The effect of income inequality on economic growth at differing levels of economic development

(1)

1

The effect of income inequality on economic growth at differing

levels of economic development

Bachelor’s Thesis

June 2016

Student Name:

Thomas David Baxendale

Student Number: 10621067

Specialisation:

Economics & Finance

Supervisor:

Dr. Dirk Damsma

Faculty:

Economics and Business

(2)

2

Statement of Originality

This document is written by Thomas David Baxendale who declares to take full responsibility

for the contents of this document.

I declare that the text and the work presented in this document is 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.

(3)

3

Abstract

This paper attempted to find a relation between income inequality and economic growth. This includes both a review of previous literature, both qualitative and quantitative, as well as empirical research on primary data. Using data for twenty countries, ranging from 1964-2012 the hypothesis that income inequality has an effect on economic growth was tested. The hypothesis was extended to include whether inequality has a different effect on growth at differing levels of economic development. The results of the regressions showed no significant linear relation when studying all countries, a positive linear relationship when studying less developed countries and no significant linear relationship when studying developed countries. Furthermore, when subject to a second degree polynomial regression the results showed that income inequality had a quadratic relationship to economic growth. This relationship was significant in the sample of all countries as well as the sample of less developed countries, but not significant in the sample of developed countries. The results suggest that not only does this quadratic relationship exist, but inequality’s effect on growth weakens as a country develops. Further research is necessary in order to affirm these conclusions.

(4)

4

Beginning in 2011, the Arab Spring caught the attention of the world, as the populations of several Arab nations organised protests and rebellions against their respective governments. There were several reasons for this movement to take place, an important one being rising inequality (Ncube & Anyanwu, 2012). People do not like to believe that they are presented with lesser opportunities than others in their own country simply due to a factor that is out of their control; such as place of birth or familial ties. Similar stories are present in Western nations. The riots in England, also in 2011, were partially attributed to perceived inequality in society (Cooper, 2012). The Occupy Wall Street movement’s slogan, “We are the 99%”, refers to the concentration of wealth and income in the top 1% of earners (Dittrich & Cheon, 2012). The presidential candidate Bernie Sanders based his campaign on reducing economic inequality in the United States.

Clearly, inequality is a heated topic in society at present. Inequality is usually discussed in a socio-political light, with the focal points being what is fair and equitable for a population. Inequality is one of the key topics that separates the left and right wings of politics.Left wingers generally lean in favour of egalitarianism, whereas right wingers are more predisposed to some form of social hierarchy. However, inequality is less often discussed from a purely economic perspective.

This paper will feature a presentation of current theory on the topic of inequality in relation to economics. This will include the derivation of income inequality measures, as well as economic reasoning as to how inequality can influence economic growth in Chapter 2. In Chapter 3, the method of gathering and analysing data will be explained. The results of the study will be presented in Chapter 4. A conclusion will then be drawn on whether or not income inequality’s has a relationship with economic growth, and if it does, the nature of the relationship will be explained. Finally, limitations of the study and potential further research will be discussed.

2. Current literature and theory surrounding this topic

This paper attempts to find a relation between income inequality and economic growth, in particular the former’s influence on the latter. While economic growth is a fairly simple variable to understand, there are several interpretations of how income inequality should be measured. Therefore, in this paper it is first essential to present and explain various measures of income inequality, in order to make a justified decision as to which measure to use.

To fully understand the debate currently surrounding inequality’s effect on growth, the reverse relationship must first be reviewed. Consequently, in this Literature Review Kuznets’ longstanding inverted U-curve will be introduced and clarified. Following this will be a review of current theories relevant to this paper’s research question. These theories will encompass both the positive and negative effect of inequality on growth.

Subsequently, literature empirically analysing the same relationship as this paper will be presented. This includes panel studies of many countries, as well as other analyses focussing on different factors that potentially have an effect on the proposed relationship. This relationship appears to be a popular topic in economics, with sufficient literature available to be assessed. This literature includes scientific articles as well as books by famous economists.

(6)

6

2.1. Measures of income inequality

There are several measures of income inequality in a population. One key feature of these measures that the literature surrounding inequality generally agrees upon is that each measure must contain the property of the Pigou-Dalton principle of transfers. This is a simple principle introduced by H. Dalton and developed further by A.C. Pigou, stating that any income inequality measure should at least decrease due to a transfer of income from the richer to the poorer (Dalton, 1920; Pigou, 1912). This principle is an important start in determining whether a measure is effective in measuring income inequality and its changes in a population.

Income inequality can be measured using the Gini coefficient. The Gini coefficient of a country is calculated using the Lorenz curve. This curve shows the cumulative distribution of either income or wealth within a country (Bellù & Liberati, 2006). The cumulative proportion of the population is on the x-axis, and the cumulative proportion of income on the y-axis. Using this curve it is possible to make a statement such as “the bottom 60% of the population earn 40% of the total income”. In a society of perfect income equality the Lorenz curve would be the 45ᵒ line y=x. The Gini coefficient is a function of the areas between this line of equality and the Lorenz curve (A), and below the Lorenz curve (B). It is given by the ratio A/(A+B) (Bellù & Liberati, 2006). The coefficient ranges between 0 (complete equality) and 1 (complete inequality). Figure 1 displays the two curves as well as areas ‘A’ and ‘B’.

