Inequality in the OECD
Political, Institutional, and Demographic Factors
Author: Rienk Jelle Vegter* Supervisor: Dr. Jochen O. Mierau
University of Groningen 24-02-2012
Abstract
This paper investigates three groups of potential factors that may have a relation with income
inequality: political factors, labor market institutions, and demographic factors. A panel study is
performed over the period 1970-1999 using top incomes data of 12 countries in the OECD. Inequality is
described by the dynamics in the shares in total income of three groups: the rich, the upper middle
class, and the rest. The results imply that more left-wing oriented governments, a working population
that is highly unionized, and high participation in elections have an egalitarian effect on the income
distribution. Especially robust relations are found with the income share of the rich.
JEL: D31, D72, J00
Keywords: Top Incomes, Income Inequality, Political Economics, Labor Market Institutions,
Demographics
*
2 1. Introduction
On 12 January 2012 Alan B. Krueger, chairman of the White House Council of Economic Advisers, gave a
speech to the Center for American Progress titled: “The Rise and Consequences of Inequality in the
United States”. In the past, Krueger preferred to refer to dispersion instead of inequality. However, the
rise in income dispersion along many dimensions has forced him to rephrase his terminology and refer
to inequality. This is an alarming process. According to Krueger the rising inequality is causing an
unhealthy division of opportunities, and therefore a threat to economic growth. “Restoring a greater
degree of fairness to the US job market would be good for business, good for the economy, and good for
the country”.
Income inequality has been a research area of interest among economists for a long time. In the early
stage of this research area the focus was primarily on the relationship between income inequality and
economic growth. In his influential paper Kuznets (1955) estimates the relationship between income
inequality and economic growth in the long run. He finds that income inequality follows an inverse-U
shape along the process of economic development. Income inequality first rises with industrialization, as
a relatively small part of de society is benefiting. Thereafter more and more people catch up with the
level of development and therefore inequality starts to decrease again.
While Kuznets (1955) studies the influence of growth on inequality other studies investigate the
influence of income inequality on economic growth. The debate about this relation is still inconclusive.
However, most economists argue that higher income inequality has a negative influence on economic
growth (see, for instance, Alesina and Perotti 1994; Alesina and Rodrik 1994; Persson and Tabellini 1994;
Birdsall et al. 1995; Clarke 1995; Deininger and Squire 1998). More recently, Galor and Moav (2004)
3 approaches named the classical approach and the credit market imperfection approach. The classical
approach states that in early stages of development income inequality is beneficial for economic
growth, as wealthier households have a higher saving propensity, which increases the rate of capital
formation. In latter stages of economic development, in which countries in the OECD (Organisation for
Economic Co-operation and Development) are, the credit market imperfection approach becomes
relevant. This approach induces that inequality has a negative effect on economic growth by impeding
access to student loans and other forms of human capital financing.
At the end of the 20th century, other potential factors correlated with income inequality are considered (see, for instance, Atkinson 1997; Acemoglu 1997; Becker and Tomes 1986; Benabou 1994; Deininger
and Squire 1996, 1998; Durlauf 1996; Galor and Zeira 1993; Galor and Tsiddon 1997; Gottschalk and
Smeeding 1997). This is motivated by the availability of new datasets. In more recent years, a stream of
research is arisen which makes use of data collected from tax statistics to investigate top income shares.
Starting with two contributions of Piketty (2001, 2003) on the long-run distribution of top incomes in
France, a succession of studies has constructed top income shares time series over the long-run for
more than twenty countries to date.
A study by Roine et al. (2009) makes use of these top incomes data to examine the long-run associations
between income inequality and economic growth, financial development, trade openness, top marginal
tax rates, and the size of the government. They propose to divide the society into three groups of
income earners: the rich, the upper middle class, and the rest of the income earners. The rich group is
represented by the fraction of total income received by the top percentile. The upper middle class is
represented by fraction of total income earned by the top decile excluding the top percentile. The rest
4 of the population. Thereafter, they investigate the relation between the variables of interest and the
share in total income of each group of income earners. In that way they can derive the relations with
inequality. The current paper builds forth on the methodology proposed by Roine et al. (2009) but
proposes three other groups of factors that may have a relation with inequality, namely political factors,
labor market institutions, and demographic factors.
The political factors that I propose are partisanship and participation in elections. Partisanship has a
potential relation with inequality as it is generally believed that left-wing political parties opt policies
associated with higher taxation, higher government spending, and more regulation to accomplish a
more equitable society (see, for instance, Hibbs and Dennis 1988; Hicks and Swank 1984; Bradley et al.
2003; Iverson and Soskice 2006). Participation in elections has a potential relation with inequality as
Mahler (2004) and Lijphart (1997) argue that higher political involvement implies that low-income
groups have better understanding of the social system and therefore are better able to reap the benefits
of social transfers, tax allowances, and favorable regulatory and economic development policies.
The labor market institutional characteristics that have a potential relation with inequality are the
degree of unionization and the degree of centralization of the wage bargaining process. Unions are
associated with an aversion against wage inequality of their members (Hibbs 1991). Furthermore, I
suggest that the degree of centralization of the wage bargaining process has a relation with inequality.
Moene and Wallerstein (2003) argue that higher centralization of wage bargaining process has an
egalitarian effect on the income distribution in two ways. It declines the wage differential between in
5 Demographic factors that I propose to have a relation with inequality are the share of the population
that is sixty-five or older and life expectancy. Beramendi and Cusack (2009) argue that when individuals
are pensioned they lose their principle source of market income, wages and salaries from employment.
Therefore inequality is expected to increase when the share of the population at retirement grows
larger. Life expectancy is factor that to my knowledge has never been linked with inequality. The
reasoning is that when life expectancy is higher individuals have more time to exploit their initial wage
differential caused by differences in abilities.
I perform a panel analysis for 12 OECD2 countries in the period 1970-1999 to study the relations of the factors mentioned above with inequality. A common problem in inequality studies is the reliability and
comparability of income inequality measures. Recently, a new version of the World Top Incomes
Database constructed by Alvaredo et al. (2011) became available. This dataset offers fairly
homogeneous, annual, long-run, and broken down by income source data for most countries and I use
this dataset to define the income share of the rich, the upper middle class, and the rest.
