The distribution of income shocks after the crises : a panel data analysis on OECD countries between 2004-2012

(1)

The distribution of income shocks after the crises: A panel data analysis

on OECD countries between 2004-2012

By: Luca Mossink

Student number: 10436383

Thesis supervisor: Mr. D.H.J. (Damiaan) Chen Economics & Business

Specialization: Economics & Finance

Amsterdam, 29-06-2016 Abstract

This paper investigated how financial development can lead to an increase in income inequality through financial development in high-income OECD countries after the 2008 financial crisis. By using two different dependent variables, namely the Gini coefficient and the Palma ratio, the evolution of average and the tails of the income distribution were estimated using a panel data analysis for 34 countries between 2004 and 2012. The main focus was the interaction between financial development and the 2008 financial crisis. The results show that the effect on the two dependent variables is not significantly different from zero at any level of testing. Financial development does have a significant and positive effect on the Gini and the Palma but it is relatively small which indicates that there are other factors having a bigger impact on changes in income inequality. This is confirmed by a rho of 97% indicating the importance of country specific circumstances in explaining income inequality.

(2)

Statement of originality

This bachelor thesis is written by Luca Mossink (student) 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

(3)

1. Introduction

One of the most important observations in economic literature is the one of increasing income inequality in the past decades. As the Financial Times (2014) noted, equality lacks relevance if the poor are growing richer. Income inequality has serious consequences for the wellbeing of people and ultimately for the functioning of society (Wilkinson & Pickett, 2009, p. 509). Piketty (2015) shows that top income shares increased in the past 30 years, especially since 1990. There are numerous channels that can lead to more income inequality such as differences in schooling and labor market circumstances for example (Albrecht, Bjorklund, & Vroman, 2003). However, this research will explore a different channel other than presented classically in this field. During the 1980s and 1990s much empirical research has been conducted on inequality and economic growth but this has been inconclusive. The literature did demonstrate the significant role that financial sector development plays with respect to equity of economies. One of the leading causes of persisting inequality was the lack of financial access (Claessens & Perotti, 2007, p. 749). Unequal access to finance can eventually lead to unequal opportunities, which can reinforce any form of economic inequality. After the 2008 financial crisis it has become more troublesome to get access to funds for investment even though the availability increased. In 08/09 the financial sector became ‘too big’, too sharply financed, with almost no equity and no buffers. The debts posed on the society were too high with serious effects on society (Boot, 2016).

This paper will investigate to what extent financial development can lead to more income inequality in high-income OECD countries and how this effect changes after the financial crisis in 2008. In particular, it would be relevant to see how economic shocks can affect inequality. The causes of the financial crisis have been described and researched extensively. Identifying the income groups who are most affected in such crises is likely to be a central issue in the design of safety nets, tax regulation and financial system restructuring (Maloney, Cunnigham, & Bosch, 2004, p. 156; Cobham & Sumner, 2013, p. 2). Attention is paid specifically to the distribution of the total income in a

country, measured by the Gini-coefficient and the Palma ratio. This research will empirically establish this relationship through a panel data analysis for the period between 2004 and 2012. With this methodology the question attempted to answer is: does financial development lead to an increase in income inequality in high-income OECD countries between 2004 and 2012 due to the financial crisis? This paper expects to observe inequality widening for this period: as financial sector develops, the wage gap between income groups increases within countries.

The structure of the paper is as follows. Theoretical considerations concerning inequality and financial development will be discussed in the literature review, section 2. In section 3 the

methodology will be explained and the data examined. The statistical findings will be presented in section 4. In section 5 the results will be discussed and reflected upon. Section 6 concludes this dissertation.

(5)

2. Literature Review

The literature suggests multiple channels through which financial development can lead to income inequality. First, two measures that quantify income inequality will be explained which are used throughout the paper: the Gini coefficient, one of the most frequently used measures, and the Palma ratio, a relatively new measure. After describing in more detail what the relevance is of studying income inequality, this section will continue with discussing the channels that can lead to income inequality, which will be used in the empirical analysis. Last, the hypothesis will be formed.

2.1 Measuring inequality

The Gini is based on the Lorenz Curve, which plots the cumulative share of total income against the cumulative proportion the various quintiles of households. The perfect-equality Lorenz curve is a 45-degree line. The divergence of a Lorenz curve for a given income distribution from the line of perfect equality is measured by the Gini coefficient (Borjas, 2012, p. 293). It must be noted that it does not capture exactly how the income is shared. Although an increase in the Gini coefficient represents an increase in income inequality, summarizing the entire shape of the income distribution into a single number gives an incomplete overview. For example, a transfer of income from the bottom quintile to the top quintile will lead to an increase of the Gini coefficient. However, by transferring income from the second and third quintiles to the top quintile one can obtain a similar numerical increase in the Gini coefficient. Although the numerical increase in the Gini coefficient is the same, the two redistributions are not identical: one cannot be sure where the income change took place within the distribution (Borjas, 2012, p. 294). This proxy for inequality is overly sensitive to variations in the middle and not very sensitive to changes in the tails of the income distribution. These are properties which arguably are acceptable within social values and make it a subjective measure: the weighting of the Gini emphasizes the middle of the distribution. (Atkinson, 1970, p. 262). It does give an overall view of the income distribution however it does not provide a complete view. It is therefore argued that the Gini will capture the average of income inequality.

To take care of the deficiency of the Gini, another model will be estimated using the Palma index as explanatory variable. The Palma ratio is defined as the ratio of the richest 10% of the population's share of gross national income divided by share of the poorest 40% of the population. Palma proposed this ratio since it was striking that the middle class incomes almost always represent about half of gross national income while the other half is split between the richest 10% and poorest 40%. However, the share of those two groups varies considerably across countries (Palma, 2011). The Palma ratio addresses the Gini index's over-sensitivity to changes in the middle of the distribution and insensitivity to changes at the top and bottom, and therefore more accurately reflects income

inequality's economic impacts on society as a whole. It is informative about the magnitude of inequality since a Palma equal to one suggests that the top 10% share earns four times the income of the bottom 40%. The Gini and the Palma express much of the same information as is empirically

(6)

established by Cobham and Summer (2013, p. 23) however the latter stresses the effects in the tails of the distribution. They also state that the Palma ratio is more relevant in making policy changes. It is limited since it leaves out 50% of the income distribution. Lastly, the Palma ratio is relatively new and has not been used in much empirical research, which means there are not many benchmarks to

compare the results to.