An alternative measure of income inequality is the Hoover index. This is also known as the Robin Hood index, as it is an analysis of taking from the rich and giving to the poor. To determine the Hoover index, the mean income level must first be found. This is the largest vertical gap between the Lorenz curve and the line of perfect equality (Charles-Coll, 2011). The Hoover index is calculated using the reported income figures of the line of perfect equality and Lorenz curve at this widest gap between them. As an example, if the Lorenz curve reports $50,000 and the line of perfect equality

Figure 1: The Lorenz Curve and Line of Perfect Equality

(7)

7

$60,000, the Hoover index is then (60000-50000)/60000=0.1667. In this example, 16.67% of total income needs to be redistributed from above the mean to below in order to achieve perfect equality. This index also ranges between 0 (complete equality) and 1 (complete inequality).

The Palma ratio measures how much richer the top 10% of a population are compared to the bottom 40% (Cobham & Sumner, 2013). José Gabriel Palma observed that the changes in income distribution of a population over time can be attributed to the tails of the distribution, as the centre of the distribution stays relatively stable (2014). The Palma ratio has no upper bound. To give some perspective, in 2012 Norway’s Palma ratio was 0.8 whereas the US has a Palma ratio of 1.75. Both the Hoover index and the Palma ratio are easier to understand at face value than the Gini coefficient while still using the Lorenz curve. All three measures satisfy the Pigou-Dalton principle of transfers; as income is transferred from the haves to the have-nots, each measure indicates a more equal society has been achieved.

2.2. Economic theory

2.2.1. Kuznets and the inverted-U hypothesis: growth’s effect on inequality

Kuznets (1955), in his seminal article in The American economic review, laid the foundation for future discussion about income inequality within the field of economics. At the time of writing the article, developed countries such as Germany, the UK and the US were experiencing a decrease in inequality, and Kuznets sought to explain this change. However, as he was an early scholar in this field, there was very little in terms of data or dominant theory. The paper frequently alludes to this lack of data, with Kuznets stating that the paper is “perhaps 5 per cent empirical information and 95 per cent speculation”. Despite this, Kuznets makes much headway into the field. The prevailing theory from this paper discusses a country’s shift from agriculture to industry as it experiences economic growth. This shift in sectors is accompanied by internal migration from rural areas to cities. It is this transitory period of a country that increases income inequality, which Kuznets demonstrates using a rudimentary numerical example.

He further speculates about why, then, underdeveloped countries (in his paper India, Ceylon and Puerto Rico) have wider income inequality than developed countries. He states that a progressive taxation and benefits system does much to reduce the income inequality present in a nation, and this system is more likely to be present in developed countries. Although not explicitly mentioned, this is a basic formulation of the Kuznets curve; income inequality first increasing, then decreasing as an economy grows.

2.2.2. Inequality’s effect on growth

Kuznets focussed on economic growth’s effect on inequality, which is not the topic of discussion in this paper. He did mention, however, inequality’s effect on savings, which in turn affected growth. In a society, the wealthy have a higher propensity to save; meaning a greater share of their income is devoted to saving as opposed to consumption (1955, p.7). Therefore, wider inequality acts as a stimulant of economic growth via savings (p. 9). Again, this was speculation and he could not test this theory using real data.

Other scholars have since discussed this topic further. There are two broad standpoints one can have: that inequality has a positive effect on growth, or a negative effect on growth. There are proponents for each standpoint.

(8)

8

a) Positive effect

Kuznets’ view, that inequality created savings, which in turn created growth, is the classical approach to this issue. It is a viewpoint shared by Kaldor (1957) and Bourguignon (1981). Kaldor incorporates this hypothesis into his all-encompassing paper A model of economic growth. Bourguignon, using a wealth distribution model created by Stiglitz (1969), shows that an un-egalitarian equilibrium can be Pareto superior, given a convex savings function and strictly positive wealth (p. 1475).

Bhattacharya (1998) presented a model whereby an economy’s income distribution matters for capital formation, and in this model assessed the effect of a policy with the intention to redistribute income to a more equal setting. Within this model, some capital investment must be externally financed, which is subject to a ‘costly state verification’ (CSV). A certain percentage of the population are capitalists, who have access to high return, but risky, production technologies. Capitalists are able to leave their wealth to their children. These children therefore are able to partially internally finance investment projects via bequests. However, these investments also demand external, costly finance. This finance is delivered by the other fragment of the population, the labourers. In this model, it is shown that the greater the capability of capitalists to bequeath their wealth, and therefore internally finance projects, the less CSV is present. These bequests therefore aid credit market efficiency and capital accumulation. At the same time, it is a mechanism to entrench inequality. A tax on the capitalists’ bequests that is given to the labourers (from the rich to the poor), in this model, ‘’necessarily reduces the steady state capital stock’’ (p. 195). Furthermore, depending on the magnitude of this effect, it may even reduce the total income of the workers. Bhattacharya therefore concludes that redistributive policies can be the detrimental to an economy, and although inequality is an entrenched quality in this model, the same strategy which creates the inequality creates a higher income for both parties.

