is important to test the validity of the assumptions underlying each method. First, a Hausman specification test indicates that fixed effects are preferred over random effects in the OLS models because there are structural specific factors in each country that may affect the
relationship between income inequality and growth. The results of this test are shown in table 7 of the Appendix. The estimated coefficients in the random effects model are therefore considered to be biased due to omitted variables. However, the estimated coefficients in the fixed effects model could still be biased due to endogeneity and the weak solution of using lagged variables at OLS regression methods (Bellamare, Masaki & Pepinsky, 2015; Grijalve, 2011). As a consequence, a generalized method of moments was performed in order to solve for possible endogeneity problems. A valid predetermined variable should satisfy the
exogeneity condition which can be examined by applying the Sargan-test of overidentifying restrictions (Stock & Watson, 2015). The 𝜒2 and the p-values are shown in table 8 of the
Appendix. The results identify that the instruments are indeed exogenous and thus estimation with system GMM yields consistent estimates. In conclusion, the GMM corrects for the upward bias and income inequality has a negative effect on economic growth which becomes larger as a country develops. This indicates that the negative effects of income inequality possibly offset the positive effects of income inequality on economic growth.
4.2 Interpretation of results 4.2.1 Control variables
The estimated coefficients of the per capita GDP are in all methods consistently negative. This supports the theory of conditional convergence that states that countries with a lower GDP per capita will tend to grow at faster rates than countries with a higher GDP per capita. The effect of human capital on growth is positive in all econometric techniques utilized, which is consistent with the Solow growth model developed by Mankiw, Romer and Weil (1992). Additionally, population growth should have a negative effect on economic growth according to the Solow growth model which appears to be the case in an OLS model with random effects and a system GMM model. Briefly, the directions of the control variables in an OLS model with random effects and a system GMM model are consistent with the
academic theory. However, it is remarkable that the system GMM model provides statistically insignificant estimators of these coefficients which should be theoretically important. A reasonable explanation for this phenomenon is that the GMM technique results in a lower
efficiency of the estimators since the desired quality of the used instruments is too weak. This can be improved by increasing the number of instruments. On the other hand, the increased number of instruments will decline the power of the Sargan-test of overidentifying
restrictions. There is therefore a trade-off between a higher efficiency of the estimators and the power of the Sargan-test. This shortcoming is exhibited in the insignificance of the log of per capita GDP, human capital and population growth.
4.2.2 Explanatory variables
Although this research finds a negative linear relationship rather than a non-linear relationship between countries with different levels of incomes, the explanation of this linear relationship does depend on a country’s initial level of income.
The significant negative effect for low income countries may indicate that inequality tends to delay growth in developing countries, which is consistent with the findings of Barro (2000). One plausible explanation for this finding is that, even though the increasing
globalization, a few low-income countries have gained access to international markets and a bulk of flows still go to a few large middle-income countries. The most low-income countries therefore still depend on their own financial sector in order to meet their financial needs and individuals tend to be more exposed to credit-market constraints (Hostland, 2009; Barro, 2000). As a consequence, greater income inequality means less investment in human capital and higher fertility rates in low-income countries, which both reduces the economic growth (Galor & Zeira, 1993; Morand, 1999).
Although many middle-income countries have gained access to international markets, they still face political instability due to corruption, civil wars as well as political disorder. For a given level of expected changes in the government or laws in a country, one may argue that particular phenomena do not have any direct impact on political uncertainty and thus investments in a country (Alesina & Perotti, 1996). However, if countries are already exposed to political instability and many changes in the government as well as constitutional laws, which is assumed to be the case in middle-income countries, greater income inequality leads to social unrest which is followed by a reduction of investments (Persson & Tabellini, 1994).
The negative effect of greater income inequality on growth in high-income countries can be explained by the theory developed by Persson and Tabellini (1994) and are also consistent with their empirical findings. High-income countries tend to be more democratic than other countries. Democratic countries succumb under the insistence of income
redistribution and increases the taxes on capital which reduces the incentives to invest and therefore economic growth.