2.2 Relevance of studying inequality

At all times it is important to monitor income inequality since large differences in earnings between certain economic groups can have societal effects. Data from the American General Social Survey from 1972 to 2008 showed that in times with less national income inequality, as measured by the Gini, Americans were on average happier. The explanatory variables for this result were ‘perceived fairness’ and ‘general trust’. In other words, the people studied trusted other people less and perceived other people to be less fair in the years with more national income inequality than in the years with less national income inequality. Striking is that the negative association between income inequality and happiness holds for lower-income segment, the bottom 20% till the bottom 60% of the distribution, but not for the top 40% income segment of the studied group (Oishi, Kesebir, & Diener, 2011, pp. 1098-1099). Alesina, Di Tella and MacCulloch (Inequality and happiness: are Europeans and

Americans different?, 2004, p. 2035) also found under European and American respondents that they perceive lower levels of happiness on average when income inequality happens to be high.

Furthermore, certain stages of income inequality have been associated with various health problems. Uneven income can have implications for the health circumstances of the population and of individuals since they lead to different life conditions (Lynch, Smith, Kaplan, & House, 2000, pp. 1201, 1203). This research established a strong correlation between the Gini-coefficient and life expectancy. However, it is important to note that the results strongly depend on the level of social investments by the government. Wilkinson and Pickett (2009, pp. 496, 509) confirm that the health of the population tends to be better where income is more equally distributed. An unequal distribution of wages is associated with numerate social problems such as mental illness, obesity, violence, lack of trust and many more. This finding holds for 56 countries high-income and low-income countries.

These findings could ultimately imply an augmentation of societal health costs as inequality rises assuming the expected age remains constant. This would be an undesired feature within a community as the costs have been rising (World Health Organization, 2010, p. 4). When extrapolating this idea, eventually income inequality could impede economic growth due to more health problems and lower levels of satisfaction among people. Possibly, this has other indirect economic

consequences that are undesired. None the less, it is about the degree of inequality; complete equal societies are not stimulating people to work harder either (Chang, 2014, p. 322). The links between income inequality, financial development and economic growth have extensively been researched as will be shown in Section 2.4.

(7)

2.3 General empirics concerning the income distribution

This section will look at the general statistics concerning the income distribution. Observations of Palma estimated that 80 per cent of the world’s population lives in countries with a median Gini around a value of 40 (2011, p. 121). Income inequality measured by the Gini coefficient appears to be relatively stable within countries and varies significantly between countries during the period of 1947 to 1994 over 112 countries (Li, Squire, & Zou, 1998, pp. 27-28). About 90% of the total variance in the Gini coefficient can be explained by variation across countries, while only a small percentage of the total variance is due to variation over time. The observed annual quantitative change in their OLS regression is small. When a significant trend was detected the country’s Gini changed with less than 1.0% on average. Cobham and Summer (2013, p. 9) confirmed that the middle-income share is more stable between 30.7% and 56.3% but that the top 10% share of income varied between 19% and 65% of total income when they examined the income deciles for the period 1990-2010.

When looking at the distribution of income after a crisis, it appears different income groups are affected depending on several initial economic and institutional circumstances. Maloney et al. (2004, pp. 155-156) employed a quintile regression by type of household during the 1995 Mexican crisis. They observed on average income shocks for household of 30%. Output fell with 6.2 percent while the price level rose with 35 percent during that period. The nominal wages remained constant leading to a decrease of 25 to 35 percent in real wages and unemployment rose from 3.9 percent to 7.4 percent. They observed that the income distribution displayed large differences in the distribution of income of household types before and during the crisis. However there were little changes in their relative positions after the crisis. It could be argued that the crisis had therefore little effect on the earning distribution. Researchers of the World Bank found in an income analysis of the United States that in general top inequality has been rising faster than bottom inequality. Overall income inequality has been on the rise since 1970 but they also find that the effect of the 2008 financial crisis reduced heterogeneity (inequality) among top incomes more than among the bottom: the majority of the states experience a drop in inequality among the top 40% between 2000 and 2010. On average the United States experienced a decrease of 0.02 in the Gini(van der Weide & Milanovic, 2014, p. 14). In a panel data analysis of Diwan (1999, p. 3) the labor share in GDP recovered only partially after a crisis occurred. This result suggested that the resolution of a crisis involves changes in distributions favoring capital and connected insiders. Income inequality will grow after such a shock, especially for the top 10% incomes.

Overall, the earnings distribution remains relatively stable for the middle-income groups but data shows that especially the tails of the distribution (the top 10% and the bottom 40%) fluctuate to a greater extent. As mentioned earlier however, it appears that income inequality is on the rise as Piketty (2015, p. 48), Palma (2011, p. 88) and Jaumotte, Lall and Papageorgiou (2013, p. 272) observe. A financial crisis does not appear to have drastic effects on the earnings distributions overall as studies in Mexico and the USA show (Maloney et al. 2004; van der Weide & Milanovic, 2014).

(8)

2.4 Income inequality, financial development & economic growth

2.4.1 Income inequality and economic growth

Early on in economic literature, Kuznets (1955, pp. 1, 26) strongly associated economic growth with the income distribution of a country. He predicted that inequality would decrease when relative productivity would increase in developed nations, a channel which promotes economic growth. Schumpeter (1934) argued that financial services play an essential role in explaining growth in his theory of economic development. Goldsmith (p.400) argued in 1968 that the financial structure of an economy accelerates economic growth and improves economic performance since it allocates funds to the user where the funds will yield the highest social return (Greenwood & Jovanovic, 1989, p. 1). At these stages it was difficult to back these ideas with empirical results and a causal relationship was not established. Nevertheless, a theory was developed that income inequality, financial development and economic growth were inextricably linked.

