What are the main determinants of income inequality?
And what to do next?
MSC Thesis
Student Hans de Jong S2592398
4
1. Introduction
Why is income inequality increasing and what determines this? And why is it more increasing in some countries than in others? Can this inequality be influenced by the government? These kinds of questions are debated with increasing frequency by scientist, economists and politicians. Recently, Piketty (2014) discusses some important causes to inequality where most problematic is the return on capital. Besides, he calls the bad diffusion of knowledge and the low investment in training and skills as an important determinant for inequality. By following other main literature and theories we can investigate the main determinants, with a focus on the reducing effects of income inequality.
The developments and its factors of income inequality have been widely studied where it is mainly focused on the economic origins of inequality in the context of growth, politics and development. Kuznets (1963) was one of the first in this area who came with the influential theory of the inverted U-shape between growth and inequality. However, Piketty (2014) argues that this curve theory was supported by very fragile empirical findings.
Gregorio and Lee (2002) and Glomm & Ravikumar (2002) use panel data in order to assess the impact of education on inequality and their conclusions are inconsistent. Piketty and Saez (2003) and Milanovic (2000) find a strong positive association between equality and redistribution by fiscal policies. Afonso et al (2010) find that redistributive government spending has a significant effect on income distritubtion. They also note that this effect is enhanced in countries with a strong education performance. Finally, there seem to be some other important factors for inequaliy. For example, the quality of institutions seems to be relevant as Gupta (2001) finds a significant increasing effect of corruption on inequality. Here, the issue of reverse causality is important to take in mind (Glaeser, Scheinkman, and Shleifer 2003) Chong and Gradstein (2007). Next, technological progress with the role of complementary skill (Acemoglu, 1998) and
immigration (Chiswick, 1982) seems to have effect on inequality through its impact on the labour market.
5
overarching lesson from the inequality literature. Next to this, we go a step further and see what works in reducing income inequality. Another contribution of this study is that it is not only looking at the correlations of income inequality but also at the drivers, both on the short-and long term. Finally, this this paper provides some political implications.
The organization of this paper is as follows. Section two reviews the literature with the
6
2. Literature review
Education
Education seems to be important for reducing income inequality, though this effect is not entirely straightforward. The balance between composition and the wage compression effect determines the impact of this factor (Knight and Sabot (1983). The effect of composition implies that the higher the tertiary education, the higher income inequality. Concerning the wage compression effect, over time, education decreases inequality. When tertiary education increases, it reduces the wages of highly educated workers because their supply goes up. The reverse effect happens at the bottom where the wages of less-educated workers increases as their supply goes down.
De Gregorio and Lee (2002) provide empirical evidence for education as a related factor to income distribution. Sylvester (2000) finds public education expenditures appear to be associated with a decrease in the level of income inequality. Glomm & Ravikumar (2002), however, do not support these results. According to them, even when the quality of public education is the same across all groups, income inequality may still increase. We will later discuss the role of public spending both on education and in general. When we are looking to the differences of effects, considering the development of the country, it is interesting to make use of the article O’Neill (1995) where he examines the role of education in determining income convergence within countries. The author finds that for developed countries, convergence in education levels has significantly reduced income inequality. When he examines the same relation for LDCs and the world as a whole, the results are significantly different, however. `Despite substantial
7
twice that of the LDCs (Barro and Lee 1993).As changes in production techniques shifts demand towards more skilled labor, higher education levels in developed countries will favor them at the expense of the LDCs, resulting in higher inequality between countries.
An interesting educational mechanism of the increasing effect on income inequality is the signaling role, explored by Spence (1973, p.361): `A greater share of highly educated workers within a cohort may signal to employers that those with less education have less ability, which may also lead to a larger wage differential between highly educated and less-educated workers and thus to higher income inequality. An increase in the levels of education of the highly
educated tends to increase income inequality as the imperfect competition for positions requiring advanced educational credentials raises the wages of educated people even more.’ Rodríguez-Pose & Tselios (2009) support this theory while focusing on the EU region. Following their results, better access to secondary and tertiary education relative to primary education, does not directly imply that this lead to lowering of income inequality.
Hendel et al. (2005) combine this ‘signaling role’ with the credit constraints mechanism. ‘As we improve opportunities for higher education, either by providing direct grants for tuition or by reducing the interest rate that households pay to borrow for an education, more high ability workers get an education and the quality of the unskilled pool drops, lowering the unskilled wage. (Hendel et al.2005, p.843)
Hypothesis 1a: In the short term, education is positively correlated to income inequality and hence, leads to a more unequal income distribution
Hypothesis 1b: In the short term education is negatively correlated to income inequality and hence, leads to a more equal income distribution
Hypothesis 1c: In the long term, education is positively correlated to income inequality and hence, leads to a more unequal income distribution
8
Government spending
Inequality is the result of political forces as much as economic ones. By following Stiglitz (2012) ‘government gives away resources, and through taxes and social expenditures, modifies the distribution of income that emerges from the market, shaped as it is by technology and politics. Governments alter the dynamics of wealth by, for instance, taxing inheritances and providing free public education. Inequality is determined not just by how much the market pays a skilled worker relative to an unskilled worker, but also by the levels of skills that an individual has acquired.’ (Stiglitz 2012, p.54) Furthermore, Stiglitz (2012, p.112) argues that inequality itself increases the gap between the rich and poor through the channel of public investment. ‘The rich don’t need to rely on government for parks or education or medical care or personal security. They can buy all these things for themselves. In the process, they become more distant from ordinary people.’