The main results of this paper are that more left-wing oriented governments and the share of the labor
force that is unionized have an egalitarian effect on the income distribution. A more left-wing oriented
government and higher unionization have a highly significant negative relation with the income share of
the rich. Furthermore, these factors have positive relation with the income share of the rest, however
the results are less significant on these relations. Participation in elections does only influence the share
of the rich in a negative way. Centralization of the wage bargaining process does not affect inequality at
all. This is also true for the share of the population that is sixty-five or older. Finally, life expectancy only
2
6 has influence on the income share of the upper middle class, although this relation is not highly
significant.
This paper contributes as it further investigates inconclusive relations in the literature with inequality.
Furthermore, a new potential relation with inequality, life expectancy, is suggested. While some of the
relations that I investigate in this paper have been extensively studied before, as can be read in Section
3, this paper contributes as it proposes the use of the top incomes data and the methodology suggested
by Roine et al. (2009) to obtain the effects on inequality.
The remainder of the paper is as follows. In the next section I shortly evaluate the different measures of
income inequality. Section 3 explains the suggested relations with inequality in more detail and gives an
overview of the findings in the literature on these factors. Furthermore, the hypotheses are elaborated.
Section 4 presents the methodology that I employ. In Section 5 I present the data that is used,
descriptive statistics, and its sources. Section 6 shows some figures about the intuition of the relation
between the variables of interest and inequality. In Section 7 I present the estimation results. In Section
8 I give concluding remarks.
2. Measures of income inequality
There are different ways to measure inequality. Over the years, several income inequality datasets are
created. Below I present the most important ones and elaborate on their advantages and disadvantages.
The most complete measure on income inequality is the Gini coefficient. Over the years an increasing
7 World Development Indicators Database3, a database from Deininger and Squire (1996), a database from the Luxembourg Income Studies Institute4, and data from the World Income Inequality Database by the United Nations University5. Using these data in cross-country analyses on income inequality turns out to be complicated. Gini coefficients suffer from problems of comparability. In some countries Gini
coefficients are provided for households whether in other countries Gini coefficient are provided for
individuals. Another problem is that the definition of income differs between the observations. Some
observations are about gross income others are about net income. This makes the use of these kinds of
databases complicated. Furthermore, series suffer from missing observations.
At the start of the 21st century researchers have focused more and more on another topic in de area of the income distribution, namely top income shares. Recently, Alvaredo et al. (2011) constructed the
World Top Incomes Database using income tax records. This dataset contributes to the debate of how
inequality best can be measured as it offers fairly homogeneous, annual, long-run, and broken down by
income source data for most countries. It offers a unique opportunity to better understand income
inequality dynamics (Atkinson et al. 2011). Unfortunately users should be aware of the limitations of the
data. Tax evasion and avoidance makes the tax reforms less reliable. Furthermore, it should be noted
that the collection process of tax data is an administrative process which is not tailored to the scientist
needs. This leads to differences in definitions of income and makes cross country comparison difficult.
In conclusion, for the purpose of the current paper I prefer to use of the World Top Incomes Database
by Alvaredo et al. (2011). In comparison to the Gini coefficients databases this datasets provides more
complete series on the countries of interest. The limitations to the World Top Incomes Database
8 explained above do not make the use of this database inappropriate. Like in all economic data, the data
measures with error the true variable of interest (Atkinson et al. 2011).
3. Potential relations with income inequality
In the introduction I already briefly explained the three groups of factors that may have a relation with
inequality. In this section I describe them in more detail. The first group consists of political factors, the
second group of labor market institutions, and third group of demographic characteristics. For each
group I explain the expected relation with income inequality and what have been the most important
findings in the literature on this relation.
3.1 Political factors
The first factor of interest is the influence of partisanship. It is widely suggested that political parties on
the left side of the political spectrum adopt policies that are more favorable for the poor compared to
political parties on the right (see, for instance, Scheve and Stasavage 2009; Mahler 2004; Beramendi and
Cusack 2009; Atkinson et al. 2011). Political parties can be seen as agents of different economic
interests. Left-wing political parties are traditionally associated with higher taxation, higher government
spending, and more regulation to accomplish a more equitable society (see, for instance, Hibbs and
Dennis 1988; Hicks and Swank 1984; Bradley et al. 2003; Iverson and Soskice 2006). Right-wing political
parties are associated with public policies that facilitate free market outcomes.
Beramendi and Cusack (2009) describe the relation between partisanship and inequality in great detail.
They distinguish between the effects on three different types of income, namely wage income, market
income, and disposable income. In the context of this paper especially market income is relevant.
9 straightforward. Disposable income is defined as total income after taxation. Policy instruments directly
influence disposable income. Among those instruments are regulations such as the minimum wage,
taxes, and transfers. This brings us to the relation with market income inequality, which is the variable of
interest. Individuals anticipate the effects of policy instruments on disposable income. Keeping the
effect of regulation on disposable income in mind, individuals formulate their investment decisions and
labor supply decisions (see, Beramendi 2001). An individual’s investment decision and labor supply
decision defines his market income. Following this line of reasoning policies initiated by politics have an
indirect effect on market income. This leads to the following hypothesis.
H.1 The more left-wing oriented the government, the lower inequality.
The empirical results on the relation between partisan politics and income inequality are mixed. Scheve
and Stasavage (2009) analyze the correlation with inequality, represented by top income shares, in 13
countries in the OECD for a period 1976-2000. Their results are inconclusive about this relation. When
they use left executive as independent variable they find no relation. The left executive variable takes a
value of one for those years where the head of government (President in a presidential system, Prime
Minister/Chancellor in a parliamentary system) was from a left party and zero otherwise. As an
alternative they measure government partisanship as equal to the government’s left-right position as
determined by the weighted (by seats in parliament) left-right positions of the parties in government
and find a significant relation in line with the theoretical prediction. Beramendi and Cusack (2009) find a
result in line with theoretical prediction using Gini coefficients for 13 countries in the OECD in the period
1978-2002. Mahler (2004) does not find an effect using Gini coefficients in a panel study of 14 countries
10 partisanship and inequality may be explained by the different measures of income inequality that are
used. Furthermore, differences in the way partisanship is measured may alter the results.
The second political factor that might have a relation with income inequality is the participation in
national elections. Mahler (2004) and Lijphart (1997) argue that higher participation in national elections
is associated with a higher share of low-income households that is involved in the political process.
Higher political involvement implies that these low-income groups have better understanding of the
social system and therefore are better able to reap the benefits of social transfers, tax allowances, and
favorable regulatory and economic development policies. Moreover, higher participation in national
elections is expected to have an egalitarian effect on the income distribution. Mahler (2004) finds
evidence in favor of this relation in panel study consisting of 14 countries in the OECD in the eighties and
nineties. This results in the following hypothesis.