2.4.2 Income inequality and financial development

In King and Levine’s (1993) paper it was empirically established that financial development was important in contributing to sustainable economic growth. They defined financial development through financial depth, which is informative about the size of the formal financial intermediary system.In their sample of 80 countries over a 29-year period financial depth measured by the ratio of liquid liabilities to GDP appeared to be robustly correlated with economic development after

controlling for other variables explaining growth1. Forbes (2000, p. 885) finds a significant positive relationship in the short and medium term of a country’s level of income inequality with respect to economic growth suggesting that inequality increases as the economy grows. Honohan (2004, p. 19) finds evidence through his modeling that financial development causes the economy to develop. Elaborating on earlier work, this research confirmed that in order for a country to have sustainable economic growth, financial development was essential in explaining this. Deeper financial markets should lead to economic growth and less inequality. Honohan makes the connection between financial development, growth and inequality and showed through a cross-country regression that finance-intensive growth measured through banking depth is empirically associated with lower poverty ratios. This observation was earlier on established by Galor & Zeira (1993, p. 35) and later confirmed by Clarke, Xu and Zou (2006, p. 578). They observe that as the financial sector deepens, inequality is reduced since the access to credit is eased as the sector deepens. In general empirical research is in accordance that financial development improves income of the poor and how income is shared (Claessens & Perotti, 2007, p. 751)2.

It is therefore argued that financial development improves distribution of income and so that of the poor in earlier stages of developing countries as suggested by eck, Demirg c-Kunt, & Levine

1_{Such as initial income level, education, government spending, inflation, trade openness, measures of}

political liberties and law and order.

(9)

(2009, p. 1). There has been a deepening of financial systems across the world, with much of the deepening concentrated in high-income countries. This deepening has taken place in banking and in stock and bond markets. King and Levine (1993, p. 719) conclude that this component of financial development is strongly and robustly correlated with improvements in economic efficiency and physical capital accumulation. The latter is of importance because it states that as the financial sector develops, the rate of physical capital accumulation increases. This possibly means that wealth is concentrating among specific income groups. In this way the financial development channel does affect the income distribution. Aghion and Bolton (1997, p. 151) predict that through this mechanism a unique invariant wealth distribution will occur within the economy.

However, another theory is proposed by Clarke et al. (2006) who discuss the inequality-widening hypothesis of financial development. They state that a plausible reason why financial development might benefit the rich, especially when institutions are weak, is that the financial system might mainly channel money to the rich and well connected, who are able to offer collateral and who might be more likely to repay the loan and so excluding the poor who are not capable of doing this. As financial sectors become more developed, they might lend more to rich households but continue to disregard the lower incomes who are incapable to provide collateral. As a result, even as the financial sector develops, the poor remain unable to migrate to invest in education or start new businesses. King and Levine (1993, p. 719) another perspective is opposed by them suggesting that inequality increases as the financial sector deepens due to credit market imperfections. The top incomes providing credit could prevent the poor from making productive investments in education or new businesses due to credit constrains arising from asymmetric information which is suggested to slow down the process of converging to income equality. Li, et al. (1998, pp. 27-28) suggest that the initial degree of inequality in the distribution of assets as measured by the distribution of land and through financial depth, are significant determinants of inequality. In theories presented by Banerjee and Newman (1993, pp. 277-278) and Galor and Zeira (1993) it is suggested that among others, trough imperfect capital markets it may lead to divergence of income for the rich and the poor in the long run. Several recent theoretical models have formalized this intuition, suggesting that capital market imperfections would increase income inequality during economic development (Clarke et al., 2006, p. 581).

An empirical study from Li et al. (1998) showed that inequality is predominantly determined by factors that change gradually within countries but this varies across countries. They find statistical evidence that the political economy argument and, of more influence, through the capital market imperfection channel. The latter is measured by the initial degree of inequality in the distribution of assets as a measure of the distribution of land and a measure of financial market development. Through their statistically robust evidence it is inferred that the rich are indeed able to maintain their wealth while the non-rich encounter capital market imperfections, limiting them to accumulate capital

(10)

and make investments. This mechanism reinforces the tendency for unequal distributions of income to endure (Li et al.,1998, pp. 27,42-43).

Aghion, Howitt, & Mayer-Foulkes (2004) find in their cross-country comparisons that

financial sector development has a stronger impact on growth in low- and middle- than in high-income countries. So a deepening of the market has less influence on growth for developed countries. They also noticed an increase in financial services globalization

.

Financial globalization, primarily through foreign direct investment, is associated with an increase in inequality (Jaumotte, Lall, & Papageorgiou, 2013).It appears financial development does not strictly follow economic activity, and the strong relationship between the level of financial development and the rate of economic growth does not simply reflect a positive association between contemporaneous shocks to both financial and economic development (King & Levine, 1993). Financial sector development is significantly and negatively correlated with the Gini coefficient (Clarke et al., 2006, p. 583). Greenwood and Jovanovic (1990) show how financial and economic development might give rise to an inverted U-shaped relationship between income inequality and financial sector development.

Evidence of Perotti and Volpin (2007, p. 14) points out that access to finance is worse in unequal countries which has consequences for the entry of new businesses. It could therefore be concluded that a widening of the income gap to a certain extent is undesirable for society. Cleassens and Perotti (2007, p. 749) state that “inequality can become self-sustained as it affects financial regulation and the evolution of the financial system”.

In conclusion, the empirics show that financial development contributes in explaining economic growth and that it reduces inequality in countries over time. Financial depth is most

important proxy for financial development and is strongly associated with explaining inequality (Li et al., 1998, p. 26; King & Levine, 1993; Beck, Demirg c-Kunt, & Levine, 2009; Clarke et al., 2006; Honohan, 2004). If there is the tendency to prevent new firms from getting access to finance,

hindering them from entering and reducing the ability of the poor to improve their economic situation will have an effect on the distribution of income. Financial development is strongly associated with a reduction in access to capital and with a concentration of funds for specific income groups as the sector deepens (King & Levine, 1993, p. 719).