5. Conclusion and discussion
The aim of this thesis is to clarify to which extent income inequality is related to economic growth in countries with different levels of incomes. Before estimating this relationship, the Kuznets’ curve is assessed which denotes that as a country develops from a primarily agricultural society to an industrialized economy, inequality in the distribution of incomes first increases and later decreases. The results in section 4 coincide with the Kuznets’ curve and appraise that the threshold where GDP per capita becomes negatively related to the Gini coefficient is approximately equal to a GDP per capita in current US dollars of 2200.
Additionally, section 4 shows that there seems to be an insignificant non-linear effect if OLS regression techniques are performed. However, the results obtained in the OLS estimates have to be considered with caution due to endogeneity problems. These problems have been solved by using a system-GMM estimator with a restricted number of instruments. The conclusion can be drawn that, as previously stated, income inequality has a negative effect on economic growth which becomes larger as a country develops. The estimated findings support previous evidence of a linear negative relationship between income inequality and economic growth (Persson & Tabellini, 1994; Perotti, 1996; Odedokun & Round, 2004). It is remarkable that there is no evidence of the existence of a non-linear effect whereby income inequality positively affects economic growth in developing countries and negatively affects economic growth in developed countries according to Galor (2000) or vice versa according to Barro (2000) and Lin et al. (2009). A theoretical explanation for the divergence in the empirical findings is that Galor (2000) and Barro (2000) explain the positive effect of income inequality on growth, in developing and developed countries respectively, by using Keynes’ theory about the marginal propensity to save: greater income inequality increases the country’s marginal propensity to save and thus investments which is beneficial for the economic
growth. However, a plausible argument to mitigate this effect is that higher income inequality results in a lower marginal propensity to consume. The decline in consumption could be detrimental for economic growth (Keynes, 1937). It would be interesting to investigate in detail which effect supersedes the other in countries with different levels of incomes. Another possible reason for the contradiction of the findings is due to differences in measures of the variables, methods as well as the selected period of time that is assessed. For instance, Lin et
al. (2009) and Barro (2000) have averaged their data over periods of three to seven years and ten years respectively, while this study has estimated the effect of income inequality on economic growth in the short-run, namely in one year. The short-run response of economic growth is not as clearly specified as the medium- and long-run response (Barro, 2000) and it is possible that this relationship can be characterized as non-linear in the medium- and long-run. A recommendation for further research is to investigate the transition of the effect of income inequality on economic growth over different time horizons in more detail.
Some limitations of this study have been mentioned before. Due to endogeneity, omitted variables and a weak quality of the predetermined instruments, the estimated effect of the included variables could be incorrect. However, obtaining strong instruments will always be a hard challenge in dynamic panel models. Another issue that may arise in this research is that there is a selection bias towards high-income countries due to data availability of the Gini coefficient. As a consequence, low-income countries are underrepresented in the sample that has been studied. Further research could focus on a proportionately distributed sample,
different datasets and other instruments in order to estimate the effect of income inequality on economic growth across countries with different levels of incomes.
6. References
Acemoglu, D., & Robinson, J. A. (2002). The political economy of the Kuznets curve. Review of development economics, 6(2), 183-203.
Alesina, A., & Perotti, R. (1996). Income distribution, political instability, and investment. European economic review, 40(6), 1203-1228.
Alvaredo, F., & Chancel, L., Piketty, T., Saez, E., & Zucman, G. (2018). World inequality report 2018.
Barro, R. J. (2000). Inequality and Growth in a Panel of Countries. Journal of economic growth, 5(1), 5-32.
Barro, R. &, Lee, J. (2013). A new data set of educational attainment in the world, 1950-2010. Journal of Development Economics, 104, 184-198.
Bellemare, M. F., Masaki, T., & Pepinsky, T. B. (2017). Lagged explanatory variables and the estimation of causal effect. The Journal of Politics, 79(3), 949-963.
Bertola, G. (1993). Factor shares and savings in endogenous growth. The American Economic Review, 83(5), 1184-1198.