H.2 The higher the participation in elections, the lower inequality.
3.2 Labor market institutions
The first labor market institution that has a potential effect on income inequality is a labor union. There
are several theoretical relations suggested in the literature. Freeman (1993) argues that unions not only
seek to raise the market income of their members but also favor social expenditures that benefit
low-income groups as a whole by providing public medical, disability, unemployment, and pension benefits.
Unions bargain for higher wages for all their members and not for differential increases for each worker,
thereby reducing wage dispersion (see, Scheve and Stasavage 2009). According to Hibbs (1991) unions
have an aversion to wage inequality. The greater the strength of the union the more aversion against
11 of it. A more organized working class is believed to have a higher ability to influence the redistributive
effort of the government. Following the arguments elaborated above the greater the share of the
working population that is member of a union, the lower inequality.
H.3 The higher the share of the working population that is unionized, the lower inequality.
Checchi and Garcia-Peñalosa (2008) find evidence that higher unionization leads to less inequality for 17
countries in the OECD in the period 1969-2004. Scheve and Stasavage (2009) also find this result for 13
countries in the OECD in the period 1976-2000. Mahler (2004) finds evidence that higher unionization of
the working population leads to lower disposable income inequality, but he finds no relation for
earnings inequality in 14 countries in the OECD in the eighties and nineties. Beramendi and Cusack
(2009) show that union membership has a significant negative relation with wage inequality for 13
countries in the OECD in the period 1978-2002.
The second labor market institution that has a potential effect on inequality is the degree of
centralization of the wage bargaining process. Moene and Wallerstein (2003) explain that the degree of
centralization of the wage bargaining process has an egalitarian effect on the income distribution. They
state that wage dispersion declines as wage bargaining is more centralized. This mechanism works in
two ways. When bargaining occurs at the industry level dispersion is reduced between different firms
within that industry, when it is at the national level dispersion is reduced between wages in different
industries. According to Wallerstein (1999) there are three reasons why more centralized wage
bargaining leads to a more equal income distribution. First, centralized bargaining is considered to be
more efficient than decentralized bargaining, because the centralized bargaining power provides more
12 off against each other. This improves the bargaining position of the workers against the employers.
Third, it is argued that centralized wage bargaining leads to a broadening of norms of redistributive
justice across society. This benefits the low-income groups.
H.4 The higher the level of centralization of the wage bargaining process, the lower
inequality.
The empirical evidence on the relation between the degree of central wage bargaining and income
inequality is mixed. In their study for 13 countries in the OECD in the period 1976-2000 Scheve and
Stasavage (2009) do not find evidence on the negative relation between the degree of centralization of
the wage bargaining process and income inequality. Checchi and Garcia-Peñalosa (2008) cannot find
evidence on this relation in a panel of 17 countries in the OECD in the period 1969-2004. In contrast,
Mahler (2004) presents evidence in favor of this relation for 14 countries in the OECD with 59
observations from the eighties and the nineties.
3.3 Demographic factors
I suggest two demographic factors that might have a relation with income inequality. The first one is the
share of the population that is sixty-five or older. According to Beramendi and Cusack (2009) the higher
the share of the population that is sixty-five or older the more market income inequality there exists.
When individuals are pensioned they lose their principle source of market income, wages and salaries
from employment. Therefore inequality is expected to increase when the share of the population at
retirement grows larger. Beramendi and Cusack (2009) present evidence on this relation for 13 countries
13 H.5 The higher the share of the population that is sixty-five or older, the higher inequality.
The second relation is one which to my knowledge is never been investigated in earlier work. I suggest
that there is a relation between income inequality and life expectancy. The reasoning behind this
relation is that individuals are born with different amounts of ability to earn income in the rest of their
life. Individuals with different ability earn almost equal income in the first period of their life. However,
the difference between their incomes increases when they become older. People with high ability
exploit their capacity and have rising income. Following this way of reasoning individuals with higher
ability also have more resources to build up a higher income after their retirement. So both at their
working age as at retirement differences in income increase. If life expectancy goes up the difference in
income between individuals has more time to diverge. The difference in income grows between
individuals with low ability and individuals with high ability.
H.6 The higher life expectancy, the higher inequality.
4. Methodology
In this section I present the methodology to test the hypotheses stated in Section 3. In Section 2 I
discussed the several income inequality datasets. I conclude that for the purpose of my research to
income inequality I prefer the World Top Incomes Database. Despite the limitations I believe this
dataset offers the best opportunity in making cross-country income inequality analyses. As already
mentioned above I follow Roine et al. (2009) and propose to use three different dependent variables to
capture the dynamics in the income distribution. They make a distinction between broadly speaking the
rich, the upper middle class, and the rest of the society. In the model these groups of income earners
14 Following Scheve and Stasavage (2009) and Roine et al. (2009) I use five year averages instead of annual
data. Averaging across five year periods follows much of the economic growth literature and allows
examining variation over time without specifying precisely how long it takes for changes in explanatory
variables to affect the dependent variable. I expect this sentence is also relevant for inequality data. For
example, I expect that the impact of a more right-wing oriented government is not immediately
captured in income inequality data. Other potential problems are serial correlation and
heteroskedasticity in the residuals. I correct the standard errors for heteroskedasticity using White’s
diagonal coefficient variance method. A remedy for serial correlation in the residuals is the use of first
differenced models6. The models are defined as follows:
(1)
(2)
(3)
Where captures time-period fixed effects and captures country specific fixed effects. I include the
time-period fixed to control for country invariant time specific effects that might have influence on the
dependent variable. I include country specific fixed effects to control for country variant time invariant
effects7.
6 In the Tables 5, 6, 7, 8, 9, and 10 in Section 7 I test for serial correlation in the residuals for each specification. I
follow the procedure proposed by Wooldridge (2002). I run the regression and keep the residuals. Thereafter I rerun the regression including the lagged residuals. A positive significant coefficient indicates serial correlation. It turns out that none of my regressions has problems with serial correlation therefore I only correct for heteroskedasticity.
7
15 The vector includes explanatory variables that correspond with the factors that I believe to have a
relation with inequality following the hypotheses. The vectors of regression parameters to be estimated
are captured by , , and . The error term is represented by .
The vector consists of control variables that are suggested in the literature to also have an impact on
inequality. The vectors , , and present the regression parameters to be estimated for the control
variables. I consider three control variables which I describe below. The data sources and the
abbreviations of the control variables can be found in Table 1 in Section 5.
