Impact of Income Distribution On Economic Growth, Rojas and Khor

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Does income distribution impact on eco-

nomic growth?
by Jorge Rojas and Ashley Khor

November 1, 2010

Abstract

This study aims to provide some insight on the relationship between income
distribution and economic growth using an overlapping generations model
where agents live for three periods and are endowed with different amounts
of human capital. We distinguish between “rich” and “poor” agents, where the
amount of human capital owned determines that agent’s labour income. We
introduce government spending as a relevant input for production, which is
financed by a lump-sum tax to capital. Through the use of regression anal-
ysis we look for verification of some implications of the model. The conclu-
sion is that the determinant in a smaller economic growth might be the reac-
tion from the governments to alleviate inequality using taxes as a mechanism
for redistribution of wealth, rather than income inequality itself.

Is equality in income really important? Or is equality in terms of access to quality


education and training of skills what really matters? In modern societies the posi-
tive relationship between educational level and income is well established. For
instance, Barro and Lee [2001] found that as of 2000, members of advanced
economies had an average of 9.80 years of school, as compared with only 4.89
years for those from developing countries. A similar pattern occurs within a
country, where on average, the greater the family income, the higher the years of
schooling for those family members.
It is accepted that human capital and education are important driving forces in
the determination of growth: see Lucas [1988], Barro [1991] or Benhabib and
Spiegel [1994]. There is less consensus on the relationship between income distri-
bution and growth. In the 1990s, theoretical and empirical studies emerged
demonstrating that inequality is negatively associated with growth across coun-
tries, including Bertola [1993], Alesina and Rodrik [1994] and Persson and
Tabellini [1994]. On the other hand, others such as Li and Zou [1998] and Forbes
[1997] have reported findings with the opposite sign. Barro [2000] found evidence
of a negative effect of inequality on growth for poor countries but a positive rela-
tionship for rich countries, with only a weak overall effect of inequality on growth
and investment. In contrast, Rehme [2007] found evidence of a U-shaped relation
between education and growth, with increases in education leading to first
increases then decreases in growth as well as income inequality. In other words,
this issue has important policy implications, since different assumptions about the
links between growth and inequality tend to produce different outcomes for the
poor [Deininger and Squire, 1997].

1
Alesina and Rodrik study the effect of inequality in income on the economic
growth through a production function similar to ours. However, the representative
agents maximise utility continuously along their lives and do not differ in their
levels of human capital. Bertola developed a model with a representative agent of
infinite life, focusing on the growth and distributional effects of fiscal policy
through taxes and subsidies. Persson and Tabellini work with an overlapping gen-
erations model with two periods of life and agents who have different endowments
of skills, but they focus on the political equilibrium of the society and conclude
that inequality leads to policies that do not protect property right, and hence the
decrease in economic growth.
We study the same issues, but from a different perspective. We focus more on the
effect of the disparity in human capital and in the relevance of government
spending. We explore the interaction between the human capital owned by the
agents, the labour income that results from this accumulation of skills, and the
economic growth achieved by the society. The main characteristics of our model
are that there are two types of representative agents, “rich” and “poor”, which
differ in their levels of human capital. As in Rehme, all income differences are due
to differences in wages. Additionally, government spending is an input for the pro-
duction of the goods in the economy. For simplicity, we have assumed that the
economy produces only one good and the production function has a Cobb-Dou-
glas form. Moreover, to incorporate some realism in the model we work with a
three overlapping generations model (childhood, working age, and retired age).
Shitovitz [1988] provides some very valuable insight about how to address a three
period OLG model. The microeconomic structure of the model is given by an
isoelastic utility function that the agent maximises during his or her life. Our
model predicts that the higher the inequality, the lower the expected economic
growth.
In section III, we present the results of cross-country regression analysis per-
formed for three different time periods. We find that income inequality had a neg-
ative and significant relationship with growth for 1960-2005, a negative but
insignificant relationship for 1975-2005 and a weak positive relationship for 1990-
2005. In section IV, we observe that these diverse results were not unexpected
given the range of findings of other research in the literature. As a caveat, we
observe that results may differ when other estimation techniques such as panel
estimation are applied. At the same time, the results provide some evidence that
taxation as a means of alleviating poverty by the government was significant
during 1960-1980, but has declined since that period. We note that further
research is required to uncover the relationship between measures to alleviate
poverty, education levels, income inequality and growth.