Bhattacharya, J. (1998). Credit market imperfections, income distribution, and capital accumulation. Economic Theory, 11(1), 171-200.
Biswas, S., Chakraborty, I., & Hai, R. (2017). Income inequality, tax policy, and economic growth. The Economic Journal, 127(601), 688-727.
Caner, M., & Hansen, B. E. (2004). Instrumental variable estimation of a threshold model. Econometric Theory, 20(5), 813-843.
Charles-Coll, J. A. (2011). Understanding income inequality: concept, causes and
measurement. International Journal of Economics and Management Sciences, 1(3), 17-28.
Cingano, F. (2014). Trends in income inequality and its impact on economic growth. Cobham, A., & Sumner, A. (2013). Is it all about the tails? The Palma measure of income
inequality.
Forbes, K. J. (2000). A Reassessment of the Relationship between Inequality and Growth. American economic review, 90(4), 869-887.
Galor, O. (2000). Income distribution and the process of development. European Economic Review, 44(4-6), 706-712.
Galor, O., & Moav, O. (2004). From physical to human capital accumulation: Inequality and the process of development. The Review of Economic Studies, 71(4), 1001-1026.
Galor, O., & Zeira, J. (1993). Income distribution and macroeconomics. The review of economic studies, 60(1), 35-52.
Grijalva, D. F. (2011). Inequality and Economic Growth: Bridging the Short-run and the Long-run.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica: Journal of the econometric society, 1251-1271.
Hostland, D. (2009). Low-income countries' access to private debt markets.
Kaldor, N. (1955). Alternative theories of distribution. The review of economic studies, 23(2), 83-100.
Keynes, J. M. (1937). The general theory of employment. The quarterly journal of
economics, 51(2), 209-223.
Knowles, S. (2005). Inequality and economic growth: The empirical relationship reconsidered in the light of comparable data. The Journal of Development Studies, 41(1), 135-159. Kuznets, S. (1955). Economic growth and income inequality. The American economic review,
Lazear, E. P., & Rosen, S. (1981). Rank-order tournaments as optimum labor contracts. Journal of political Economy, 89(5), 841-864.
Li, H., & Zou, H. F. (1998). Income inequality is not harmful for growth: theory and evidence. Review of development economics, 2(3), 318-334.
Lin, S. C., Huang, H. C., Kim, D. H., & Yeh, C. C. (2009). Nonlinearity between inequality and growth. Studies in Nonlinear Dynamics & Econometrics, 13(2).
Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A contribution to the empirics of economic growth. The quarterly journal of economics, 107(2), 407-437.
Mirzaei, S., Borzadaran, G. R. M., Amini, M., & Jabbari, H. (2017). A comparative study of the Gini coefficient estimators based on the regression approach. Communications for Statistical Applications and Methods, 24(4), 339-351.
Morand, O. F. (1999). Endogenous fertility, income distribution, and growth. Journal of Economic Growth, 4(3), 331-349.
Naguib, C. (2015). The relationship between inequality and GDP growth: An empirical approach. LIS Working Paper Series.
Odedokun, M. O., & Round, J. I. (2004). Determinants of income inequality and its effects on economic growth: evidence from African countries. African Development Review, 16(2), 287-327.
Palma, J. G. (2011). Homogeneous Middles vs. Heterogeneous Tails, and the End of the ‘Inverted‐U’: It's All About the Share of the Rich. Development and Change, 42(1), 87-153.
Panizza, U. (2002). Income inequality and economic growth: evidence from American data. Journal of Economic Growth, 7(1), 25-41.
Perotti, R. (1996). Growth, income distribution, and democracy: What the data say. Journal of Economic growth, 1(2), 149-187.
Persson, T., & Tabellini, G. (1994). Is inequality harmful for growth? The American Economic Review, 600-621.
Piketty, T. (2014). Capital in the twenty-first century. Cambridge, MA: Harvard University Press.
Piketty, T. (2015). About capital in the twenty-first century. American Economic Review, 105(5), 48-53.