Globalization measured by the ratio of imports plus exports as a share of GDP is suggested by several
authors (see, for instance, Hurrel and Woods 1995; Reich 1992; Tonelson 2000) to have a relation with
inequality. According to Mahler (2004) the growing movement of goods and capital throughout the
world has driven a wedge into domestic economies, separating those who are well positioned to gain
from globalization from those whose status is increasingly undermined by it. This drives a wedge
between the rich and the poor leading to increasing inequality.
The second control variable I consider is education. Several studies (see, for instance, Adelman and
Morris 1973; Chenery and Syrquin 1975; Ahluwalia 1976; Marin and Psacharopoulos 1976; Winegarden
1979) make notice of the effect of education on inequality. These studies found egalitarian effects of
education. The theoretical relation is less clear. Knight and Sabot (1983) distinguish between the
composition effect and the wage compression effect. The composition effect increases the relative size
of the group with more education and in that way tends to raise income inequality. The wage
coordination effect works in the opposite direction as the higher supply of educated workers tends to
16 The third control variable is GDP per capita. Several authors (see, for instance, Roine et al. 2009; Scheve
and Stasavage 2009) argue that in periods of high economic growth the rich benefit relatively more than
the poor. Income of the rich is relatively more dependent on economic growth because in general they
have jobs that are relatively more dependent on the economic development. Furthermore, their income
consists for a larger part of capital income which makes their income also more dependent on economic
growth.
5. Data
In this section I explain the data I use and how the variables are defined. Table 1 presents an overview of
the dependent variables and the explanatory variables, their abbreviations, and their sources.
[Insert Table 1 here]
5.1 Dependent variables
The top incomes data are obtained from The World Top Incomes Database by Alvaredo et al. (2011). The
main source of this dataset is personal income tax returns. The income data represent gross total
income. This includes labor income, capital income, and business income before taxes and transfers. In a
few cases realized capital gains are included as well. I use three dependent variables that together
capture the dynamics in the income distribution as proposed by Roine et al. (2009). The Top1 variable
measures the fraction of total income received by the top percentile. The Top(10-1) variable measures
the fraction of total income earned by the top decile excluding the top percentile. The Bot90 variable
measures the residual share of income received by the lowest ninety percent of the population. These
variables correspond with the rich, the upper middle class, and the rest of the population, respectively.
17 shares of capital incomes. The upper middle class consists mainly of high wage earners. The income of
the rest of the population is mostly wage income. In Table 2 in I present descriptive statistics of the five
year averaged dependent variables.
[Insert Table 2 here]
5.2 Explanatory variables
Below I sum up how the explanatory variables are defined and what their sources are. Table 3 gives
descriptive statistics on the five year averages of these variables.
[Insert Table 3 here]
5.2.1 Political factors
To measure political color or partisanship I obtain data from the Comparative Political Data Set by
Armingeon et al. (2011) from the Institute of Political Science at the University of Berne. This variable
measures the cabinet composition and is also referred to as the Schmidt-Index. In this paper I refer to
this variable as the GOV variable. This variable is ranged from one to five. Where a one denotes
hegemony of the right-wing (and center) parties, a two denotes dominance of the right-wing (and
center) parties, a three denotes balance of power between right and left, a four denotes dominance of
social-democratic and other left-wing parties, and a five denotes hegemony of social-democratic and
other left-wing parties. The variable is used to test H.1.
The second political factor that might have a relation with income inequality is the participation in
18 variable is obtained from the Comparative Political Data Set by Armingeon et al. (2011) from the
Institute of Political Science at the University of Berne. I refer to this variable as the VT variable. The
variable is used to test H.2.
5.2.2 Labor market institutions
The degree of unionization is represented by union density which is also obtained from the Comparative
Political Data Set by Armingeon et al. (2011) from the Institute of Political Science at the University of
Berne. I refer to this variable as the UD variable. This variable is defined as net union membership as a
proportion of all wage and salary earners in employment and is used to test H.3.
The degree of wage bargaining coordination is gathered from the ICTWSS 3.0 database by the
Amsterdam Institute of Advanced Labour Studies (2011). I refer to this variable as the WC variable. The
variable is used to evaluate H.4. The variable has a range from one to five. A one denotes fragmented
bargaining mostly at company level, a two denotes mixed or alternating industry- and firm level
bargaining, a three denotes industry bargaining with no or irregular pattern setting, a four denotes
mixed industry and economy-wide bargaining, and a five denotes economy-wide bargaining. So the
higher the number, the more centralized the wage bargaining process is.
5.2.3 Demographic factors
The data on the share of the population that is sixty-five or older I obtain from the Comparative Political
Data Set by Armingeon et al. (2011) from the Institute of Political Science at the University of Berne. I
refer to this variable as PAP (pension-aged-population) in the rest of the paper. H.5 is tested with the
19 Life expectancy data is acquired from the Comparative Political Data Set by Armingeon et al. (2011) from
the Institute of Political Science at the University of Berne. I refer to this variable as LE. This variable is
used to evaluate H.6.
6. A first look at the data
To get some feeling with the data I analyze the data in this section. First, I present plots of the
dependent variables to explain how inequality developed over time. Second, I show scatterplots of the
dependent variables with the explanatory variables to get some idea of the tentative relation between
these variables. Third, I present a correlation matrix of the explanatory variables to give information on
a potential problem with multicollinearity.
6.1 Top incomes data
Figure 1 shows how the Bot90 variable developed for each country over time. The overall pattern is that
the share of the bottom ninety percent in total income increased from the seventies to the eighties.
Thereafter it decreased again. For more than half of the countries Bot90 is lower at the end of the
period than at the start. For France, the Netherlands, Norway, Sweden, and Switzerland Bot90 was
higher at the end of the period. Where it is interesting to note that Switzerland is the only country
where the trend at the end of the period was upward. This confirms the observation that the share of
the rest in total income followed a downward pattern at the end of the period.
[Insert Figure 1 here]
Figure 2 presents the share of the upper middle class for each country over time. An overall trend for
20 United Kingdom, and the United States the upper middle class share in total income increased almost
the whole period.
[Insert Figure 2 here]
Figure 3 indicates how the share of the rich developed over time. I observe that at the start of the
period there was a downward trend which means that the share in total income of the rich decreased.
However, in the eighties the share started to increase again. This resulted in higher shares in total
income at the end of the period. In all countries except France, Germany, the Netherlands, and
Switzerland the share of the rich was higher at the end of the period than at the start. In the
Netherlands, Switzerland and Germany the lower share of the rich was the consequence of het
downward trend in the previous periods.