An alternative opinion on why income inequality is a good thing is held by Mirrlees (1971). Efforts to create a more equal distribution of income have an effect on workers’ motivations. In the extreme, a situation of complete equality, everybody would receive the same wage. As this wage is a constant, not linked to output, there is no incentive to work hard. Mirrlees argues that wage needs to be dependent on final output in order to incentivise workers and maximise aggregate production. Inequality is simply a natural product of this incentivising strategy.

b) Negative effect

Alesina & Perotti (1996) researched income inequality’s effect on socio-political instability. They argued that income inequality was a component of social unrest, which had a negative effect on the political and economic environment of a country. This lack of faith reduces investment, which in turn reduces economic growth. Therefore, social unrest is a channel via which income inequality reduces economic growth.

Persson & Tabellini (1994) also suggest that inequality is harmful for growth, specifically in democracies. In a democratic society which puts importance on distribution of income, the resultant policies tax investment and other growth-promoting activities in an attempt to reallocate income. Thus, it is not the inequality itself but the policies to redistribute income that have an effect on growth. Persson & Tabellini created a theoretical model reflecting this position.

In his 2012 book The Price of Inequality, the economist Joseph E. Stiglitz offered multiple economic arguments about inequality and its negative effect on the economy, specifically in the United States. They are summarised below.

(9)

9

He first argues that the wealthy have a lesser need for public goods than the poor. Features such as education, health, and personal security are public goods that the rich do not need from their government; they can purchase these things themselves. As a country becomes more unequal, the rich are more reluctant to spend money on these public goods, and so public investment decreases. This has a detrimental effect on the economy as poorer individuals’ health, education, mobility and other features are reduced, affecting the productivity of the populace.

This increase in inequality not only reduces investment in education, it reduces economic mobility. Economic mobility is the ability for poor people to improve their incomes. Expensive tuition in the United States, combined with a student loan system that does not help but hinder the poor, can mean the poor have the choice of tens of thousands of dollars of debt or no further. As the poor become increasingly poor, they have less time to invest in their children, both time-wise and financially. This lack of time not only reduces mobility, and hence entrenches inequality, but also has an effect on educational attainment. This is because the home life of a child has a significant effect on their grades. Stiglitz posits that this has an effect on the long-term productivity of a nation.

Stiglitz also discusses rent seeking at length. Rent seeking is an effort to increase one’s own wealth without creating new wealth, and is especially present in many industries in the US. It is resources spent on capturing wealth that already exists, as opposed to producing new wealth. An example of rent seeking would be lobbying for maintaining monopoly power, or preferential tax treatment. Despite rent seekers creating profit for their own corporation, the overall benefit to the economy may not even be positive. Rent seeking industries include the financial sector and pharmaceutical corporations. Financial corporations attract large amounts of young talent, who create ‘financial innovations’ to circumvent current regulations. These can come at a detriment to the economy, and have little to no societal benefit. Pharmaceutical corporations spend more money on marketing than research. They spend money creating the same drug as their competitors to take a share of their profits. If the money spent on marketing and copy-cat research was spent on real research the economy would benefit much more. This is an inefficient allocation of resources, which has a negative effect on growth. A central theme of Stiglitz’ book is that this rent seeking increases income inequality in a society. Therefore, an increase in inequality is indicative of the presence of rent seeking and a misallocation of resources.

Poverty affects productivity. This is true in both relative and absolute senses. Stiglitz presented an experiment in which workers who were doing the same job were first paid the same. After wages for some were increased and for others decreased, one might suggest that the increase in productivity of the richer worker would offset the decrease in productivity of the poorer workers, but this turned out not to be true; total productivity decreased (Cohn et al., 2009). Workers who feel like they are being treated unfairly will be much less productive. This can be extended to the aggregate workforce: if the bottom portion of society feels as though it’s being unfairly compensated, the productivity of an economy will drop.

The above paragraph presents relative poverty – poverty relative to others in the same society. Absolute poverty is an income level with which an individual struggles to buy basic goods, such as food and shelter. This also has an effect on productivity. Malnourished workers are less productive (Leibenstein, 1960). Poor people devote more of their cognitive resources to worrying about their situation. Anxiety over children, retirement, mortgages and other worries can limit the amount of thinking a person can devote to improving their own situation, and has even been shown to make individuals make irrational decisions when there is a better option available (Mullainathan & Shafir, 2013).

(10)

10

Aside from a drop in 2009 GDP per capita is consistently rising in the US. This benefit is not realised by the poor, however, as income inequality has also been increasing. This rise in inequality is increasing levels of both absolute and relative poverty, and causing the damaging effects explained above.

c) Unification of effects

Galor (2000) attempted to unify these conflicting views by discussing the intertemporal aspects of the models. He stressed the importance of the development of a country and its effect on income inequality.

He posited that inequality has two contradictory effects on development. The positive effect of increased physical capital accumulation is offset by the negative effect of human capital accumulation. A developing country has limited access to physical capital, and as such the return on physical capital is much higher than that on human capital. Therefore, a developing country’s development is driven primarily by capital accumulation. Galor also uses the increasing marginal propensity to save in wealthy people as the reasoning for why inequality aids capital accumulation.