3. Empirical Research

This section will describe the data and the methodology used for the statistical analysis to investigate the relationship proposed in section 2 regarding financial development and inequality in high-income OECD countries and the effect of the 2008 financial crisis.

(11)

3.1 Descriptive statistics

All data is retrieved from the World Bank (The World Bank, 2015). The Palma ratio has been calculated independently using the methodology of Cobham and Summer (2013, p. 7). The definition is as follows:

Equation 1

Where ‘Income share by highest 10%’ is the percentage share of income or consumption that accrues to the top decile. Whether the data came from income figures or consumption figures depends on availability. ‘Income share by lowest 40%’ is the percentage share of income or consumption that accrues to the two bottom quintiles, which is a sum of the bottom and second 20% share of income as reported. Income shares where gathered from the World Bank database (2015). The figures are subject to rounding (two decimal places). In table 1 all the variables are defined which are used in the panel data analysis.

Variable Definition

GiniWBE World Bank Estimates of the Gini coefficient, Development Research Group.

Data are based on primary household survey data obtained from government’s statistical agencies and World Bank country departments (Milanovic, 2014).

Palma The ratio of the richest 10% of the population's share of gross national income divided by share of the poorest 40% of the population.

llgdp Liquid Liabilities to GDP (%)

Crisis A dummy variable for the 2008 crisis that will have a value of zero before 2008 and will be equal to one from 2008 to 2012

Depthcrisis An interaction variable between llgdp and Crisis Unemployment Total unemployment (% of total labor force)

Education Labor force with secondary education (% of total). Labor force with secondary education is the share of the total labor force that attained or completed secondary education as the highest level of education. Inflation Inflation as measured by the consumer price index reflects the annual

percentage change in the cost to the average consumer of acquiring a basket of goods and services that may be fixed or changed at specified intervals per year.

(12)

The sample consists of 34 high-income OECD countries for the period between 2004 and 2012. This interval has been chosen since it includes the time before the crisis, which should give an indication of the distribution during ‘normal periods’ and four years after the start of the crisis to see if the crisis led to changes in the distribution through financial development. Important to stress is that all the data is denoted in percentages. Some observations were not available for every country making the dataset unbalanced for 34 countries. The established results are from a balanced dataset since Stata (the statistical software used for the analysis) takes out the countries for which no observations were available for the specific period.

Two countries at opposite ends of Gini distribution are compared in table 2. The summary statistics are displayed in table 3.

Table 2; Comparison of relevant variables in 2010

When looking at the distribution of the Gini coefficient it could be inferred that it remains stable with a standard deviation of 5.267 among countries. This confirms the observations of Li et al. (1998) and Palma (2011). The most unequal country in the particular sample is Chile with an average Gini of 51.54 over time. On the contrary, Slovenia is relatively the most equal country with an average Gini of 25.40 over the specified period. The Gini and the Palma have a correlation of 0.73 and

strongly move together. It is therefore not striking that the countries with a higher Gini also have a higher Palma ratio even though for the three control variables (unemployment, inflation and education) there is not much difference as can be seen in table 2. The highest Palma is observed for Canada at 3.49.

The proxy for financial depth, liquid liabilities to GDP, varies strongly among countries. Chili has the least deep markets where Canada is at the other bound of the distribution when looking at column four and five. On average, markets are deepening in this sample with 5.1 percentage points per year per country. Striking is the deflation of 4.48% in Ireland in 2009. This was directly after the financial crisis which had a major impact on the growth of the country (The World Bank, 2015). There is a decreasing trend in inflation after 2008, but overall it stays close to the mean plus or minus one standard deviation as presented in table three, column two and three. The same behavior can be observed in the variable ‘unemployment’. Spain and Greece have the highest level of unemployment, but this is observed only after the crisis. Last, education varies strongly among countries but remains stable over time. Portugal is at the lower bound of the distribution whereas the Czech Republic has the highest percentage of people who finished secondary education.

Country Gini Palma llgdp Depthcrisis Unemployment Inflation Education

Slovenia 24,94 0,85 65,89% 65,89 7,20% 1,84% 60,40%

(13)

3.2 Methodology

To make empirical observations this paper will make use of a panel data analysis. This will describe the development of the financial sector and income inequality over time for 34 different countries. Clarke et al. (2006, p. 581) suggest that the relationship between the evolution of income inequality and financial development would be most likely to show up in short- or medium-run time-series or panel data. Also repeated panel data can be used to estimate the distribution of shocks during normal times (pre-crisis) Maloney et al. (2004, pp. 161-162). Since there are many unobserved variables other than financial development that contribute to explaining income inequality, the development would most likely be explained through a fixed effect panel data analysis.In order to use this, certain assumptions need to be made.

First of all, it is presumed that every country has certain characteristics that bias the dependent variable. This seems reasonable as observed by Li et al. (1998) who emphasize the importance of country specific circumstances in explaining inequality. The fixed effect model will remove this bias of these time-invariant characteristics. Lastly, the time invariant characteristics are unique to the individual country and this should not correlate with other countries. This would suggest that while all other things equal, changes in the dependent variable are due to changes in the independent variables of interest (Stock & Watson, 2015, pp. 401-403). To test whether a fixed or a random effects model is appropriate (more efficient), a Hausman specification test was performed. Taking the null hypothesis in consideration where it should hold that Cov (αi, Xit) = 0, the Chi-squared was equal to 7.67 with a p-value of 0.2635, which is larger than alpha equal to 0.05. Therefore the null hypothesis is rejected and it is inferred that it is unlikely that one will observe a more extreme test statistic under the null

hypothesis. The results from a random panel data analysis do not significantly differ from that of a fixed panel data analysis and would be more efficient than a fixed effect model. However, a random effects model assumes that unobserved variables are uncorrelated with the observed variables. This assumption is unlikely to hold since variables such as llgdp, unemployment and inflation strongly

(1) (2) (3) (4) (5)

VARIABLES N Mean St.dev Min Max

llgdp 273 93.63 57.56 22.81 399.1 Unemployment 306 7.353 3.609 2.300 25.20 Crisis 306 0.556 0.498 0 1 GiniWBE 237 32.12 5.267 23.72 52 Palma 239 1.283 0.447 0.803 3.491 Education 267 47.79 14.92 13.40 79.10 Depthcrisis 273 54.34 67.26 0 399.1 Inflation 306 2.764 2.121 -4.480 12.68

(14)

move with economic growth for example. Since this is the case, based on theory, it is preferred to use the fixed effect model, which will be consistent.