The government can influence income distribution by spending on public education. We already discussed above that there is little consistency about the public education channel (Sylvester (2000), Glom & Ravikumar (2002) and Hendel et al (2005)). To continue on this, Afonso et al. (2010) empirically examine whether public investment and the efficiency of this really affect the distribution of income. They find that public spending aiming on reducing income inequality and education performance shows a significant effect.
These results are not entirely consistent across countries from their sample though. In terms of affecting income distribution via social public expenditure are the Nordic countries plus Japan, the Netherlands and Slovakia. The less efficient countries appear to be the Anglo-Saxon and Southern European countries plus Germany and France. The efficiency seems to be enhanced in countries with a strong education performance.
Hypothesis 2a: In the short term, government spending is negatively correlated to income inequality and hence, leads to a more equal income distribution
9
R&D expenditures
When we follow the literature about the impact of technology on the labour market and
inequality, economist refer many times to the role of skill-biased technical change (SBTC). This theory emphasizes the idea that technology is more beneficial to skilled workers compared to unskilled workers. This advantage is determined by the size of the market for different inventions. When the group of skilled workers is larger, the market for technologies that
complement skills is also larger, hence more of them will be invented, and new technologies will complementary to skill (Acemoglu, 1998). The same author theoretically shows thatincome inequality could increase due to an increase in the ratio of skilled workers, or in a reduction of cost in acquiring skills. Card & DiNardo (2002) highlight this theory empirically on the U.S. When they look at the period of 1979-1999 they find that some of the early rise in inequality may have been due to rapid technological change, but suspect that the increase in the early 1980s is largely explained by other plausible factors. Goos and Manning (2004) come with the same conclusion for the U.K. Besides, in examing their research objective (relationship between financial globalization and rising income inequality),Jaumotte et al. (2009)find however, that technological progress (their control variable) has had a greater impact than globalization on inequality. Finally, Piketty (2014, p263) emphasizes the role of technology by combining it with the development of education. He describes the idea of a race between education and technology as follows: ‘if the supply of skills does not increase at the same pace as the needs of technology, then groups whose training is not sufficiently advanced will earn less and be relegated to devalued lines of work, and inequality with respect to labor will increase. In order to avoid this, the educational system must increase its supply of new types of training and its output of new skills at a sufficiently rapid pace. If equality is to decrease, moreover, the supply of new skills must increase even more rapidly, especially for the least well educated.’
10
Hypothesis 3a: In the short term, technological progress is positively correlated to income inequality and hence, leads to a more unequal income distribution
Hypothesis 3b: In the long term, technological progress is negatively correlated to income inequality and hence, leads to a more equal income distribution
Hypothesis 3c: In the long term, technological progress is negatively correlated to income inequality and hence, leads to a more equal income distribution
Hypothesis 3d: In the long term, technological progress is negatively correlated to income inequality and hence, leads to a more equal income distribution
Tax and transfer systems
Fiscal policies are an important determinant for income inequality. This seems logically as it is the main redistribution tool regarding income inequality. This redistribution impact varies across OECD countries. Piketty & Saez (2003, p.203) elaborate in their paper the effect redistribution during great shocks: ‘the large shocks that capital owners experienced during the Great
Depression and World War II seem to have had a permanent effect: top capital incomes are still lower in the late 1990s than before World War I. We have tentatively suggested that steep progressive taxation, by reducing the rate of wealth accumulation, has yet prevented the large fortunes to recover fully from these shocks. The evidence for wage series shows that top wage shares were at before World War II and dropped precipitously during the war.’
11
pure inequality considerations put an upper bound on redistribution when the heterogeneity in individuals’ preferences is taken into account.