[Insert Figure 3 here]
In conclusion, for a majority of countries there is a tendency towards more inequality at the end of the
period. This tendency can be derived from a decreasing share of the rest of the population in total
income and the increasing share of the rich in total income. Concerning the income share of the upper
middle class there is no change in the overall pattern.
6.2 Relations obtained from scatterplots
To examine the hypotheses proposed in Section 3 I create scatterplots to illustrate whether at least the
21 therefore only about the tentative relations. Note that I suggest that inequality becomes smaller when
the share in total income of the rich and the upper middle class declines and the share of the rest rises.
6.2.1 Political factors
The relation between the political variables and income inequality is tested using H.1 and H.2. For H.1
Figure 4 shows the scatterplots. The figure indicates that inequality declines when governments are
more left-wing oriented. This is demostrated by the positive relation between the GOV and Bot90 and
the negative relation between GOV and Top(10-1), and GOV and Top1.
[Insert Figure 4 here]
Figure 5 shows the scatterplot for H.2. The hypothesis is confirmed from the observations in the graph.
Higher voter turnover, or higher participation in elections, is associated with less income inequality.
Top1 and Top(10-1) are negatively related to VT. Bot90 has a positive relation with VT.
[Insert Figure 5 here]
6.2.2 Labor market institutions
H.3 and H.4 are used to test the relation between labor market institutions and income inequality.
Figure 6 shows a scatterplot for H.3. The figure supports the relation suggested by H.3. The stronger
unions or the higher union density, the higher the share of the rest of the population in total income,
the lower the share of the upper middle class in total income, and the lower the share of the rich in total
22 [Insert Figure 6 here]
For H.4 Figure 7 demostrates that a higher degree of centralized wage bargaining is associated with
lower inequality, indicated by the postive relation between Bot90 and WC, and the negative relation
with Top1, and Top(10-1).
[Insert Figure 7 here]
6.2.3 Demographic factors
The hypotheses H.5 and H.6 suggest the relations between demographic factors and income inequality.
Figure 8 illustrates that Bot90 is positively related with PAP, and that Top(10-1) and Top1 are negatively
related with PAP. This means that according to the Figure 9 the share of the rest in total income
increases with a higher share of the population that is sixty-five or older, and that the share of the upper
middle class and the rich declines when the share of the population that is sixty-five or older increases.
This can be seen as an indication that ineqaulity becomes smaller when the share of the population that
is sixty-five or older grows larger. This indication of the relation is in contrast with H.5. Interesting is if
this contrasting relation is also significant. This is evaluated in the section that follows.
[Insert Figure 8 here]
Figure 9 illustrates that H.6 is only partly confirmed. LE is positively related to Top1. This is in line with
H.6 that states that inequality increases with life expectancy. However, the positive relation with Bot90
23 [Insert Figure 9 here]
From the analysis above I conlude from the scatterplots that there is some evidence in support of H.1,
H.2, H.3, and H.4. Regarding H.5 the evidence is even contrasting the hypothesis. H.6 is only confirmed
concerning the relation with the rich. A further emperical analysis is employed in Section 7 to see if the
relations that are illustrated in the scatterplots are significant.
6.3 Multicollinearity
Another relevent table to show is the correlation matrix of the explanatory variables. High correlation
coefficients between the explanatory variables can give a potential problem of multicollinearity. The
correlation matrix is presented in Table 4. Regression results suffer from multicollinearity as regression
coefficients change largly when highly correlated explanatory variables are added in a single regression.
Furthermore, standard errors are high. Table 4 shows that there is no need to worry about
multicollinearity as the explanatory variables are not highly correlated.
[Insert Table 4 here]
7. Estimation results
In this section I present the results from the estimations. First, I estimate the one-on-one relation
between the suggested factors that might have a relation with the income shares. Second, I estimate a
model involving all factors to test if the results are mutually exclusive, in other words, whether the
coefficients and significance of the explanatory variables change when I include all explanatory variables
24 The focus in this section is on the relations with the rich group of income earners. I do this because in
the empirical analyses below it turns out that I find most robust results concerning this group of income
earners. Thereafter, I describe in a short sub-section the relations with the upper middle class and the
rest. Finally, I briefly evaluate the results with respect to the control variables.
7.1. Relations with the rich group of income earners
Table 5 presents the results for the baseline estimation between the political factors, the labor markets
institutional characteristics, and the demographic factors and the income share of the rich.
[Insert Table 5 and Table 6 here]
7.1.1 Political factors
Column 1 of Table 5 presents the results concerning H.1. The baseline estimation results are in line with
H.1. A more left-wing oriented government has a negative significant relation with the income share of
the rich indicated by a negative coefficient of the GOV variable. Column 7 indicates that when all
explanatory variables are regressed together the result only slightly changes. The relation is significant
at a higher level and the coefficient becomes somewhat smaller. This gives additional evidence in favor
of H.1.
In Table 6 I check the robustness of the results by the inclusion of control variables. In column 1, 2, and 3
I first include the suggested control variables individually. In column 4 I include them all together. The
25 Next to statistical significance it is also important to consider the economic significance. The result in
column 4 of Table 6 for the share of the rich economically implies that when the cabinet composition
index becomes one point higher between two five year averages, so the cabinet becomes more
left-wing oriented, the share of the rich in total income declines with 0.206 percentage points.
In conclusion, I find strong evidence in favor of H.1. The share of the rich in total income declines when
the government becomes more left-wing oriented. Broadly speaking, left-wing politicians favor
redistributive policies more than right-wing politicians. The results are in line with Beramendi and
Cusack (2009) and Scheve and Stasavage (2009).
The column 2 of Table 5 shows the results for H.2. The results confirm H.2. Higher participation in
elections, represented by VT, has a negative significant relation with the share of the rich in total
income. The negative effect on the income share of the rich indicates an egalitarian effect. Theory
argues that higher political involvement implies that low-income groups have better understanding of
the social system and therefore are better able to reap the benefits of social transfers, tax allowances,
and favorable regulatory and economic development policies (see, for instance, Mahler 2004; Lijphart
1997). Note that the result is not highly significant. Column 7 provides additional evidence in favor of
H.2 when I check for mutual exclusivity.