3.2.2 Estimated model

Two models will be estimated because of the use of two different dependent variables, namely the Gini-coefficient and the Palma ratio. These two will be estimated using the same explanatory variables as portrayed in equation 1 and equation 2.

One of the three main explanatory variables will be financial depth, measured by liquid liabilities to GDP (%). This concerns currency plus demand and interest-bearing liabilities of banks and other financial intermediaries divided by GDP. This indicator of financial development is the broadest available because it includes all banks, bank-like and nonbank financial institutions. It is a measure of the quantity and quality of financial services that households, firms, and governments received in total. In general, this number varies across countries and on average increases over time (Beck et al., 2009, p. 4). This measure to proxy financial development is also used by (King & Levine, 1993, p. 720; Beck et al., 2009; Honohan, 2004, p. 7; Li et al., 1998). The second explanatory variable will be a dummy for the 2008 crisis that will have a value of zero before 2008 and will be equal to one from 2008 to 2012. The last explanatory variable of interest will be an interaction variable of the above two.

The literature also elaborates on other variables that explain income inequality through channels different from financial development for which this analysis will control. They are summarized in table 2. The justification for their use will be explained below. First, the level of inflation could affect the distribution of wealth according to Clarke et al. (2006, p. 585). They argue that monetary instability primarily affects the lower and middle segment of the income distribution relatively more than the top segment because the latter have better access to financial instruments that allow them to hedge their exposure to inflation. They hypothesize that inflation will have a positive coefficient. This is also in line with (Diwan, 1999, p. 3). Second, including unemployment in the regression is based on observations of Maloney et al. (2004). During and right after a crisis,

unemployment increases, often drastically. This has effect on the earnings of workers and therefore could affect the income distribution as is argued in section 2.2.1. Third, education is used to proxy investment in human capital using the mean years of secondary schooling. One would expect that a more educated population is able to exert more influence on they pay they receive and have more influence on policy (Li et al, 1998, p. 37; Forbes, 2000, p. 876). It is expected to have a negative sign: as investment in human capital increases, the degree of inequality should decrease. Taking the above into consideration, the following regressions will be estimated:

(15)

Equation 3; Palma ratio as the dependent variable

4. Results

This section will describe the results for the two estimated models based on the methodology

explained in section three. The results of the panel data analysis will be discussed separately as these cannot be compared directly due to the use of a different dependent variable. The primary coefficient of interest is the interaction term ‘Depthcrisis’ since this would describe the effect of a crisis on the income distribution through financial development.

4.1 Gini Estimates

In table 3 the main estimation results are presented regarding the coefficients of the independent variables. First of all, the results show that the depth of the financial sector measured by the ratio of liquid liabilities to GDP is positive: as the financial sector deepens, inequality rises. More specifically, when looking at column three which excludes education since it does not significantly contribute, 1 percentage point increase in llgdp is associated with a 0.0027 increase in the Gini coefficient on average for a given period in time for a specific country. When controlling for inflation and unemployment this result is significant and robust at the 5 per cent level. This is coherent with the literature which predicted a positive effect of financial development on inequality. However, the impact of the depth of the sector on income inequality is relatively small. For the Gini to increase with one full point, the financial sector would have to deepen with 370 percentage points on average keeping all other things equal which is close to the relative size of the financial sector of Luxemburg in this sample.

When looking at the second dependent variable, the crisis has a negative effect on the Gini, meaning that as the crisis occurs and in the years after 2008 till 2012, the Gini coefficient decreases on average keeping all other things equal. More specifically, in a year during or after 2008, the Gini is 0.778 points lower. The sign of the coefficient is equivalent to findings of van der Weide and Milanovic (2014). However, the magnitude of the effect strongly differs since these estimates are almost four times higher than their analysis suggests.

The third variable, financial development interacting with the crisis, captures the focus of this research. The effect of financial development on inequality during and after the 2008 crisis has positive effect on inequality: if the depth of the sector increases with one percentage point, inequality increases with 0.00160 points on average keeping all other things equal. This result is not significantly different from zero at any level of the testing. Furthermore the effect of the interaction term is

approximately 1.69 times smaller than the coefficient of llgdp and the results of an F-test to see whether these coefficients significantly differ suggest that they do not. The probability of finding a

(16)

more extreme outcome was equal to 0.7891. The null hypothesis for this test being that llgdp-depthcrisis = 0. It could therefore be inferred that there is no significant impact of financial development after the crisis on the Gini.

The alpha, indicating the entity fixed effects, appears to be highly significant. This suggests that there are country specific circumstances that effect inequality across entities but not over time. Additionally this is indicated by the value of rho equal to 0.965. This indicates that 96.5% of the variance is due to differences across panels. The alpha of Australia is displayed as reference country under the variable name ‘constant’.

Evidence shows for this period of time that as the financial sector deepens inequality increases which is conflicting with observations of Galor & Zeira (1993, p. 35), Clarke et al. (2006, p. 578) and of the meta-analysis of Cleassens and Perotti (2007, p. 751). Additionally, through financial

development in the time after a crisis inequality increases, as predicted, however this result is insignificant.

Table 4; fixed effect panel data analysis with the Gini-coefficient as dependent variable.