Hypothesis 4a: In the short term, taxes are negatively correlated to income inequality and hence, leads to a more equal income distribution
Hypothesis 4b: In the long term, taxes are negatively correlated to income inequality and hence, leads to a more equal income distribution
Corruption
According to Gupta (2001) corruption is also an important factor for a significant increase of income inequality, which affects it through several channels. Examples of these channels are tax evasion, poor tax administration. In addition, exemptions that favor the ‘well-connected and wealthy’ population groups, reduces the tax base and the progressivity of the tax system. This might increase income inequality. Second, corruption can affect the targeting group of social programs. The programs can be used to extend the benefits of the wealthy groups, or the transfer of funds by well-connect individuals, which was actually used for poverty-reducing programs. As a third reason the author discusses a society where only a small group of people own the majority of assets: ‘asset owners can use their wealth to lobby the government for favorable trade policies, including exchange rate, spending programs, and preferential tax treatment of their assets. These policies will result in higher returns to the assets owned by the wealthy and lower returns to the assets owned by the less well-to-do, thereby increasing income inequality’(Gupta 2001, p.26). Finally, corruption can affect income distribution via its impact on human capital. As discussed above, corruption can lead to tax evasion and poor tax administration. Therefore, in a given tax system, higher levels of corruption leads to lower tax revenues which in turn leads to lower resources available for public spending, including education.
12
are conducive to lower income inequality, but lower income inequality may be conducive to better institutional quality, as well. There seems to be a vicious circle here.
The effect of corruption on income inequality is not entirely consistent across regions. Apergis et al. (2009) explored this relationship for the U.S and find that long-run corruption has an
increasing effect on income inequality. The effect of this relationship is contrary in Latin-America however. Studies for this region show that lower corruption is associated with higher income inequality (Dobson & Ramlogan, 2009, Andres & Dobson, 2011). The first give the following possible explanation: ‘institutional reform (lowering of corruption) is likely to exacerbate inequality in countries where there is a large informal sector. Firms in this sector have low operating costs arising from their lack of compliance to rules and regulations. It is for this reason that the sector tends to employ the poorest members of society. Since compliance comes with institutional reform and corruption reducing measures, firms will incur rising costs. Furthermore, the actual process of reform requires better trained personnel and support
infrastructure, necessitating new taxes. Higher costs of production, new taxes and more vigilant policing will have a direct impact on employment in the informal sector. A second plausible explanation for the trade-off focuses on the impact of reform on redistributive measures. In many developing countries income redistribution policies are promoted by corrupt elements in society whose primary interest is political power.’ (Dobson & Ramlogan 2009, p.104)
This informal sector explanation, however, does not count for the African region. Increased corruption is positively correlated with income inequality (Gyimah-Brempong, 2001). Besides, a study that specifically looks to the regional differences in this relationship finds that, though there are differences in strengths, there are no differences in positive or negative effects of corruption on income inequality across regions. (Gyimah-Brempong & Munoz Gyimah-Brempong, 2006). Hypothesis 5a: In the short term, corruption is positively correlated to income inequality and hence, leads to a more unequal income distribution
13
Hypothesis 5c: In the long term, corruption is positively correlated to income inequality and hence, leads to a more unequal income distribution
Hypothesis 5d: In the long term, corruption is negatively correlated to income inequality and hence, leads to a more equal income distribution
Economic Growth
Kuznets (1955) was one of the first scientists who elaborated the relation between economic growth and inequality by introducing the theory of the inverted U-curve. He argues that income inequality should increase during the early stages of development (due to urbanization and industrialization) and decrease later on (as industries would already attract a large fraction of the rural labor force). Robinson (1977) demonstrates that overall inequality will first rise and then fall as the share of urban population increases. However, there are also some critics on the widely used Kuznets curve. Piketty (2014) argues that this curve theory was supported by very fragile empirical findings. Kuznets’s findings were established from the data period 1913-1948. Piketty investigated this period for the developed countries and saw that the sharp reduction in income inequality that was observed in almost all the rich countries between 1914 and 1945 was due above all to the world wars and the violent economic and political shocks they entailed. (Piketty, 2014, p15).
What Piketty (2014) shows, however, is a grand theory which describes a positive relationship between economic growth and inequality. When economic growth is lower than the return on capital inequality will rise and, according to Piketty (2014) this is what happening currently. On the other hand (Higins & Williamson (1999) (Ravallion & Chen (1997) find a negative relation between economic growth and income inequality. According to the latter authors distribution improved as often as it worsened in growing economies, and negative growth was often more detrimental to distribution than positive growth.
14
Hypothesis 6b: In the short term, economic growth is negatively correlated to income inequality and hence, leads to a more equal income distribution
Hypothesis 6c: In the long term, economic growth is positively correlated to income inequality and hence, leads to a more unequal income distribution
Hypothesis 6d: In the long term, economic growth is negatively correlated to income inequality and hence, leads to a more equal income distribution
Immigration
The OECD (2012) discusses in its report several recommendations in reducing income inequality. One of them is the promotion of integration of immigrants in the labour market. They are not going into detail but referring to a previous study (Causa and Jean, 2007) which showed that immigrants suffer disproportionately from contract-related labour market dualism.A model on this topic, formulated by Chiswick (1982), puts forward a three-factor economy including capital, high-skilled labour and low-skilled labour. The model assumes and scales immigrants to
low-skilled labour. Hence, immigration increases the supply of low-skilled labour, which in turn reduces their marginal product, reducing their wages. On the other hand, the increased supply of low-skilled labour act as a complement to high-skilled labour, raising the latter’s employment and/or wage rates. This increases the gap between the two categories. This complementarity mechanism is also active with respect to capital, so that higher supply of low-skilled workers increases returns to capital,assuming concentrated ownership of capital, Hence, income inequality is further heightened.