In Table 6 the robustness of the result is checked by the inclusion of control variables. The coefficients
and the significance levels of VT only marginally change. Economically, column 4 states that an increase
of participation in elections of 10 percentage points leads to a decrease of 0.53 percentage points in the
26 7.1.2 Labor market institutions
The baseline estimation result concerning H.3 is presented in column 3 of Table 5. H.3 states that when
a higher share of the working population is unionized inequality declines. The coefficient of UD in
column 3 indicates that union density has a negative relation with the share of the rich in total income.
This is a sign of an egalitarian effect by unions. When all independent variables are regressed together in
column 7 the coefficient marginally changes and becomes more significant.
In Table 6 the robustness of the results is checked by the inclusion of control variables. Colum 1, 2, 3,
and 4 indicate that the coefficient of UD variable only changes marginally and the corresponding
significance levels do not alter. Column 4 shows that a 10 percentage point increase in the share of the
labor force that is unionized leads a 0.85 percentage point decrease in the share of the rich in total
income. This is in line with the theoretical predication. According to Hibbs (1991) the greater the
strength of the union the more aversion against inequality there exists. Furthermore, the result confirms
earlier work by Checchi and Garcia-Peñalosa (2008), Scheve and Stasavage (2009), Beramendi and
Cusack (2009), and Mahler (2004).
Colum 4 of Table 5 shows the baseline estimation result for H.4. The hypothesis states that the higher
the degree of centralization is the more equality there exists. I find no support in favor of this hypothesis
in column 4. The coefficient of WC has the correct sign but is not significant. As expected the coefficient
in column 7, where I check for mutual exclusivity, is not significant either. Moreover, in Table 6 where
control variable are added, the coefficient is also not significant in all regressions. This result confirms
27 7.1.3 Demographic factors
Column 5 of Table 5 shows the baseline estimation result concerning H.5. The result implies that an
ageing society has a negative significant impact on the share of the rich in total income. This means H.5
is rejected as the share of the rich declines as PAP increases. This is a result that could be expected from
Figure 8 in Section 6 where a negative relation between the share of the rich in total income and the
share of the population that is sixty-five or older is shown. In column 7 the significance of the coefficient
disappears and in the robustness checks in Table 6 I also do not find significant results. This means that
in conclusion I do not find a relation between the share of the population that is sixty-five or older and
the share in total income of the rich. This is in contrast to the findings by Beramendi and Cusack (2009).
Column 6 of Table 5 shows the coefficient of the one-on-one relation between life expectancy
represented by LE and the share in income of the rich. H.6 states that a higher life expectancy increases
inequality. The coefficient in column 6 has the correct sign but is not significant. Column 7 confirms this
result. As expected from the previous results I find no significant relation between life expectancy and
the share of the rich in Table 6 where I include control variables.
7.2 Relations with the upper middle class and the rest of the income earners
Below I shortly evaluate H.1 to H.6 concerning the share of the upper middle class and the rest in total
income.
28 7.2.1 Political factors
In Table 7 column 1 and column 7 show some evidence in favor of H.1. The coefficient of the GOV
variable is positive and significant indicating that the income share of the rest grows as the government
becomes more left-wing oriented. Table 8, where I include control variables, shows additional evidence.
Table 9 and Table 10 give no evidence in favor of H.1. So there is no relation between the GOV variable
and the income share of the upper middle class.
H.2 states that when participation in elections is higher, there is less income inequality. Column 2 and
column 7 of Table 7 provide evidence in favor of this hypothesis, however, in Table 8 the significance of
the relation between VT and Bot90 disappears. I conclude that H.2 is rejected concerning the income
share of the rest.
In Table 9 and Table 10 none of the regressions shows a significant relation between VT and Top(10-1).
So there is no relation between participation in elections and the income share of the upper middle
class.
7.2.2 Labor market institutions
Column 3 and column 7 of Table 7 indicate a positive significant relation between UD and Bot90.
Furthermore, when control variables are added in Table 8 the relation is confirmed. This means that
when higher share of the working population is member of a union the share in total income of the rest
29 In none of the regressions in Table 9 and Table 10 I find a significant relation between the share of the
upper class in total income and union density.
H.4 states that when centralization of the wage bargaining process in higher inequality is lower. This
relation is confirmed in none of the regressions in Table 7, Table 8, Table 9, and Table 10. So concerning
the share of the rest and the share of the upper middle class in total income I reject H.4.
7.2.3 Demographic factors
The relation between the percentage of the population that is sixty-five or older and the share in total
income of the rest is indicated in column 5 and column 7 of Table 7 and in Table 8. I find no evidence in
favor of H.5 as there is no significant relation between PAP and Bot90 in any regression.
Column 5 and column 7 in Table 9 and Table 10 indicate that there is also no relation with the upper
middle class of income earners.
H.6 states that when life expectancy is higher, inequality rises. I find evidence in favor of this relation is
column 6 of Table 7. However, in column 7 of Table 7 the significance of this relation disappears.
Furthermore, in Table 8 I find no relation in any regression. I reject H.6 concerning the income share of
the rest.
When I consider the relation with the income share of the upper middle class column 6 and column 7 of
Table 9 indicate that there is significant positive relation between LE and Top(10-1). In Table 9 this
relation is confirmed when I add control variables, although the relation is not highly significant. In line
30 middle class which is an indication of lower inequality. Note that this relation is only significant at a ten
percent level.
7.3 Relations with the control variables
The estimation results of the Top1 variable are shown in Table 6. Column 1 and column 4 indicate that
the only control variable that has significant effect on the share in total income of the rich is
globalization, represented by OP. Column 1 and column 4 of Table 8 show that globalization has a
negative significant relation with the share of the rest in total income, and column 1 and column 4 in
Table 10 show that globalization has positive significant relation with the share in total income of the
upper middle class. These results are in line with the theoretical predication. According to Mahler
(2004) the growing movement of goods and capital throughout the world has driven a wedge into
domestic economies, separating those who are well positioned to gain from globalization from those
whose status is increasingly undermined by it. This drives a wedge between the rich and the poor
leading to increasing inequality.
The other two control variables education, represented by EDU and GDP per capita, represented by
GDP, do not have significant relations with the groups of income earners as can be derived from Table 6, Table 8, and Table 10.
8. Concluding remarks
In this paper I investigate several factors that might have a relation with income inequality in OECD
countries. I propose three groups of factors: political factors, labor market institutions, and demographic
31 represented by union density and the centralization of the wage bargaining process. Demographic
factors are the share of the population that is sixty-five or older and life expectancy.