(1) (2) (3) (4)

VARIABLES giniWBE giniWBE giniWBE giniWBE

llgdp 0.00298* 0.00296* 0.00266** 0.00257 (0.00154) (0.00155) (0.00114) (0.00154) Crisis -0.727* -0.776** -0.778* -0.772* (0.384) (0.373) (0.388) (0.424) Depthcrisis 0.00288 0.00305 0.00160 0.00159 (0.00281) (0.00283) (0.00310) (0.00327) Inflation 0.152** 0.189*** 0.189*** (0.0632) (0.0482) (0.0526) Unemployment 0.140*** 0.142*** (0.0431) (0.0423) Education -0.000677 (0.114) Constant 32.12*** 31.69*** 30.60*** 30.16*** (0.208) (0.240) (0.416) (5.571) Observations 211 211 211 203 R-squared 0.039 0.070 0.139 0.136 Number of country 31 31 31 30

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

4.2 Palma Estimates

The effect of a deeper financial system on the Palma ratio is statistically significant at the 5% level of testing. After a 1 percentage point increase in llgdp the Palma ratio increases with 0.000188

percentage points on average keeping all other things constant. This suggests that the top 10% of the population earns 4.000188 times more than the bottom 40% assuming country i has a Palma equal to one at t = 0. If the observations were in line with those of van der Weide and Milanovic (2014, p. 14) one would expect the Palma ratio to decrease after the crisis and when the financial sector deepens

(17)

since heterogeneity (inequality) among top incomes decreases more than among the bottom according to their observations. This would mean that the numerator in equation one is lowered to a greater extent than the denominator, hence a lower Palma. When observing the results in table five, there is no such movement. In fact there is a positive relation between the Palma and llgdp. The crisis does have a negative effect on the dependent variable. This finding is coherent with those of Diwan (1999, p. 3)

It must be noted that a smaller Palma can be due to an increase in the share of income for the bottom 40 percent or a decrease in top 10% share of income. The interaction variable is not significant when tested. The effect of a crisis through financial development is equal to zero and thus has no effect on the earnings distribution. Again, the constant is significant at any level of testing. The same reasoning for the Gini estimates is applied to explain why this occurred. Country specific

circumstances play an important role in explaining income inequality within a country. Table 5; fixed effect panel data analysis with the Palma ratio as dependent variable.

(1) (2) (3) (4)

VARIABLES Palma Palma Palma Palma

llgdp 0.000205* 0.000221** 0.000188** 0.000139

(0.000101) (0.000107) (7.98e-05) (0.000119)

Crisis -0.0483* -0.0494** -0.0474* -0.0448*

(0.0258) (0.0241) (0.0242) (0.0257)

Depthcrisis 0.000194 0.000192 8.92e-05 4.97e-05

(0.000194) (0.000195) (0.000193) (0.000201) Inflation 0.00905* 0.0133*** 0.0140*** (0.00485) (0.00447) (0.00482) Unemployment 0.0103*** 0.0110*** (0.00284) (0.00261) Education -0.00527 (0.00617) Constant 1.278*** 1.252*** 1.166*** 1.534*** (0.0154) (0.0187) (0.0348) (0.340) Observations 212 212 212 203 R-squared 0.031 0.046 0.113 0.095 Number of country 30 30 30 29

Robust standard errors in parentheses

* p<0.01, p<0.05, * p<0.1

(18)

5. Discussion

This section will reflect on the results in section four, the implications of these findings and give a possible explanation on why the hypothesis was not confirmed. Second, the limitations of this study will be discussed. The last paragraph will give an overview and suggestions in which direction further research should continue.

5.1 Results and implications

In observations of Palma (2011) and Cobham and Summer (2013) the bottom 40 percent and top 10 percent of the income distribution experienced much more fluctuations in income shares compared to the middle of the distribution. It was therefore expected that in the Palma model the interaction term would bear more significance than in the Gini model. This observation does not seem to hold. It can be concluded from the data that a deepening of the financial sector is strongly

associated with more inequality over time. This is consistent with the literature and more specifically, it is consistent for high income developed countries since this effect was predominantly for developing countries (Claessens & Perotti, 2007). This implies that as size of the financial sector increases, countries become more unequal regarding the income distribution, but the magnitude of the effect does not lead to large fluctuations in inequality. Furthermore, empirical results of Piketty (2015) and Beck et al. (2009) suggest that between 2004 and 2012 income inequality has risen and so it is not clear whether there is causal relationship between financial depth and income inequality. In this sample, financial market deepen on average 6.08 percentage points per year. Despite that, in order to determine causality, these results should be tested during a period where inequality is decreasing.

In table four and five, column three displayed the model with more predicting power than the model including ‘Education’ as explanatory variable. In fact, the R-squared is 0.018 points lower in case of the Palma and 0.003 points lower in case of the Gini when education was added. The reason why Education might not have contributed to the model is because it is a variable that moves slowly over time. Also, it might not fully capture investment in education since secondary schooling is demanded by law and therefore not a human capital investment decision by the student (Angrist & Krueger, 1991, p. 979). Therefore incorporating it in the model which estimates the effect of a shock such as a financial crisis does not change to the extent as other economic indicators such as inflation and unemployment.

However it can be concluded that both dependent variables interact similarly with the independent variable when looking at the sign and the significance levels. These are equal for every variable across the two models using a different dependent variable, however their magnitude cannot be compared because of the difference in units. None the less, it was expected that the tails of the distribution as measured by the Palma would experience much more variations.