As far as known to the author, there are only a few empirical studies regarding the role of
15
adjust to the labour market. If recent immigrants are excluded from their sample, inequality is still increasing, but a lower rate.’ (Moore & Pacey 2003, p.33)
Hypothesis 7a: In the short term, immigration is positively correlated to income inequality and hence, leads to a more unequal income distribution
Hypothesis 7b: In the long term, immigration is positively correlated to income inequality and hence, leads to a more unequal income distribution
16
3. Methodology
This study investigates the period 1993-2007. This period was chosen because it is the most recent period that did not experience a massive economic shock (from 2008) that might influence the results. Besides, for some important indicators like corruption and immigration, the data collection has only started from around 1993. In addition, a period of 15 years should be sufficient to catch the major effects of the factors on income inequality.
A well-known issue it the limited availability of inequality data. Initially, the database offers information for almost all of the countries in the world. However, after filtering the dataset (at least one data point in every 5 years) along the investigated period and to household income, 63 countries (Appendix A) remain left. In the next section we will explain all the variables in our dataset which is based on the literature review above. In the final section we will discuss the model and methods.
3.1 Dependent variable
17
measures only top income shares and hence is silent on how inequality evolves elsewhere in the distribution (Piketty & Saez, 2005). Considering the richer availability of data (especially the number of countries) for the Gini coefficient, we will use this indicator in our model. The World Income Inequality Database (WIID) will be used for this variable. The units in this database are based on households and others on individuals. This measurement indicator could be very
different because someone could have a very low income but can live in a household where other persons earn a high income which ‘offsets’ her income. Mainly because this is the main used unit for this indicator across countries, we choose consistently for the household income
3.2 Independent variables Education
From the literature review, it can be concluded that human capital, inequality and economic growth are closely linked. The World Data Bank will be used as the source for these data where they make use of the data developed by Barro and Jong W. Lee (1996). In their database there are many indicators that are focused on education. One of the best indicators could be the graduation ratio from secondary schooling, because this can be linked to the working population, which directly influences the income distribution. Besides, following the ‘signaling role’ described in our literature review, the proportion of people with higher degrees seem next ‘average years of schooling’ a relevant variable. However, only a few countries from the sample have data
available for this indicator. Therefore, the regression for this paper will make use of the average years of secondary education among people over age 25. This is also a widely used indicator for education in inequality studies. Barro and Lee (1996) constructed data for this indicator for every 5 years. Hence, the starting point for our study is 1993. This implies that we start for this variable from 1995.
Public investment
18
railways, including schools, offices, hospitals, private residential dwellings, and commercial and industrial buildings. It is calculated as a percentage of GDP.
Corruption
For the corruption variable we make use of the Corruption Perception Index (CPI), from the Quality of Government Institute (QOG). QOG defines corruption as the abuse of public office for private gain. The index is all based on perceptions from business people, risk analysis and the general public regarding the degree of corruption. These perceptions are collected through surveys which translates to a perception index which ranges between 10 (highly clean) and 0 (highly corrupt). It should be mentioned that for this variable, data is collected from 1995 rather than 1992. The reason is that, before 1995 no data exist for this index.
R&D expenditures
The data for this variable is taken from the World Bank. It is defined as expenditures for research and development are current and capital expenditures (both public and private) on creative work undertaken systematically to increase knowledge, including knowledge of humanity, culture, and society, and the use of knowledge for new applications. R&D covers basic research, applied research, and experimental development.
Economic growth
For this variable we use the annual percentage growth rate of GDP at market prices based on constant local currency. The data is again from the World Bank.
Immigration
19
Tax & Transfer systems
For this variable we use the Fiscal freedom indicator, again from the QOG. Fiscal freedom is composed of three quantitative components in equal measure: The top tax rate on individual income, the top tax rate on corporate income, Total tax revenue as a percentage of GDP. In scoring the fiscal freedom factor, each of these numerical variables is weighted equally as one-third of the factor. The country’s fiscal freedom ranges between 0 and 100, where 100 represent the maximum degree of fiscal freedom.
This indicator is based on the redistribution effect, discussed in the literature review. However, it should be mentioned that this using indicator in our model is not directly related to redistribution. This effect is such a complex mechanism, that it in our opinion it is worth a paper apart about this topic. It seems inappropriate to do this following our research topic. The taxing factor is included in our model though, as it such an important reducing factor. However, at the end, the conclusion related to the redistribution effect will discussed but might be weaker.