I. Theory
The production function is given by:
1−α
Yt = Ktα h1t L1t + h2t L2t G1−α
t (1)

2
where Kt is capital, Lit is the population of type i, hi,t is the stock of human cap-
ital owned by agents of type i, and Gt is government spending. We define govern-
ment spending as a fraction of capital, given by the lump-sum tax τ . For sim-
plicity, we assume that population is equal to labour, where the population
growth rate is given by n. There are only two types of agents. A proportion σ of
agents are type 1 (“rich”), and (1 − σ) type 2 (“poor”). These definitions are sum-
marized in the following equations:

Gt = τ Kt
Lt = (1 + n) Lt−1
L1t = σ Lt
L2t = (1 − σ) Lt

The stock of human capital owned by the two type of agents is given by:

h1t = η ht
h2t = (1 − η) ht

where η is the fraction of the total human capital in the economy that belongs to
type 1 with η > 0.5. The total stock of human capital follows the law of motion
given by equation (2)1:

dht
= µ eψu Atγ h1−γ
t , µ > 0 and γ ∈ (0, 1] (2)
dt

where u is the number of years of schooling, At is the world technology frontier at


time t, and ψ is a positive coefficient that takes into account the effect of an extra
year of education in the change of human capital. This parameter has been esti-
mated in around 10%.

We can see from equation (1) that hi,t is just a Harrod-neutral productivity factor
for agent i. The total stock of human capital grows according to equation (2). We
further assume that u is constant and corresponds to the average for the society.

In this economy, markets are competitive, i.e. labour and capital earn their
marginal products, and firms earn zero profits. We calculate the marginal prod-
ucts in the usual way:

∂Yt
rt = (3)
∂Kt

1. Equation (2) is a relationship analyzed by Nelson and Phelps (1996) and Bils and Klenow (2000),
and cited in Jones [2002].

3
∂Yt ∂Yt
Wt,i = i
= hit = hit Wt
∂Lt ∂Lt

To calculate these expressions we need to define explicitely our production func-


tion, hence we solve the equation given in (2) by separation of variables. First,

Ȧt
= φ (4)
At
⇒ At = A0 e φt

where A0 is the initial technology frontier and φ is the growth rate of A. With
this solution we obtain:

 1/γ  
µ ψu+φγt
ht = A0 exp (5)
φ γ

Therefore, our Cobb-Douglas function can be written as:

 1−α  
1−α µ γ (1 − α) (ψ u + φ γ t)
Yt = τ 1−α (η σ + (1 − η) (1 − σ)) A1−α
0 exp Kt L1−α
t
φ γ

 Yt = Bt Kt L1−α
t (6)

where:
 1−α  
1−α µ (1 − α) (ψ u + φ γ t)
1−α
A1−α
γ
Bt = τ (η σ + (1 − η) (1 − σ)) 0 φ
exp γ

These parameters are exogenously determined. Bt could be regarded as a Hick-


neutral productivity factor that depends on the proportion of agents who are type
1, the level of human capital that they own, the amount of taxes, the average
years of schooling, the growth rate of technological progress, the return to each
additional year of education, the share of capital and the constants µ and γ. Fur-
thermore, let us assume that the parameters A0, µ, γ are equal to 1.

 Bt =
 1−α
τ
φ
[σ (2η − 1) + 1 − η]
1−α
exp[(1 − α) (ψ u + φ t)] (7)

Notice that if η = 1/2, then the proportion of the population plays no role. This is
expected because if η = 1/2, then according to our definition there is only one
type of agent . In addition, if σ = 1/2, then the expression σ (2η − 1) + 1 − η is
always equal to 1/2 irrespective of the value of η.

4
Figure 1. Productivity factor in a society with perfect equality. The dot line repre-
sents higher taxes, while the continuous line represents lower taxes.

From figure (1), we see that the higher the taxes, the higher the productivity
factor. This makes sense since government spending is productive. Nevertheless,
notice that the income per effective worker might decrease since rich workers pay
more taxes and spend less on consumption.

Figure 2. Productivity factor in a society with inequality. The dot line represents
higher inequality, while the continuous line represents lower inequality.

From figure (2), we see that more inequality leads to a decrease in productivity.

5
Next, we explore the law of motion for the capital accumulation. This is given by:

Kt+1 = (1 − δ) Kt + It (8)

where δ is the depreciation rate and It is the total investment in the society. For
our closed economy, investment is equal to savings:

s1t L1t + s2t L2t


 
I t = St = s t Y t = Yt
Lt

 It = s1t σ + s2t (1 − σ) Yt
 
(9)

where st is a weighted average of the optimal savings rate of the two types of
agents from the household’s problem. This allows to write the law of motion as:

Kt+1 = (1 − δ) Kt + st Yt

 Kt+1 = (1 − δ) Kt + st Bt Kt L1−α
t (10)

dividing equation (10) by Bt Lt we obtain:

kt+1 = (1 + g) (1 + n) (1 − δ) kt + s1t σ + s2t (1 − σ) Kt Lt−α


 
(11)

kt represents capital per unit of effective labour, g is the growth rate of Bt. The
next step is to calculate the saving rates s1t and s2t to obtain the expression that
will provide the steady-state.
The marginal productivities are given by:

rt = Bt L1−α
t (12)
Wt = (1 − α) Bt Kt Lt−α

The productivity of labour for agent of type i is given by Wt,i = Wt hit


The agents in this model are assumed to be altruistic towards their children.
Therefore, their utility is related not only to their own consumption, but also to
the utility of future generations. Thus, the utility function may be written as:

1 1 1
Vti = U i(C1t,i) + U i(C2t+1,i) + 2 U i(C3t+2,i) + i
Vt+1 (13)
1+ ρ (1 + ρ) (1 + R)2

where ρ > − 1 is the discount rate for utility due to consumption. The greater ρ,
the less the household values future consumption relative to current consumption.
Cjt,i is the consumption at period j of life of an agent type i at time t. R > 0 is
the rate at which parents discount the utility of their children. For simplicity, it is
assumed that agents have children in the last period of their lives. However, this
assumption could be relaxed to consider situations in which agents have children
in the second period or the first period and subsequently they might “exploit”
their children.

6
By recursive iteration we obtain:
∞  2j " #
X 1 1 1
Vti = U i(C1t+j ,i) + U i(C2t+1+j ,i) + U i(C3t+2+j ,i) (14)
j=0
1+R 1+ ρ (1 + ρ)2

We assume agents have isoelastic utility i.e. constant-relative-risk aversion utility:


1−θ
Ct,i
Ui = (15)
1−θ
where θ > 0 is the relative risk aversion coefficient. Notice that 1/θ2 is the elas-
ticity of substitution between consumption at any two points in time.
The instantaneous budget constraints of an agent type i born at time t are given
by:

C1t,i + S1t,i = Bt,i (16)


C2t+1,i + S2t+1,i = Wt+1,i + (1 + rt+1) S1t,i
C3t+2,i + Bt+2,i = (1 + rt+2) S2t+1,i

C is consumption, S is savings, B is bequest, W is labour income and rt is the


interest rate in period t. Rearranging the equations we obtain the intertemporal
budget constraint of an individual born in period t as follows:
C2t+1,i C3t+2,i Wt+1,i Bt+2,i
C1t,i + + = Bt + − (17)
(1 + rt+1) (1 + rt+1) (1 + rt+2) (1 + rt+1) (1 + rt+1) (1 + rt+2)

We assume that the bequest is non-negative for all periods, i.e. Bt > 0 ∀t.
To obtain the optimal saving rates and the steady-state is necessary to solve the
maximization problem of the household:

Max Vti
subject to The budget constraint
− Bt 6 0

To solve the problem, we deploy the Lagrange as:


∞  2j " #
1 1 1
U (C3t+2+ j ,i) +

X
i i i
L= U (C1t+ j ,i) + U (C2t+1+j ,i) + (18)
j =0
1+R 1+ρ (1 + ρ)2
 
Wt+1,i Bt+2,i C2t+1,i C3t+2,i
λ Bt + − − C1t,i − − − µ Bt
(1 + rt+1) (1 + rt+1) (1 + rt+2) (1 + rt+1) (1 + rt+1) (1 + rt+2)

From the Lagrange we can obtain 5 first-order conditions.


∂L ∂L
=0 =0
∂C1t ∂C2t+1
∂L ∂L
=0 =0
∂C3t+2 ∂λ
∂L
=0
∂µ

7
For simplicity, we assume that labour is supplied inelastically. Thus, Lt = 1 ∀t.
This implies the following equations:

kt+1 = (1 + g) (1 − δ) kt + s1t σ + s2t (1 − σ) kt Bt


 
(19)
rt = Bt (20)
wt = (1 − α) rt kt (21)

We know that sit is a function of rt, and in turns rt is a function of kt. Solving
explicitely for the steady-state is not done because the math is very involved.
However, we can still draw some general conclusions from the above model.
Firstly, this three-period OLG model may have no Pareto-optimal stationary
competitive equilibrium (among them, it would be the trivial equilibrium at zero).
There may be an inefficient steady state, with increasing accumulation of capital.
Secondly, the inefficiency in the outcome is due to the fact that future genera-
tions’ preferences do not affect current decisions when the bequest is equal to
zero. Thirdly, there are missing markets because different generations cannot
trade. Finally, our model predicts that higher inequality would lead to lower eco-
nomic growth and with inefficient outcomes.