However, as a country develops, the physical capital accumulated raises the rate of return on human capital. There is therefore a new interest in human capital accumulation. As an individual’s human capital accumulation is subject to diminishing marginal returns, the total return on investment in human capital is maximised if marginal returns are equal across all individuals. In this situation, increased equality has a positive effect on economic growth. By definition, a developed country has a higher wage rate than that of a developing country. The differences between individuals’ marginal propensities to save reduces as income increases, and so the negative effect of equality on savings accumulation is diminished. So as a country develops, the positive effect of inequality on savings is reduced, and the negative effect of reduced investment in human capital dominates. Therefore, a move towards equality stimulates growth.

2.3. Empirical testing

An investigation into the Kuznets curve was conducted by Thornton (2001). He tests the hypothesis using a regression in the following manner:

𝐼𝑁𝐸𝑄𝑖𝑡 = 𝛼0+ 𝛽1𝑙𝑛𝑌𝑖𝑡+ Ω1(𝑙𝑛𝑌)𝑖𝑡2 + 𝜀𝑖𝑡

INEQ is a measure of inequality; in this case the Gini coefficient. Y is real GDP per capita. The Kuznets curves implies β>1, Ω<1. Using panel data of 96 countries, Thornton found both coefficients to be significant at the 95% significance level. The range of data is from the post-war period to 1995. This large number of observations makes a powerful statistical test. However, it is possible conditions have changed over the past two decades, with factors such as the financial crisis upsetting the relationship between economic growth and unemployment. It is also relevant to note that Thornton regresses growth on inequality in the same period, implying no delay between the factors’ movements.

Barro (2000) asked whether there was a reverse causality in this relationship between economic growth and income inequality. That is, if income inequality has an influence on economic growth. In a regression of the Gini coefficient on a country’s growth, roughly 100 countries were included, observed from 1960 to 1995. The same limitation that applied to Thornton’s research applies to this study; the latest observations are from over two decades ago and may not fully apply to present day. The regressions results showed ‘little overall relation between income inequality and rates of growth’. However, at different levels of per capita GDP, inequality was indicated to have a

(11)

11

different effect. Below $2000 (in 1985 terms), wider income inequality tended to slow growth, whereas it had the opposite effect above that level. This is the opposite of the prediction of Galor (2000). Furthermore, the Kuznets curve – showing how income inequality is first increasing, then decreasing as an economy grows – was shown to be present in the sample studied.

Voitchovsky (2005) hypothesised that the influence of income inequality on economic growth differs at separate areas of the income distribution. She studied only democratic, relatively wealthy countries, and had a limited data sample. Her sample comprised of income inequality metrics in these countries, measured every five years, with some data needing to be estimated based on adjacent data, or taken from the previous or subsequent year. In addition to this, data was taken from differing types of household income surveys, potentially causing discontinuity in the data. These limitations may have an effect on the validity of the results. Her results implied that progressive policies, such as a progressive tax or benefits system, “are likely to facilitate growth through their impact on the bottom of the distribution, and to inhibit growth through their impact on the top of the distribution”. So a tax on the rich given to the poor as welfare would both negatively affect the economy via the tax and positively affect the economy via the welfare. This leads to an ambiguous aggregate effect depending on the magnitude of each individual effect. The extent of these effects varied with factors such as the countries chosen and estimation technique.

While Voitchovsky asked whether income inequality’s influence differed at different areas of the income distribution, Lin et al. (2014) asked if it differed with different levels of economic development. They used a large sample of the 48 contiguous US states, from 1945-2004, and found ‘overwhelming evidence’ supporting their hypothesis. Their hypothesis was based on Hansen’s (1999) panel threshold regression and tested the existence of endogenous income thresholds in the relationship between inequality and growth. Their paper was a continuation of the studies Barro (2000) and Lin et al. (2009), who found evidence for this hypothesis using country-level data. The reasoning for this hypothesis was based on previous theoretical works, some of which are already discussed in this paper, such as Galor (2000). Lin et al. also provide their own reasoning for this hypothesis: under a median voter framework, the median voter’s preference to redistributive policies changes according the state of economic development.

Inequality was found to have a significantly negative effect on growth at lower levels of economic development. This negative effect decreased then became significantly positive as a country further developed. Again, this is at odds with what Galor predicted to happen. In their conclusion Lin et al. provide a number of reasons for this result. If economic development precedes financial development, then the negative effects of inequality on capital accumulation are more pronounced at lower levels of development and hence lower levels of financial sophistication (Robinson, 1952). Secondly, in rural areas inequality tends to have a more personal effect. As less developed countries are generally more rural, inequality is more likely to have an impact on social capital and economic growth (Fallah & Partridge, 2007). Conversely, urban areas benefit economically from inequality as it acts as an incentive for entrepreneurs and specialised labour.

2.4. Conclusion

Income inequality as a concept may be relatively intuitive to understand, but providing a clear metric for it is not as simple. The Lorenz curve is useful in graphically representing income distribution, and from this curve one can derive each of the measures discussed in this Literature Review – the Gini coefficient, the Hoover index and the Palma ratio.