3.3 Control variables
We control for trade openness in order to capture the effect of globalization. Globalization itself is a broad definition and therefore, its effect is difficult to capture. Following the current
literature, the degree of openness to trade seems to be the best indicator in doing this job. The existing research on the impact of trade openness on inequality derives inconclusive results. Sharma and Morrissey (2006) finds that trade liberalization does appear to be associated with increased inequality. Rodrik (1997) argues that the winners from international trade could compensate the losers hence reducing inequality as a final result. Jaumotte et al (2013) did not find any significant effects at all in this case.
The openness is measures as the sum (% of GDP of) exports and imports of goods and services. The source of this variable is also the World Bank.
20
A summary of descriptive statistics of the relevant variables can be found in table 1 below. A point to note from the table is the lack of observations on the education – and immigration coefficient. As mentioned in our methodology part, the dataset for both variables contain values for every 5 years.
Table 1. Basic descriptive statistics
Variable Obs. Mean Std. Dev. Min Max Inequality 474 42.21986 10.1286 24.32 63 Education 198 2.717222 1.429091 .25 6.89 Public investment 975 21.9132 5.455655 5.385321 43.58616 R&D Exps 604 3.543192 68.4131 .00444 1682 Corruption 647 42.44173 20.63602 4 95 Taxes 842 70.77755 13.33681 29.8 96.7 GDPGrowth 742 1.984027 2.605949 -13.12672 18.8691 Immigration 198 6.68906 8.173654 .036 43.34199 Trade 976 74.5334 36.00734 15.63556 220.4073 Unemployment 990 8.527576 4.528125 .9 35.9 3.4 Estimation model
By following the literature review and our elaboration of the variables, we arrive at the following estimation equation for this model:
Gini= β1+ β2 (Education) + β3 (Publicinvestment ) + β4 (Corruption) + β5 (R&D) +
Β6 (Immigration) + β7 (Economic Growth) + β8 (Taxes) + β9 (Trade) + β10 (Unemployment) +
ε
As shortly discussed above, we will look at the determinants of income inequality. For the determinants we are not just looking at the correlations but also at the drivers of income
21
we look at the correlation based on the yearly values for every indicator. Next, for exploring the short term drivers in section 2 (model 3 & 4), we will look at the first differences in our chosen period. In the third section (model 5 & 6) we will look to the long term correlations for income inequality. More specifically, for every indicator in our model the average value of all years available in the period 1993-2007. Finally, in section 4 (model 7 & 8) the long term drivers of income inequality will be explored. This will be done by looking at the overall change between year 2007 and 1993. However, for some countries there is no data available for one of our indicators in that particular year just mentioned. Hence, when there is at least one data point in both the 3 latest and first years, the country is included in our sample. Thus, section (1) and (3) will look at the indicators that correlate with inequality whereas section (2) and (4) focus on the drivers of inequality.
Since our study examines the main determinants and drivers of income inequality across 63 countries over a 15-year period, our sample is considered to be panel data. Therefore, for our models estimating yearly data the unobserved effect might be correlated with the covariates. In that case, a fixed effect model is applied to this panel in order to control for the issue of
22
4. Results
Table 2 shows a check that was done for collinearity between the independent variables. The correlation coefficient can range from -1, indicating a perfect negative correlation, to +1, indicating a perfect positive correlation, and 0 indicating no correlation at all
According to Cohen (1988) an absolute value of r of 0.1 is classified as small, an absolute value of 0.3 is classified as medium and of 0.5 is classified as large We can see that there exist a large correlation between corruption and education, immigration and education, immigration and public investment, corruption and taxes.
Table 2. Collinearity check
Note Correlation coefficients with p-values underneath in italic and * indicates that it was significant for p <0.05
Using panel- and cross-sectional data for 63 countries we assess the short- and long-run effects of several determinants of income inequality. Both the correlations and drivers will be investigated. The results are shown in table 3. Model (1) & (2) and (5) & (6) are focused on the correlating factors (on the short- and long term, respectively) for income inequality. Model (3) & (4) and (7) & (8) assessing the drivers (on the short- and long term, respectively) of inequality. In all cases
Education Public
Investment R&D exp. Taxes Corruption GDP Growth Immigration Trade Unemployment
23
we start with the basic model where in the next model we control for the effects of trade (globalization) and unemployment. Starting with the short-term correlations, we see in model (1) that education shows surprisingly a positive coefficient, though insignificant. Furthermore the effect of taxes is unexpected, which shows a significantly negative sign. The adding of control variables does not result in noticeable changes. With this different regression now education, public investment and trade show significant negative correlation with income inequality. In addition, the role of taxes and corruption seems to have a significantly increasing effect. When moving to drivers of inequality on the short term in model (3) education shows again a negative effect on inequality. Besides, technological progress (R&D) and corruption increases inequality (on 10% and 5% significance, respectively). Another interesting result is that, when adding the globalization variable in model (4), immigration turns out to be an increasing inequality driver on a 1% significance level.