Regarding internal validity, it is often an issue in a panel data analysis that autocorrelation is present because of the time dimension included (Worrall, 2010, p. 183). When testing for this using

(19)

the Woolridge test, autocorrelation was present in the Gini model (p>F=0.0033) but not for the Palma model (p>F=0.5232). Because of the time dimension data is more likely to correlate with other observations however this seems to be more a problem for the Gini estimates and not for the Palma estimates. As can be seen in table six (see appendix) the Gini displays stronger correlations with the other variables included. However, because robust standard errors were used in the regressions, this did not complicate the final results

Additionally, caution must be taken on drawing conclusions regarding external validity on the effect of a crisis. After a shock such as the 2008 financial crisis one cannot exactly conclude which income group was affected most but because of changes in the distribution it will not be informative on whether the same group will be affected during another similar crisis. It would require repeated observations of crises in order to establish this. Even if that would be accomplished, due to different causes and contexts, different patterns of shocks to the same income groups could occur. To illustrate, labor markets could adjust primarily through wages and not through the size of the working force. As a result when wages are held fixed in nominal terms while through inflation the real value decrease, does not experience a large shock. During a crisis, as can be deducted from the data, labor markets did adjust through quantities and so unemployment is temporarily unevenly distributed across different income groups Maloney et al. (2004, pp. 172-173)

Lastly, the interval of the dataset is of importance. Cleassens and Perotti (2007) give an overview of studies which conclude that inequality decreases as the financial sector develops. The focus of these studies are low-income countries and they infer that financial development through financial depth does contribute to a more equal distribution of income. The focus here is on high-income countries and so these results cannot be compared directly.

5.2 Limitations

It must be noted that there are some limitations concerning the methodology presented. First of all, there are some caveats concerning the Gini coefficient as mentioned in section 2.1. Atkinson

(Atkinson, 1970, p. 262) emphasizes the need of a complete ranking of distributions which can only be attained by examining the complete distribution of income. Additionally the Palma ratio was used in order to get a more complete view of the income distribution. This statistic is much more informative about the magnitude of inequality within a country and focus on the tails of the distribution. Overall, these measures do give an indication of the skewness of the income distribution and so whether a country is growing more equal or not, but for further research it is suggested to analyze all the quintiles to have a complete overview of the distribution especially when the focus of the research is on the behavior of the income distribution and it is not intended to give an overview.

When looking at financial development there are some limitations as to measure it through the depth of the sector. It is a proxy that mainly measures the quantity of financial services that

(20)

be subject to policy influence and may therefore be endogenous Li et al. (1998, p. 37). Furthermore Honohan (2004, p. 17) observes that financial development cannot fully be explained by financial depth. He argues that a proxy of financial development must be more subtle and complex than to be informative about the function of the sector.

For further research it would be interesting to make use of a similar approach using a more complete measure of financial development which is more specific about how it captures access to finance. Additionally, it would be interesting to see if the effect holds for crises other than the financial crisis of 2008 and when inequality is decreasing while the financial sector deepens.

6. Conclusion

In conclusion, this research explored whether financial development lead to an increase in income inequality in OECD countries between 2004 and 2012 due to the financial crisis. This was done by estimating two different models where one estimated the Gini coefficient and the other estimated the Palma ratio. The literature predicted different results depending on the type of country. Low income countries would specifically benefit from a deepening of the financial sector whereas the result among high income countries was ambiguous. One of the most famous theories in this field by Kuznets (1955, pp. 16-17) predicted that richer countries would experience lower levels of income inequality. Empirical studies showed that financial development contributed in explaining economic growth and that it reduced inequality in countries over time. Li et al (1998, pp. 27-28) found that about 90% of the total variance in the Gini coefficient was explained by variation across countries, while only a small percentage of the total variance was due to variation over time. Financial depth was the important proxy for financial development and is strongly associated, theoretically and empirically, with explaining inequality (Li et al., 1998, p. 2 King Levine, 199 eck, Demirg c-Kunt, & Levine, 2009; Clarke et al., 2006; Honohan, 2004). Some of these studies also highlighted the fact that there was the tendency to prevent new firms from getting access to finance, hindering them from entering and reducing the ability of the poor to improve their economic situation. This would eventually have an effect on the distribution of income. Financial development is strongly associated with a reduction in access to capital and with a concentration of funds for specific income groups as the sector deepens (King & Levine, 1993, p. 719). The effect of a crisis and the 2008 financial crisis in the United States did not explain major changes in the income distribution (Maloney et al., 2004; van der Weide & Milanovic, 2014). Some evidence contradicted these findings (Diwan, 1999).

The literature predicted to see variations in the tails of the income distribution as the financial sector deepened. In this case, as measured through the Palma ratio, there were significant results indicating such a movement. Additionally, the average inequality as measured by the Gini coefficient showed a similar movement. Therefore financial development, measured through liquid liabilities to GDP, was significant in explaining changes in the Gini and the Palma. The main variable of interest,

(21)

an interaction between financial development and the crisis, was not significant. It can therefore be concluded that there is no significant impact of the 2008 financial crisis on explaining variations in income inequality for the 34 OECD countries. Country specific circumstances play an important in explaining income inequality within a country. The constant in the regression was highly significant at any level of testing and this was in line with observations of Li et al. (1998). Therefore it is suggested that future research focusses on a period of time where inequality is decreasing while the financial sector deepens for similar countries. Finally, more research should be done on country specific circumstances other than financial sector development which appeared to have a relatively small impact on the distribution of income.

(22)

Bibliography

Aghion, P., & Bolton, P. (1997). A Theory of Trickle-Down Growth and Development. Review of Economic Studies, 64, 151-172.

Aghion, P., Howitt, P., & Mayer-Foulkes, D. (2004). The effect of financial development on convergence: Theory and evidence. Working Paper 10358, National Bureau of Economic Research, 1-49.

Albrecht, J., Bjorklund, A., & Vroman, S. (2003). Is There a Glass Ceiling in Sweden? Journal of Labor Economics, 1(21), 145-177.

Alesina, A., Di Tella, R., & MacCulloch, R. (2004). Inequality and happiness: are Europeans and Americans different? Journal of Public Economics, 88(9), 2009– 2042.

Angrist, J. D., & Krueger, A. B. (1991). Does compulsory school attendance affect schooling and earnings? The Quarterly Journal of Economics, 979-1014.

Atkinson, A. B. (1970). On the Measurement of Inequality. Journal of Economic Theory, 2(3), 244-263.

Banerjee, A. V., & Newman, A. F. (1993, April). Occupational Choice and the Process of Development. Journal of Political Economy, 101(2), 274-298.