(12)

12

Although inequality has always been a topic of debate, Kuznets’ inverted-U hypothesis framed the debate in an economic manner. Kuznets briefly touched upon the classical argument of the relationship between income inequality and economic growth; that inequality increases savings, which in turn creates investment and economic development. There are other arguments supporting inequality. Bhattacharya creates a model in which it is more efficient for rich families to finance investments internally, via bequests, as opposed to externally, which in this case is costly. In this model, a redistributive policy may even have a negative effect on income. Mirrlees also noted that a move towards equality disincentivises hard work, as compensation is less connected to final output, and not at all connected to final output in the case of total equality.

There are also many arguments on the other side of the issue; that inequality is has an adverse effect. Social unrest, resulting from inequality, increases political and economic uncertainty in an economy and as a result reduces investment. In a democracy, an unequal distribution of income will result in redistributive efforts, which take funds away from growth-promoting activities. Stiglitz wrote an entire book about the negative effect of inequality and listed several intuitive arguments against it. Reduced public investment, reduced economic mobility, increased rent seeking, and increased poverty all have a detrimental effect on the economy and are all caused by or, in the case of rent seeking, cause inequality.

There is also the point of view that the relationship can change depending on the development of the country. This view is held by Galor. He discusses a theory in which physical capital is more prized in earlier stages of development, whereas human capital is more valuable in later stages of development. Because of this, savings and physical capital accumulation are more important at first, but as a country develops more equality must be attained in order to invest in the human element of the economy.

The empirical results of preceding literature on this topic have tended to agree. There is evidence for economic growth affecting inequality as Kuznets suggested. Barro and Lin et al. both concluded that income inequality retarded growth at low income levels and had the opposite effect at higher income levels. Voitchovsky found that the effect differed at different segments of the distribution of income inequality.

There appears to be a consistent discord between the empirical findings of previous research and economic theory outlining the relationship. Frequently, the projected relationship is not found when analysing real data. This discord may also be present in this study after the results are found.

Fifty years after Kuznets lamented the lack of data (1955, p. 3), Voitchovsky discussed the limitations of the data in her study (2005, pp. 293-294). This flags up data quality and availability as potential concerns in this paper. However, another limitation that affected the above studies and will not affect this study is age of the data. The most recent data analysed in all of these papers is 2004, by Lin et al. This paper will use the most recent data available. The onset of the financial crisis, as well as inequality having increased further still may mean that there are more recent, relevant conclusions to be drawn.

3. Methodology

3.1. The Model

In this paper, the influence of income inequality on economic growth is studied. Theories about this relationship were discussed above, as well as empirical tests relating to this topic. In this section an

(13)

13

empirical test will be carried out to build upon previous studies. The nature of this topic requires a regression to be performed, using economic growth as the dependent variable and income inequality as the independent variable. Income inequality will be used both in the linear and quadratic form, to test if there is non-linearity present in the relationship.

Therefore, the following regressions will be performed:

𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡= 𝛼 + 𝛽1∙ 𝐺𝑖𝑛𝑖𝑖,𝑡+ 𝜀𝑖,𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑡= 𝛼 + 𝛽1∙ 𝐺𝑖𝑛𝑖𝑖,𝑡+ 𝛽2∙ 𝐺𝑖𝑛𝑖𝑖,𝑡2 + 𝜀𝑖,𝑡

Regressing country t’s growth at time i on its Gini coefficient at time i. The exact formulation of growth and the Gini will be described in the ‘Data’ section of the Methodology below.

Regressions will be performed using the Ordinary Least Squares (OLS) method. In order to use this method there are three least squares assumptions that must be made (Stock & Watson, 2003):

1. The conditional distribution of εi given Xi (the independent variable, in this case Gini) is zero.

That is, although there are differences between the predicted value and actual value of the dependent variable, this is on average zero.

2. Both the dependent and independent variables are independently and identically distributed (i.i.d.).

3. Large outliers are unlikely.

There has been much discussion about whether the state of economic development of a nation has an effect on the relationship between inequality and growth. Barro concludes that at a GDP per capita of 2000 (1985) USD this relationship changes from negative to positive. Using the US Bureau of Labour Statistics’ CPI Inflation Calculator, it is calculated that $2000 in 1985 is has the same value as $4,465.35 in 2016. Barro also used a regression of the Gini coefficient on GDP, making his study similar to mine. Because of this, I expect to find similar results to his. This paper will therefore have two sub-sections of the sample: countries with GDP per capitas consistently below $4,465.35, and countries with GDP per capitas consistently above $4,465.35. Regressions will be performed on the two sub-sections individually as well as the sample as a whole.

Each sub-section contains ten countries, for a total of twenty countries studied. The countries chosen did not only need GDP per capitas consistently below or above the threshold of $4,465.35, but also needed to have many data points available. As a result of these requirements, the countries studied with relatively low GDP per capitas are: Bangladesh, Cote d’Ivoire, Dominican Republic, El Salvador, Mauritania, Morocco, Nigeria, Pakistan, Sri Lanka and Uganda. The countries studied with relatively high GDP per capitas are: Canada, Denmark, Finland, France, Ireland, Israel, Italy, Japan, the United Kingdom and the United States.