From column (5) and on we start with estimating the long-term determinants for income inequality. Similarly as for the short-term, education seems to play an important role in inequality in the long term. Besides, taxes show a significant positive sign whereas surprisingly enough, on the long term corruption seems to have a decreasing effect on inequality. Finally, looking at the long-term drivers (7), none of the variables has a significant effect. However, when including our controls (8) (where unemployment is significant on 1%), education returns to be a key factor while also immigration shows a positive sign on a 10% significance level.
When looking at the goodness of fit, mainly the models focusing on the correlations doing a good job. The models which are estimating the drivers showing a lower R2,, meaning that those are less
24
Table 3. Estimation results for dependent variable: Gini
(1) (2) (3) (4) (5) (6) (7) (8) Education 2.375 2.132 -0.759 -1.027 -2.487 -2.145 -2.615 -3.679 (5.271) (7.956) (0.360)** (0.411)** (0.863)*** (0.941)** (2.053) (1.815)** Publicinvestment -0.429 -0.447 -0.045 -0.040 -0.367 -0.345 -0.008 -0.080 (0.360) (0.526) (0.095) (0.141) (0.201)* (0.230) (0.127) (0.115) RDExpenditures -13.974 -14.017 4.320 6.193 -0.347 -0.820 (9.304) (13.437) (2.464)* (3.000)** (1.825) (1.866) Taxes 0.268 0.266 0.108 0.117 0.300 0.302 0.002 -0.046 (0.088)** (0.124) (0.116) (0.071) (0.129)** (0.132)** (0.030) (0.059) Corruption 0.279 0.270 0.020 0.016 -0.299 -0.305 -0.014 -0.026 (0.277) (0.396) (0.010)** (0.011) (0.105)*** (0.105)*** (0.049) (0.033) GDPGrowth 1.185 1.237 0.029 0.094 -0.490 -0.485 -0.047 0.174 (0.545)* (0.858) (0.191) (0.165) (0.665) (0.667) (0.340) (0.329) Immigration 0.112 0.140 0.197 0.352 -0.043 -0.022 0.448 0.577 (1.083) (1.697) (0.131) (0.131)*** (0.113) (0.118) (0.389) (0.323)* Trade -0.003 -0.137 -0.009 0.015 (0.111) (0.054)** (0.034) (0.039) Unemployment -0.063 0.321 -0.339 -0.507 (0.477) (0.289) (0.249) (0.225)** _cons 18.865 21.135 -1.400 -0.858 35.626 37.801 1.397 2.154 (15.685) (29.517) (0.565)** (0.660) (11.867)*** (12.403)*** (1.809) (1.764) R2 0.82 0.54 0.16 0.28 0.47 0.49 0.07 0.18 N 63 63 62 62 59 59 59 54 F Prob. >F 0.183 2.64 0.000 8.98 0.1662 1.56 0.0330 2.25 0.0000 6.45 0.0001 5.24 0.3866 1.08 0.1015 1.80 Modified Wald test (chi2 Breusch-Pagan Test 3.33e+28*** 2.85* 5.41** 0.73 1.49 1.52 7.14*** 6.23**
25
4.2 Robustness check
26
Table 4. Robustness estimation results for dependent variable: Gini
(1) (2) (3) (4) (5) (6) (7) (8) Education 1.573 -1.266 -0.527 -0.842 -1.682 -1.504 -2.270 -4.325 (3.230) (1.211) (0.421) (0.409)** (0.681)** (0.767)* (3.727) (4.314) Corruption 0.357 0.341 0.012 0.002 -0.187 -0.210 0.063 -0.002 (0.152)** (0.059)*** (0.014) (0.013) (0.086)** (0.089)** Publicinvestment -0.354 -0.335 -0.026 0.051 -0.140 -0.147 (0.107)*** (0.087)*** (0.152) (0.144) (0.166) (0.193) (0.169) (0.264) RDExpenditures -16.491 -14.989 4.213 7.896 -0.040 -0.498 -15.832 -13.087 (3.006)*** (1.484)*** (3.212) (3.275)** (1.528) (1.644) (8.778)* (12.688) GDPGrowth 1.240 1.431 0.022 0.172 -0.009 0.073 16.276 12.988 (0.200)*** (0.104)*** (0.203) (0.196) (0.594) (0.617) (7.899)* (14.933) Immigration -0.111 0.268 0.135 0.338 -0.028 0.005 0.230 0.570 (0.310) (0.211) (0.131) (0.143)** (0.087) (0.093) (0.448) (0.579) Taxes 0.235 0.238 0.215 0.189 0.197 0.185 -0.030 -0.027 (0.025)*** (0.024)*** (0.104)** (0.097)* (0.109)* (0.112) (0.119) (0.131) Trade 0.071 -0.185 0.000 -0.027 (0.013)*** (0.069)** (0.030) (0.115) Unemployment -0.267 0.025 -0.292 -0.705 (0.084) ** (0.301) (0.214) (0.635) _cons 21.349 21.161 -1.370 -1.274 33.787 36.741 -2.519 0.887 (11.061)* (3.395)*** (0.675)* (0.699)* (9.382)*** (9.935)*** (4.927) (7.425) R2 0.94 0.99 0.30 0.44 0.38 0.42 0.14 0.29 N 43 43 42 42 39 39 39 34 F Prob. >F 0.2528 8.98 0.2528 8.98 0.1042 1.90 0.0283 2.53 0.2186 2.82 0.0352 2.39 0.8301 0.49 0.6363 0.78 Modified Wald test (chi2 Breusch-Pagan Test 2.8e+28*** 2.2e+27*** 3.98** 2.47 1.51 1.71 8.91*** 4.24**
27
5. Conclusion
This study examines empirically the main determinants of income distribution from a cross-country perspective during the period of 1993-2007. The study first discusses conceptually the determinants of income equality. It then studies empirically the relation between distribution indicators and inequality.