I perform a panel study for 12 OECD countries in the period 1970-1999. As an indicator for inequality I
use the shares in total income of the rich, the upper middle class, and the rest. In that way this paper
builds forth on the Roine et al. (2009) in using the same dependent variables to describe the shape of
the income distribution. While some of the relations investigated in this paper have been extensively
studied before, this paper contributes as it proposes the use of the top incomes data to obtain the
effects on inequality, while previous studies used mainly Gini coefficients as a measure for inequality.
Furthermore, some relations that have been investigated extensively in previous work were
inconclusive. In addition, this paper considers a new factor, life expectancy. Below I sum up the main
conclusions regarding H.1 to H.6. Note, that I find most robust results on the relations with the income
share of the rich.
The results in this paper indicate that more left-wing oriented governments are associated with a lower
the share of the rich and a higher of the rest in total income (H.1). I find the most significant relation
with the share of the rich in total income. Broadly speaking, left-wing politicians favor redistributive
policies more than right-wing politicians. The results in this paper imply that left-wing involvement in
the government has an egalitarian effect. This result is complementary to the findings of Scheve and
Stasavage (2009). Moreover, the current paper contributes as it also considers the effects on the share
of the rest. Beramendi and Cusack (2009) found no effect of partisanship on inequality using the Gini
32 Higher participation in elections has a negative relation with the share in income of the rich (H.2). Higher
participation in elections indicates higher awareness of the social system by the bottom part of the
income distribution. When individuals at the bottom part of the income distribution are more aware of
the social system they are believed to be better able to reap the benefits of the social system. I find no
effects on the share of the upper middle class and the rest. The detrimental effect on the share of the
rich is in line with the theoretical prediction, as a lower share of the rich indicates less inequality.
However, it is striking that I find no effect on the share of the rest, because the theory about
participation in elections is especially relevant for this group of income earners. The result is partly in
line with Mahler (2004) who found an egalitarian effect of participation in elections on earnings. To my
knowledge this is a novel result as previous research by Mahler (2004) used other indicators for
inequality.
The main argument about unions is that they have an aversion to inequality. When a larger share of the
labor force is unionized the share of the rest in total income rises and the share of the rich in total
income declines (H.3). Although there is no relation with the upper middle class I conclude that strength
of the union has an egalitarian effect. This result confirms a large body of literature on this relation (see,
for instance, Checchi and Garcia-Peñalosa 2008; Scheve and Stasavage 2009; Beramendi and Cusack
2009; Mahler 2004).
Theory argues that higher centralization of the wage bargaining process leads to less wage dispersion
and in that way to less income inequality. The conclusion on the effect of wage bargaining coordination
is very clear (H.4). In none of the estimations I find a significant relation. There is no relation between
centralization of wage bargaining process and income inequality. This result is in contrast to the result
33 confirm my findings. So the debate on this potential factor that has a relation with inequality cannot be
settled.
A larger share of the population that is sixty-five or older is suggested to increase inequality, because
people at that age lose their main source of income, namely salaries and wage income. I find that the
share of the population that is sixty-five or older does not have influence on inequality (H.5). This result
is in contrast with the earlier finding of Beramendi and Cusack (2009) who found that a higher share of
the population that is sixty-five or older leads to more inequality. Important to note is that the relations
obtained from the scatterplot indicate that the sign of the relation is even in contrast to H.5.
Life expectancy only has a relation with the share in income of the upper middle class (H.6). This is a
relation that to my knowledge is novel in the inequality literature. I suggest that when life expectancy
becomes higher differences in income become bigger. When individuals have a higher life expectancy
people in the upper middle class have more time to increase their share in total income. This is in line
with the theoretical prediction. I find no relation with the share of the rich and the share of the rest, so
the hypothesis is only partly confirmed.
In conclusion, the most important findings in this paper are that more left-wing oriented governments
and higher unionization of the working population have an egalitarian effect on the income distribution.
This relation is most robust with the rich group of income earners. I find a less significant relation with
the income share of the rest for these two factors. Furthermore, higher participation in elections also
34 For further research it may be desirable to make use of more groups of income earners to get an even
better insight on the relations with different groups of income earners. One could think of investigating
the relation with the share of income held by the top 0.1 percent. A limitation to this paper is that
because I consider several relations I choose to use a time period where all of the factors had data
available. When one considers the effects individually, more recent observation can be taken into
account. In addition, a dataset that provides data over a longer time span may lead to more robust
35 References
Acemoglu, Daron (1997), “Matching, Heterogeneity, and the Evolution of Income Distribution”, Journal
of Economic Growth, vol. 2, pp. 61–92.
Adelman, Irma and Cynthia T. Morris (1973), “Economic Growth and Social Equity in Developing
Countries”, Stanford University Press, Stanford, California.
Ahluwalia, Montek S. (1976), ‘‘Income Distribution and Development: Some Stylized Facts,’’ American
Economic Review, vol. 66, pp. 128–135.
Alesina, Alberto and Roberto Perotti (1994), "The Political Economy of Growth: A Critical Survey of the
Recent Literature", World Bank Economic Review, vol. 8, pp. 351-371.
Alesina, Alberto and Dani Rodrik (1994), "Distributive Politics and Economic Growth", Quarterly Journal
of Economics, vol. 109, pp. 465-490.
Atkinson, Anthony B. (1997), ‘‘Bringing Income Distribution in from the Cold’’, Economic Journal, vol.
107, pp. 297-321.
Atkinson, Anthony B., Thomas Piketty, and Emmanuel Saez (2011), “Top Incomes in the Long Run of
History”, Journal of Economic Literature, vol. 49, pp. 3-71.
Becker, Gary S. and Nigel Tomes (1986), “Human Capital and the Rise and Fall of Families”, Journal of
Labor Economics, vol. 4, pp. S1–39.
Benabou, Roland (1994), “Education, Income Distribution, and Growth: The Local Connection”, NBER
Working Paper 4798, National Bureau of Economic Research, Cambridge, Massachusetts. Beramendi, Pablo (2001), “The politics of income inequality in the OECD: The role of second order
effects”, Luxembourg Income Study Working Paper, vol. 284.
Beramendi, Pablo, and Thomas R. Cusack (2009), “Diverse Disparities”, Political Research Quarterly, vol.
36 Birdsall, Nancy, David R. Ross and Richard Sabot (1995), "Inequality and Growth Reconsidered: Lessons
from East Asia", World Bank Economic Review, vol. 9, pp. 477-508.
Bradley, David, Evelyne Huber, Stephanie Moller, François Nielsen, and John D. Stephens (2003),
“Distribution and Redistribution in Postindustrial Democracies”, World Politics, vol. 55, pp.