3.2. The Data

3.2.1. Income inequality

As mentioned in the Literature Review, there are several ways to measure income inequality. This paper will use the Gini coefficient as its measure of choice. This is because the Gini coefficient is the most widely used and available metric. The fact that it is used in many previous studies, such as Thornton (2001) and Barro (2000), makes the results of this paper easier to compare to earlier findings.

The Gini coefficient has been historically difficult to measure. Data from institutions including the World Bank, United Nations and the OECD are severely limited – many countries have

(14)

14

no Ginis available and even more have a limited amount, making analysis difficult if not impossible using these databanks. However, there is one source available online which provides accurate historical Gini coefficients for many countries over many years – as far back as the 1960s. This is the Standardized World Income Inequality Database, constructed by Frederick Solt. It was constructed for the purpose of studies like those in this paper – broad cross-national research. It uses the Luxembourg Income Study (LIS), the ‘gold standard’ in comparability of cross-country inequality, as a benchmark for comparability, and subsequently creates a dataset with more than double the amount of Ginis than the next largest dataset. The dataset provides a pair of Ginis for each country for each year: gross and net of taxes and transfer payments. This paper will use the net figure as it accounts for redistributive policies enacted by the country and better reflects the real income inequality that is present.

3.2.2. Economic growth

Although there are many variants of what constitutes the income of a nation, gross domestic product (GDP) is without a doubt the most widely used and accepted form of national income. This paper studies many countries across a large time series. Because of this, an issue arises when comparing growth rates. This is the issue of local currency. Two main concerns are present when using local currency: exchange rates and inflation. Both of these concerns, and hence the issue of using local currency, can be eliminated with help from a dataset of the World Bank. This dataset contains the GDP of nations in 2016 US dollars. Using one unit of value for every country in every year means there is no inflation bias present, as well as confusion due to exchange rates. Additionally, it is possible to calculate GDP per capita using population data provided by the World Bank. As a result, using these datasets allows for cross-country comparison of real GDP growth on a per capita basis.

Using OLS regressions is it possible to analyse country-level data to find a relationship between income inequality and economic growth. Furthermore, a more specific relationship can be hypothesised: whether the level of economic development has an effect on growth. This is achieved by using an income threshold previously hypothesised by Barro (2000). The data for the Gini coefficient was compiled specifically for cross-country comparisons in mind, making it suitable for this study. GDP and population statistics are provided by the World Bank and are used in a way to compare real GDP per capita growth across countries.

4. Results

Table 1 shows the results of regressions performed on the dataset, as well as the two subsections of the dataset. The process by which Wealthy and Non-Wealthy countries were defined is explained above in the Methodology. Two regressions were performed on each sample: regressing a country’s growth on its Gini coefficient for the corresponding year (1), and regressing a country’s growth on the Gini coefficient and the square of the Gini coefficient (2).

(15)

15

Table 1

Sample All Countries Non-Wealthy Wealthy

Regression (1) (2) (1) (2) (1) (2) Gini 0.0951435 -2.003865** 0.3706357* -5.728955* -0.1546228 1.538875 (1.22) (-2.90) (2.19) (-2.40) (-1.19) (0.98) Gini2 - 3.005035** - 7.828193* - -2.893391 - (3.06) - (2.56) - (-1.08) Constant 0.0347632 0.3882919** -0.0751951 1.09084* 0.111835*** -0.130917 (1.29) (3.27) (-1.13) (2.37) (2.85) (-0.57) N 678 678 312 312 366 366 R-Squared 0.0022 0.0158 0.0152 0.0356 0.0039 0.0070 t-statistics in parentheses, *: p<0.05, **: p<0.01, ***: p<0.001

The results were expected to be similar to those of Barro (2000). This means that there was an expected negative effect of inequality on growth for the Non-Wealthy countries and an expected positive effect of inequality on growth for the Wealthy countries. This translates into a negative β1

for Regression (1) in the sample of Non-Wealthy and the reverse in the sample of Wealthy.

However, opposite results were found when these regressions were performed. As shown in Table 1, β1 is positive for Non-Wealthy and negative for Wealthy. Additionally, the coefficient was

only significant at the 5% level for Non-Wealthy. The coefficient was not significant for Wealthy. The results for Non-Wealthy could be explained by the classical argument; that increased inequality generates savings which allows for increased investment. However, why this or any other effect is not present in the Wealthy countries is unknown. One would expect the opposite relation to hold: that Wealthy countries’ inequality generates more savings, as those who are better off have a much higher income than those who are better off in Non-Wealthy countries.

The coefficient was also not significant for All Countries in Regression (1). This result was expected, as previous literature has concluded that inequality may have a different effect on growth at differing levels of GDP per capita. Therefore, when studying countries with a wide variety of GDP per capitas, there is no expected significant relationship. This theory was reflected in the results.