In summary, following our estimation results, the most important determinant for income inequality is education. Both on the short and- long term it seems to be factor that can have a large impact in reducing inequality. It is negatively correlated and is a main driver in reducing inequality, both on the short and- long term. This is also in line with hypotheses 1b and 1d. Following our second hypothesis, we can carefully say that the outcomes are in line with the expectations. The coefficients are not all entirely convincing but both on the short and- long term is government spending negatively correlated with income inequality. This is also stated in both hypotheses for this variable. Whether it really leads to a more equally income distribution is difficult to say as the coefficients for this variable does not significantly turn out to be a reducing driver for inequality. Moving to the role of technological progress; our R&D variable seems to be only short term related to inequality. This is slightly surprising as you would expect that e.g. the process of complementary skill (Acemoglu, 1998) to workers takes some time and hence, the effect on inequality is feasible on the long run. Following our hypothesis 3a, the effect is
positively related to inequality and leads to a more unequal income distribution.
As mentioned earlier, our variable regarding to the redistribution effect of taxes is not the most targeted variable. Using the fiscal freedom indicator as a proxy for redistribution we see that it is positively correlated (and not a driver) to inequality. This is quite intuitively but the explanation might lie in our variable self, where 2 out of 3 components are based on a country’s top tax rate. These taxing rates might not be progressively and hence, favor the top earners. Thus, we reject both hypotheses 4a and 4b.
Corruption seems to be positively correlated with inequality in the short run whereas it benefits (though not significantly in all columns) the equal income distribution in the long run.
28
the short term. On the other hand, the institutional reform explanation for developing countries (Dobson & Ramlogan, 2009, Andres & Dobson, 2011) regarding high reform costs is more plausible for the long term. As the developing countries are slightly predominant in our sample, might this be an explanation for the outcomes. With some carefulness, we follow hypothesis 5a and 5d. Next, GDP growth shows no convincing results in being a significant driver for income inequality. Finally, only when we control for the effect of globalization, immigration turns out to be a driver in increasing inequality both on the short- and long term. Our results are not consistent with those of Moore & Pacey (2003) who are seeing (in their case study) a lower effect of
immigration on the long term. On the short term, trade itself seems to have an important effect on inequality. On the other hand, unemployment only impacts inequality on the long term.
Finally, when empirically looking at the determinants of income inequality, endogenity should be taking into account. Especially for the short term correlations and long term drivers it seems to be an issue when looking at the robustness of our results. Nevertheless, the majority of our
estimations still hold. 5.1 Implications
There are a few relevant policy conclusions that could be drawn from our study. It emerges as an empirical regularity that political determinants of income inequality carry more weight than the economic ones. Firstly, education is the factor which shows the largest impact in reducing income inequality. Rather than in e.g. the case of technological progress governments can exert relatively easily influence in the quality and quantity of educational values. Furthermore, a direct political channel for reducing inequality is public investment. Besides, as it is quite intuitively to recommend though, institutional reform does not seem to work in reducing inequality on the long run; it might even have be an adverse effect. However, this adverse effect is probably mainly applied to developing countries. Next, in line with the recent developments in Europe, the impact that immigrations has on inequality is an interesting one. The increasing effect on inequality, which is even larger on the long run, is a point that matters. As the OECD (2012) advised as well, integration of immigrants might reduce the impact on inequality and raise GDP per capita
29
close the gap between immigrants and non-immigrants’ labour market performance’ (OECD 2012, p.196).
5.2 Limitations
30 References
Acemoglu, D. (1998). Why do new technologies complement skills? Directed technical change and wage inequality. Quarterly journal of economics, 1055-1089.
Afonso et al. (2010) Income distribution determinants and public spending efficiency. The Journal of Economic Inequality, 8(3), 367-389
Andres, A. R., & Ramlogan-Dobson, C. (2011). Is corruption really bad for inequality? Evidence from Latin America. Journal of Development Studies, 47(7), 959-976.