193-228.
Checchi, Daniele and Cecilia García-Peñalosa (2008), “Labour Market Institutions and Income
Inequality”, Economic Politics, vol. 23, pp. 601-649.
Chenery, Hollis B. and Moshe Syrquin (1975), “Patterns of Development, 1950–1970”, Oxford University
Press for the World Bank, London.
Clarke, George R. (1995), "More Evidence on Income Distribution and Growth", Journal of Development
Economics, vol. 7, pp. 403-427.
De Gregorio, José and Jong-Wha Lee (2002), “Education and Income Inequality: New Evidence from
Cross-Country Data”, Review of Income and Wealth, vol. 48, pp. 395-416.
Deininger, Klaus and Lyn Squire (1996), "A New Data Set Measuring Income Inequality" World Bank
Economic Review, vol. 10, pp. 565-591.
Deininger, Klaus and Lyn Squire (1998), “New Ways of Looking at Old Issues: Inequality and Growth”,
Journal of Development Economics, vol. 57, pp. 259–287.
Durlauf, Steven N. (1996), “A Theory of Persistent Income Inequality”, Journal of Economic Growth, vol.
1, pp. 75–93.
Freeman, Richard B. (1993), “How much has de-unionization contributed to the rise in male earnings
Inequality?”, Uneven tides: Rising inequality in America, pp. 133-163, New York: Russell Sage.
Galor, Oded and Omer Moav (2004), “From Physical to Human Capital Accumulation: Inequality and the
37 Galor, Oded and Daniel Tsiddon (1997), “The Distribution of Human Capital and Economic Growth”,
Journal of Economic Growth, vol. 2, pp. 93–124.
Galor, Oded and Joseph Zeira (1993), “Income Distribution and Macroeconomics”, Review of Economic
Studies, vol. 60, pp.35–53.
Gottschalk, Peter and Timothy M. Smeeding (1997), “Cross-National Comparisons of Earnings and
Income Inequality”, Journal of Economic Literature, vol. 35, pp. 633–687.
Hibbs, Douglas A. and Christopher Dennis (1988), “Income distribution in the United States”, The
American Political Science Review, vol. 82, pp. 467-490
Hibbs, Douglas A. (1991), “Market Forces, Trade Union Ideology and Trends in Wage
Dispersion”, Acta Sociologica, vol.34, pp. 89-102.
Hicks, Alexander and Duane H. Swank (1984), “Governmental Redistribution in Rich Capitalist
Democracies”, Policy Studies Journal, vol. 13, pp. 265-286.
Hurrel, Andrew and Ngaire Woods (1995), “Globalisation and Inequality”, Millenium: Journal of
International Studies, vol. 24, pp. 447-470.
Iverson, Torben and David Soskice (2006), “Electoral Institutions and the Politics of Coalitions: Why
Some Democracies Redistribute More Than Others”, American Political Science Review, vol. 100,
pp. 168-181.
Knight, John B. and Richard H. Sabot (1983), ‘‘Educational Expansion and the Kuznets Effect’’, American
Economic Review, vol. 73, pp. 1132–1136.
Krueger, Alan B. (2012), “The Rise and Consequences of Inequality in the United States”, The Center for
Economic Progress, http://www.americanprogress.org/events/2012/01/pdf/krueger.pdf. Kuznets, Simon (1955),” Economic Growth and Income Inequality”, The American Economic Review, vol.
38 Leigh, Andrew (2007), “How Closely Do Top Income Shares Track Other Measures of Inequality?”, The
Economic Journal, vol. 117, pp. 619-633.
Lijphart, Arend (1997), “Unequal participation: Democracy’s unsolved dilemma”, American
Political Science Review, vol. 91, pp. 1-14.
Mahler, Vincent A. (2004), “Economic Globalization, Domestic Politics, and Income Inequality in the
Developed Countries: A Cross-National Study”, Comparative Political Studies, vol. 37, pp. 1025
1053.
Marin, Alan and George Psacharopoulos (1976), ‘‘Schooling and Income Distribution’’, Review of
Economics and Statistics, vol. 58, pp. 332–338.
Moene, Karl Ove and Michael Wallerstein (2003), “Earnings Inequality and Welfare Spending: A
Disaggregated Analysis”, World Politics, vol. 55, pp. 485-516.
Persson, Torsten and Guido Tabellini (1994), "Is Inequality Harmful for Growth?", American Economic
Review, vol. 84, pp. 600-621.
Piketty, Thomas (2001), “Les Hauts Revenus en France au 20ème siècle”, Paris: Grasset.
Piketty, Thomas (2003), “Income Inequality in France, 1901-1998”, Journal of Political Economy, vol. 111,
pp. 1004-1042.
Reich, Robert B. (1992), “The work of nations: Preparing ourselves to 21st century capitalism”. New York: Vintage.
Roine, Jesper, Jonas Valchos, and Daniel Waldenström (2009), “The long-run determinants of inequality:
What can we learn from top income data? Journal of Public Economics, vol. 93, pp. 974-988.
Scheve, Kenneth and David Stasavage (2009), “Institutions, Partisanship, and Inequality in the Long
Run”, World Politics, vol. 61, pp. 215-253.
Tonelson, Alan (2000), “The race to the bottom: Why a world worker surplus and uncontrolled
39 Verbeek, M. (2008), “A guide to modern econometrics”, John Wiley and Sons, Ltd.
Wallerstein, Michael. (1999), “Wage-setting institutions and pay inequality in advanced industrial
Societies”, American Journal of Political Science, vol. 43, pp. 649-680.
Winegarden, C. R. (1979), ‘‘Schooling and Income Distribution: Evidence from International Data,’’
Economica, vol. 46, pp. 83–87.
Wooldridge, J.M. (2002), “Econometric Analysis of Cross-Section and Panel Data”, MIT Press,
40 Appendix
Figure 1: Share of Bot90 in total income
56
60
64
68
72
76
80
70
72
74
76
78
80
82
84
86
88
90
92
94
96
98
Australia
Canada
France
Germany
Italy
Netherlands
New Zealand
Norway
Sweden
Switzerland
41 Figure 2: Share of Top(10-1) in total income
16
18
20
22
24
26
28
30
70
72
74
76
78
80
82
84
86
88
90
92
94
96
98
Australia
Canada
France
Germany
Italy
Netherlands
New Zealand
Norway
Sweden
Switzerland
42 Figure 3: Share of Top1 in total income