What is most interesting, however, are the results of Regression (2). This regression tested whether there was a quadratic relationship between the two variables. Both coefficients were significant at the 1% level for All Countries, implying a quadratic relationship. At first glance, a quadratic relationship would seem to imply that increasing the Gini coefficient has positive and negative effects at differing levels of the Gini. Indeed, the positive Gini2_{coefficient would seem to}

confirm this. However, this is not guaranteed by looking at only the Gini2_{coefficient. It is important}

to remember that the Gini is strictly positive, ranging between 0 and 1. If the regression coefficient for ‘Gini’ was positive, above 0 an increase in the Gini would always mean an increase in growth. However, in this regression the coefficient for ‘Gini’ is not positive, meaning that in the range of 0 and 1, the Gini first decreases, then increases growth.

This was an unexpected result, as this implies that it is not the level of economic development that affects the relationship, but the level of the Gini itself, and therefore the level on inequality in the country. Although this conclusion was not the original intent of the paper, this is certainly relevant when answering this paper’s research question. Using the coefficients, it is possible to determine the lowest point of this ‘U’ curve: When the Gini coefficient is equal to 0.3334, a movement to in either direction would be predicted to stimulate growth.

(16)

16

A similar situation arises when looking at the results of Regression (2) on the Non-Wealthy sample. In this sample, the lowest point of the ‘U’ curve is at a Gini coefficient of 0.3659. The idea of a quadratic relationship is reinforced by the R2_{figures. For every sample, the R}2_{of Regression (2) is}

higher than the corresponding R2_{in Regression (1). Scatter plots, with the corresponding regression}

lines, are presented below. Note the coefficients in Table 1 are scaled differently to those in the graphs below. In the dataset, a growth of 10% would be charted as 0.1, whereas in the graphs the same growth rate would be 10%. This is for illustrative purposes.

(17)

17

This quadratic relationship does not hold, however, with the Wealthy countries. The two coefficients in the regression hold low t-values, indicating insignificance. Some speculation is necessary to explain why this occurs. It may be the case that the relationship becomes weaker as a country becomes wealthier. That is, the coefficients in Regression (2) decrease as GDP per capita increases. This hypothesis is supported by the results in Table 1. Regression (2)’s β1 and β2 are higher

for Non-Wealthy than All Countries, meaning a more steeply-sloped parabola. This may be as wealthier countries are included, the relationship weakens. This is further supported by the lack of significance in Wealthy countries in Regression (1). Looking at the R2_{results also aids this theory. For}

Regression (2):

𝑅𝑊𝑒𝑎𝑙𝑡ℎ𝑦2 < 𝑅𝐴𝑙𝑙 𝐶𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠2 < 𝑅𝑁𝑜𝑛−𝑊𝑒𝑎𝑙𝑡ℎ𝑦2

This implies that the proposed relationship has a better fit as the countries studied become less developed.

It is therefore possible that the results are displaying a mix of effects: a fundamental quadratic relationship between income inequality and economic growth combined with a weakening relationship between these two variables as a country develops.

5. Conclusion

Income inequality can be a very divisive topic. It can cause social unrest and can be indicative of an unfair society. However, its economic relation is much less clear. This paper sought to objectively analyse inequality’s effect on real economic growth, using previously established economic arguments as well as data from the World Bank and Frederic Solt’s Gini databank. The debate on this topic began with discussion of the reverse relationship, with Kuznets and his inverted-U hypothesis, which has been argued over since its early formulation in 1955. However, there has also been development on the relationship this paper has attempted to study. There are proponents on both sides of the argument: those who believe income inequality is a positive force for the economy, and those who believe it is detrimental to the economy. There are also several nuances to arguments that must also be considered: that a change in inequality has a different effect at different segments of income, or that the relationship can change as an economy becomes more developed. The Literature Review of this paper also showed the disparity between theory and results from real data. This disparity was reflected in the results of this paper.

Although the Methodology of this study attempted to reconstruct that of Barro’s (2000), the resulting coefficients from the regressions were surprisingly different. Below his threshold of $2000 in 1985, the countries studied had an opposite, and significant, relationship to income inequality than was expected. Above this threshold no significant relationship at all was found. The results became even more surprising when testing whether a quadratic relationship exists. This relationship was found to be strongest in the Non-Wealthy sub-sample, and weakened as the countries studied became more developed. This suggests two unexpected influences on the relationship: that there is seemingly a least-optimal point of income inequality and that income inequality appears to have less of an influence with increasing development. This suggests further research can be performed to test the validity of these proposed influences.

6. Further research and potential limitations of the study

From these results a number of questions arise. What is the cause of these effects? Why does income inequality seem to have less of an influence on growth as a country develops? Why is there apparently a least-optimal point of income inequality? The regressions performed above are simply

(18)

18

quantitative analyses of data. While these results are certainly useful, economic theory is needed to answer these questions. The Literature Review highlighted the disparities that can arise between these two forms of analysis; the results found from empirical analysis frequently contrasted with economic reasoning provided by other economists. For example, Barro (2005) found the exact opposite relationship that Galor (2000) expected to hold.

There is currently little to no literature available discussing a quadratic relationship between inequality and growth. The same is true for the second effect theorised in this paper: inequality’s influence on growth diminishing as a country develops. Therefore, further research is needed in order to affirm both of these effects. It is also possible to conduct further quantitative research on this topic by using different measures of income inequality, such as the Palma Ratio or Hoover Index, as well as different measures of economic growth, such as Gross National Product or National Income.