Beck, T., Demirg c-Kunt, A., & Levine, R. (2009). Policy Research Working Paper. Washington: The World Bank Development Research Group Finance and Private Sector Team.

Boot, A. (2016, February 14). Buitenhof 14 februari: Minister Edith Schippers, Kelderende bankenaandelen, Links over 6 april. (P. Witteman, Interviewer)

Borjas, G. J. (2012). In Labor Economics (Vol. 6, pp. 1-592). New York: McGraw-Hill. Chang, H.-J. (2014). Economics: The User's Guide. Penguin Group.

Claessens, S., & Perotti, E. (2007, July). Finance and inequality: Channels and evidence. Journal of Comparative Economics(35), 748–773.

Clarke, G. R., Xu, L., & Zou, H.-f. (2006, Januari). Finance and Income Inequality: What Do the Data Tell Us? Southern Economic Journal, 3(72), 578-596.

Cobham, A., & Sumner, A. (2013, September). Is It All About the Tails? The Palma Measure of Income Inequality. Center for Global Development Working Paper(343), 1-38.

Diwan, I. (1999). Labor shares and financial crises. The World Bank, Washington, DC. Mimeo, 1-39. Forbes, K. J. (2000). A Reassessment of the Relationship between Inequality and Growth. The

American Economic Review , 90(4), 869-887.

Galor, O., & Zeira, J. (1993). Income distribution and macroeconomics. The review of economic studies, 60(1), 35-52.

Greenwood, J., & Jovanovic, B. (1989). Financial Development, Growth, and the Distribution of Income. NBER Working Paper Series, N. 3189.

Honohan, P. (2004, February). Financial Development, Growth and Poverty: How Close Are The Links? World Bank Policy Research Working Paper 3203, 1-31.

Jaumotte, F., Lall, S., & Papageorgiou, C. (2013). Rising Income Inequality: Technology, or Trade and Financial Globalization? IMF Economic Review, 61, 271–309.

King, R. G., & Levine, R. (1993). Finance and growth: Schumpeter might be right. The Quarterly Journal of Economics, 103(3), 717-737.

Kuznets, S. (1955, March). Economic Growth and Income Inequality. The American Economic Review, 1(45), 1-28.

Li, H., Squire, L., & Zou, H.-f. (1998). Explaining International and Intertemporal Variations in Income Inequality. The Economic Journal, 108(406), 26-43.

(23)

Lynch, J. W., Smith, G. D., Kaplan, G. A., & House, J. S. (2000). Education And Debate Income inequality and mortality: importance to health of individual income, psychosocial environment, or material conditions. British Medical Journal, 320(7243), 1200-1204. Maloney, W. F., Cunnigham, V. W., & Bosch, M. (2004). The Distribution of Income Shocks during

Crises: An Application of Quantile Analysis to Mexico, 1992-95. The World Bank Economic Review , 2(18), 155-174.

McCloskey, D. (2014, August 11). Inequality lacks relevance if the poor are growing richer. Retrieved from Financial Times: https://next.ft.com/content/4c62ddaa-e698-11e3-9a20-00144feabdc0

Milanovic, B. L. (2014, November). All the Ginis, 1950-2012. Retrieved May 5, 2016, from World Bank Data:

http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/0,,contentMD K:22301380~pagePK:64214825~piPK:64214943~theSitePK:469382,00.html

Oishi, S., Kesebir, S., & Diener, E. (2011). Income Inequality and Happiness. Psychological Science, 22(9), 1095– 1100.

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.

Perotti, E., & Volpin, P. (2007, November). Investor Protection and Entry. Tinbergen Institute discussion paper(No. 07-006/2), 1-40.

Piketty, T. (2015). About Capital in the Twenty-First Century. American Economic Review, 105(5), 48-53.

Schumpeter, J. A. (1934). The theory of economic development: An inquiry into profits, capital, credit, interest, and the business cycle . Cambridge, MA: Transaction publishers. Stock, J. H., & Watson, M. W. (2015). Introductin to Econometrics, Second Edition. Pearson

Education Limited.

The World Bank. (2015). World DataBank, World Development Indicators. Retrieved April 30, 2016, from The World Bank:

http://databank.worldbank.org/data/reports.aspx?source=2&country=&series=SI.POV.GINI& period=

van der Weide, R., & Milanovic, B. (2014, July). Inequality Is Bad for Growth of the Poor (But Not for That of the Rich). Policy Research Working Paper 6963, 1-38.

Wilkinson, R. G., & Pickett, K. E. (2009). Income Inequality and Social Dysfunction. Annual Review of Sociology, 35, 493–511.

World Health Organization. (2010). World health report. Retrieved June 8, 2016, from Health systems financing: the path to universal coverage: http://www.who.int/whr/2010/en/

Worrall, J. L. (2010). A User‐Friendly Introduction to Panel Data. Journal of Criminal Justice Education, 2( 21), 182-196.

(24)

Appendix

giniWB

E

Palm a

llgdp Crisis Depthcrisis Inflation Unemploymen t Educatio n giniWBE 1,00 Palma 0,731 1,000 llgdp 0,093 0,019 1,000 Crisis -0,027 -0,075 0,169 1,000 depthcrisis 0,050 -0,039 0,661 0,729 1,000 Inflation 0,323 0,405 -0,113 -0,284 -0,257 1,000 Unemploym ent 0,249 0,078 -0,116 0,136 0,069 -0,179 1,000 education -0,576 -0,508 -0,210 -0,018 -0,135 -0,196 -0,088 1,00

Table 6; Correlations between included variables

List of countries

Australia Germany Mexico Switzerland

Austria Greece Netherlands Turkey

Belgium Hungary New Zealand United Kingdom

Canada Iceland Norway United States

Chile Ireland Poland

Czech Republic Israel Portugal

Denmark Italy Slovak Republic

Estonia Japan Slovenia

Finland Korea Spain

France Luxembourg Sweden

The distribution of income shocks after the crises : a panel data analysis on OECD countries between 2004-2012