Apergis, N., Dincer, O. C., & Payne, J. E. (2010). The relationship between corruption and income inequality in US states: evidence from a panel cointegration and error correction model. Public choice, 145(1-2), 125-135.
Barro, R. J., & John-Wha, L. (1993). International comparisons of educational attainment. Journal Of Monetary Economics, 32(3), 363-394.
Card, D., & DiNardo, J. E. (2002). Skill biased technological change and rising wage inequality: some problems and puzzles (No. w8769). National Bureau of Economic Research.
Causa, O., & Jean, S. (2007). Integration of immigrants in OECD countries: Do policies matter? OECD Papers, 7(7), 1-33
Chong, A., & Gradstein, M. (2007). Inequality and institutions. The Review of Economics and Statistics, 89(3), 454-465.
Cohen, J. (1988) Statistical Power Analysis for the Behavioral Sciences, 2nd ed. Hillsdale, NJ: Erlbaum.
De Gregorio and Lee (2002. Education and income inequality: new evidence from cross-country data. Review Of Income & Wealth, 48(3), 395-416.
Dobson, S., & Ramlogan-Dobson, C. (2010). Is there a trade-off between income inequality and corruption? Evidence from Latin America. Economics letters, 107(2), 102-104.
Forbes, K. J. (2000). A Reassessment of the Relationship between Inequality and Growth. American economic review, 869-887.
Glaeser, E., Scheinkman, J., & Shleifer, A. (2003). The injustice of inequality. Journal of Monetary Economics, 50(1), 199-222.
Glomm, G. & Ravikumar, B. (2002), Public education and income inequality. European Journal of Political Economy, 19 (2), 289-300
31
Gyimah-Brempong, K., & de Gyimah-Brempong, S. M. (2006). Corruption, growth, and income distribution: Are there regional differences?. Economics of Governance, 7(3), 245-269.
Gyimah-Brempong, K. (2002). Corruption, economic growth, and income inequality in Africa. Economics of Governance, 3(3), 183-209.
Hendel et al. (2005). Educational opportunity and income inequality. Journal of Public Economics, 89(5-6), 841-870.
Higgins, M., & Williamson, J. G. (1999). Explaining Inequality the World Round: Cohort Size, Kuznets Curves, and Openness. National bureau of economic research. (No. w7224).
Hodler, R. (2009). Redistribution and Inequality in a Heterogeneous society. Economica 76(304), 704-718.
Jaumotte, F., Lall, S., & Papageorgiou, C. (2013). Rising Income Inequality: Technology, or Trade and Financial Globalization&quest. IMF Economic Review, 61(2), 271-309.
Kuznets, S. (1955). Economic growth and income inequality. American economic review, 45(1), 1-28.
Milanovic, B. (2000). The median-voter hypothesis, income inequality, and income
redistribution: an empirical test with the required data. European journal of political economy. 16(3), 367-410
Manning, A. (2004). We Can Work It Out: The Impact of Technological Change on the Demand for Low‐Skill Workers. Scottish Journal of Political Economy, 51(5), 581-608.
Moore, E. G., & Pacey, M. A. (2003). Changing income inequality and immigration in Canada, 1980-1995. Canadian Public Policy/Analyse de Politiques, 33-52.
O'Neill, D. (1995). Education and income growth: Implications for cross-country inequality. Journal Of Political Economy, 103(6), 1289.
Piketty, T. (2014). Capital in the twenty-first century. London: The Bellknap Press of Harvard University Press.
Piketty, T., & Saez, E. (2006). The evolution of top incomes: a historical and international perspective (No. w11955). National Bureau of Economic Research
32
Robinson (1977). A note on the U hypothesis relating income inequality and economic development. The American Review. 60(3), 437-440
Rodríguez-Pose, A. & Tselios,V. (2009) Education and Income Inequality in the Regions of the European Union. Journal of regional science, 49(3), 411-437
Rodrik, D. (1997). Has globalization gone too far?. California Management Review, 39(3), 29-53.
Sharma, K., & Morrissey, O. (Eds.). (2006). Trade, growth and inequality in the era of globalization. Routledge.
Spence, M. (1973). Job market Signaling. Quarterly. Journal Of Economics, 87(3), 355-374. Stiglitz, J. (2012). The price of inequality. New York: W.W. Norton & Co
33
Appendix
Appendix A. List of countries
Argentina Latvia Armenia Lithuania Australia Malaysia (1966-) Bangladesh Mauritius Belize Mexico Bolivia Moldova Brazil Pakistan (1971-) Bulgaria Panama Canada Paraguay Chile Peru China Philippines Colombia Poland
Costa Rica Portugal
Cote d'Ivoire Romania
Croatia Russia
Czech Republic Senegal
Dominican Republic Slovakia
Ecuador Slovenia El Salvador Sweden Estonia Tanzania France (1963-) Thailand Ghana Tunisia Guatemala Turkey Honduras Uganda Hungary Ukraine
India United Kingdom
Indonesia United States