One issue that may arise during this study is that of omitted variable bias. Economic growth is, of course, an extremely complicated variable to control for as most economic activities would appear to have some effect on it. However, this sample, and the two sub-samples, appear to be sufficiently large (N≥312) to mitigate any risk of omitted variable bias.

Another issue that may need to be addressed is endogeneity of the variables. This paper aims to study income inequality’s effect on economic growth, not the reverse relationship. If economic growth has a significant effect on inequality, which some see as likely (Kuznets, 1955), endogeneity will be a problem in the regression. This can be fixed by using instrumental variables. However, obtaining specific data for instrumental variables may be difficult for this dataset as it studies, in some cases, extremely poor countries as far back as the 1960s.

The choice of income inequality measure may affect the outcome of the study. This was one of the findings of Voitchovsky, who highlighted the fact that using only one statistic may ‘mask the underlying complexity of the relationship’ (2005, p. 290). Therefore, the results may not show the effect of income inequality on growth, but only the effect of a proxy of income inequality, in this case the Gini coefficient, on growth. This can be rectified by testing whether the results are consistent across multiple indexes of inequality.

(19)

19

Bibliography

Alesina, A., & Perotti, R. (1996). Income distribution, political instability, and investment. European

economic review, 40(6), 1203-1228.

Barro, R. J. (2000). Inequality and Growth in a Panel of Countries. Journal of economic growth, 5(1), 5-32.

Bellù, L. G., & Liberati, P. (2006). Inequality Analysis: The Gini Index (EASYPol module No. 040). Retrieved from Food and Agriculture Organization of the United Nations website: http://www.fao.org/docs/up/easypol/329/gini_index_040EN.pdf

Bhattacharya, J. (1998). Credit market imperfections, income distribution, and capital accumulation. Economic Theory, 11(1), 171-200.

Bourguignon, F. (1981). Pareto superiority of unegalitarian equilibria in Stiglitz' model of wealth distribution with convex saving function. Econometrica: Journal of the Econometric Society, 1469-1475.

Charles-Coll, J. A. (2011). Understanding income inequality: concept, causes and measurement.International Journal of Economics and Management Sciences,1(3), 17-28.

Cobham, A., & Sumner, A. (2013). Putting the Gini back in the bottle? “The Palma” as a policy-relevant measure of inequality. London: King’s College London.

Cohn, A., Fehr, E., & Goette, L. (2009). Fairness and effort-evidence from a field experiment.Unpublished,102(4).

Cooper, C. (2012). Understanding the English “riots” of 2011:“Mindless criminality” or youth “Mekin Histri” in austerity Britain?.Youth and Policy, 109(6), 6.

Dalton, H. (1920). The measurement of the inequality of incomes. The Economic Journal, 30(119), 348-361.

Dittrich, S., & Cheon, M. (2012). Occupy Wall Street.

Fallah, B. N., & Partridge, M. (2007). The elusive inequality-economic growth relationship: are there differences between cities and the countryside?.The Annals of Regional Science,41(2),

375-400.

Galor, O. (2000). Income distribution and the process of development. European Economic

Review, 44(4), 706-712.

Hansen, B. E. (1999). Threshold effects in non-dynamic panels: Estimation, testing, and inference.Journal of econometrics,93(2), 345-368.

Kaldor, N. (1957). A model of economic growth.The economic journal, 67(268), 591-624.

Kuznets, S. (1955). Economic growth and income inequality.The American economic review,45(1),

1-28.

Leibenstein, H. (1960). Economic backwardness and economic growth.

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).

Lin, Y. C., Huang, H. C. R., & Yeh, C. C. (2014). Inequality-growth nexus along the development process. Studies in Nonlinear Dynamics & Econometrics, 18(3), 237-252.

Mirrlees, J. A. (1971). An exploration in the theory of optimum income taxation.The review of economic studies,38(2), 175-208.

Mullainathan, S., & Shafir, E. (2013).Scarcity: Why having too little means so much. Macmillan.

Ncube, M., & Anyanwu, J. C. (2012). Inequality and Arab spring revolutions in North Africa and the Middle East.Africa Economic Brief,3(7), 8.

Palma, J. G. (2014). Has the Income Share of the Middle and Upper‐middle Been Stable around the ‘50/50 Rule’, or Has it Converged towards that Level? The ‘Palma Ratio’ Revisited.Development and Change,45(6), 1416-1448.

Persson, T., & Tabellini, G. (1994). Is inequality harmful for growth?. The American Economic Review, 600-621.

Pigou, A. C. (1912). Wealth and welfare. Macmillan and Company, limited.

(20)

20

Stiglitz, J. E. (1969). Distribution of income and wealth among individuals. Econometrica: Journal of

the Econometric Society, 382-397.

Stiglitz, J. (2012).The price of inequality. Penguin UK.

Stock, J. H., & Watson, M. W. (2003).Introduction to econometrics(Vol. 104). Boston: Addison Wesley.

Thornton, J. (2001). The Kuznets inverted-U hypothesis: panel data evidence from 96 countries.Applied Economics Letters,8(1), 15-16.

Voitchovsky, S. (2005). Does the profile of income inequality matter for economic growth?. Journal

The effect of income inequality on economic growth at differing levels of economic development