Pimentel, D. A. V., Aymar, I. M., & Lawson, M. (2018). Reward work, not wealth. Retrieved from https://d1tn3vj7xz9fdh.cloudfront.net/s3fs-public/file_attachments/bp-reward-work-not-wealth-220118-en.pdf
Psacharopoulos, G. (1994). Returns to investment in education: A global update. World development, 22(9), 1325-1343.
Roodman, D. (2006). How to do xtabond2: An introduction to difference and system GMM in Stata.
Solt, F. (2016). The standardized world income inequality database. Social Science Quarterly, 97(5), 1267-1281.
Stock, J. H., &, Watson, M.W. (2015). Introduction to Econometrics (3rd ed.). Boston, MA:
Pearson Education
Sukiassyan, G. (2007). Inequality and growth: What does the transition economy data say? Journal of comparative economics, 35(1), 35-56.
Worldbank (2018). The World Bank in middle income countries. Retrieved from http://www.worldbank.org/en/country/mic/overview#2
7. Appendix
7.1 Descriptive statistics
Table 3: descriptive statistics of the variables
Variable Obs Mean Std.Dev. Min Max
Per capita GDP 1582 13556.98 15855.06 120.63 102000
Gini coefficient 1584 .384 .094 .205 .587
Human capital 1584 2.538 .641 1.211 3.719
Population growth 1536 .013 .009 -.004 .063
Table 4: list of all countries including their mean of the Gini coefficient and their median of the income group between 1980-2012.
Countries Average Gini coefficient Classification of income level
Argentina 0.4255 UM Australia 0.3104 H Belgium 0.2482 H Brazil 0.5064 UM Canada 0.2981 H Chile 0.4964 UM China 0.4228 L
Colombia 0.5142 LM Costa Rica 0.4264 LM Denmark 0.2373 H Egypt 0.4733 LM Fiji 0.4541 LM Finland 0.2315 H France 0.2959 H Germany 0.2690 H Greece 0.3406 UM Guatemala 0.4668 LM Hong Kong 0.3952 H India 0.4470 L Indonesia 0.3926 L Iran 0.4392 LM Ireland 0.3166 H Israel 0.3351 H Italy 0.3258 H Japan 0.2825 H Korea 0.2884 UM Madagascar 0.3992 L Malawi 0.4653 L Malaysia 0.4557 UM Mexico 0.4657 UM Netherlands 0.2556 H Norway 0.2413 H Pakistan 0.3575 L Panama 0.4898 LM Peru 0.5256 LM Philippines 0.4712 LM Portugal 0.3440 H Singapore 0.3880 H South Africa 0.5675 UM Spain 0.3240 H
Sri Lanka 0.4824 L Sweden 0.2304 H Switzerland 0.3003 H Thailand 0.4619 LM United Kingdom 0.3279 H United States 0.3538 H Venezuela 0.4068 UM Zambia 0.4858 L
Table 5: percentages of the observed income levels in the sample Income levels Percentage in the sample
Low income 13.194%
Low-middle income 23.485% Upper-middle income 18.371%
High income 44.949%
7.2 Hausman specification tests
The Hausman specification test evaluates the consistency of estimators. De null hypothesis of the Hausman specification test suggests that there are no systematic differences in the
coefficients that are estimated with random effects and fixed effects (Hausman, 1978). The Hausman specification test rejects the null hypothesis and only an estimation method with fixed effects is consistent.
Table 6: Hausman specification test (table 1) Coefficient
𝜒2(2) 79.92
Prob>𝜒2 0.000***
Table 7: Hausman specification test (table 2) Coefficient
𝜒2(7) 87.74
7.3 Sargan-test of overidentifying restrictions
A Sargan-test is used to identify if the instruments are indeed exogenous and thus
uncorrelated to the error-terms. A rejection of the null hypothesis indicates that the model is weakened by many instruments and that the estimated coefficients are inconsistent. There is not enough evidence to reject the null hypothesis. Therefore, we can conclude that the estimates are consistent coefficients.
Table 8: Sargan-test of overidentifying restrictions Coefficient
𝜒2(9) 8.73