II. Empirical evidence


This section provides a preliminary analysis of the link between growth and
income inequality. As the baseline regression, we re-examine the Alesina-Rodrik
regression using an updated high quality dataset. Our approach considers the
cross-sectional regression approach of Alesina and Rodrik and Persson and
Tabellini.
Data
Gini coefficients are commonly used as measures of inequality since comparable
data on wealth distribution for a large enough sample of countries do not exist.
We use the high-quality dataset of Gini coefficients by Deininger and Squire
[1996), which minimises measurement error in inequality and any resulting coeffi-
cient bias, as well as increases the efficiency of estimates. Following Barro [2000],
Forbes and Li and Zou [1998], we have not included the Gini coefficient for land.
Appendix 1 lists countries and Gini coefficients. Primary school enrolment ratios
are from the 2010 Barro-Lee dataset. Real per capita GDP and the resultant
growth rates are taken from the Penn World Tables. We adopt the approach
taken by Li and Zou and classify a country as democratic if its civil liberty index
is less than or equal to two. We have used the civil liberty index published by
Freedom House.
We have adopted the same selection of countries used by Forbes. While the data
set reflects 45 countries for which sufficient data was available, nevertheless lim-
ited number of observations are available for many countries and earlier time
periods. As such, data quality is still a significant issue in this empirical estima-
tion.

8
Table 1. Base Regression Results 1960-2005: Cross-Sectional Data

Baseline regression

For our two period regressions, we consider regressions for there different time
periods:

Gi,1960−2005 = β0 + β1Ginii,65 + β2GDPi,65 + β3PRi,65 + ui

Gi,1975−2005 = β0 + β1Ginii,80 + β2GDPi,80 + β3PRi,80 + ui

Gi,1990−2005 = β0 + β1Ginii,95 + β2GDPi,95 + β3PRi,95 + ui

Gi,1960−2005 is the average real GDP growth rate between 1960 and 2005. Ginii,65
is the Gini coefficient, GDPi,65 is the per capital real GDP, and PRi,65 is the pri-
mary school enrollment ratio, presented as average initial values over 5 years from
1961 to 1965. Similarly, Ginii,80, GDPi,80 and PRi,80 are average initial values
from 1976 to 1980. If there is no available Gini for that period, we use the data
available for the next most recent period.

Additionally, for each period, we consider several variations. D is included as a


democracy dummy variable. APR, the average years of primary schooling, is
included as an alternative to primary school enrolment ratio. The results are pre-
sented in Tables 1, 2 and 3 for the three time periods.

9
Table 2. Base Regression Results 1975-2005: Cross-Sectional Data

Table 3. Base Regression Results 1990-2005: Cross-Sectional Data

The results differ for our different time periods. For 1960-2005, the results are
similar to those in Alesina and Rodrik, where the income distribution GINI as a
negative and significant coefficient. On the other hand, for 1975-2005, we find a
negative but insignificant coefficient for the GINI. For 1990-2005, we find a posi-
tive coefficient for the GINI. However, while the first two regressions pass the

10
diagnostics at the 5% level of significance (except for 1960-2005 Reg (1) which
finds no evidence of heteroskedasticity at 1% level of significance and 1975-2005
Regs (3) and (4) which satisfy normality at 1% level of significance), there is
strong evidence that normality is violated for all of the 1990-2005 regressions.
Thus, we cannot rely on hypothesis testing for this period and we focus our anal-
ysis on the regressions for 1960-2005 and 1975-2005. Lagged GDP has a negative
coefficient in all the regressions for the two periods. This coefficient is generally
significant, except when the democracy dummy variable is included for 1975-2000.
Primary school enrolment ratios, average years of school and the democracy
dummy were not significant in any of the models. This accords with Alesina and
Rodrik?s finding that the relationship between inequality and growth is not dif-
ferent in democracies and nondemocracies but differs from Persson and Tabellini’s
result of an inverse relationship for democracies but not nondemocracies.

III. Discussion
We note that our results may have been different had we conducted other estima-
tion methods, such as sensitivity analysis and panel estimation. Forbes, and
Deininger and Squire [1998], observe that when sensitivity analysis is performed,
the negative coefficient on inequality in cross-country regressions often becomes
insignificant. Panel estimation as compared to cross-country regression has the
advantage of revealing dynamic interaction between distribution and growth [Li
and Zou, 1998] and can also specifically estimate how changes in a country’s level
of income predicts a change in that country’s growth rate [Forbes, 2000]. In addi-
tion, panel estimation often produced positive associations in the empirical litera-
ture: Barro [2000], Forbes and Li and Zou. Notably, while Li and Zou found a
positive relationship using panel estimation, they were able to produce negative
links with cross-country regressions using the same data set.
Rehme [2007] contended that while most empirical work focus on linear relation-
ships (as has ours), the growth-inequality relationship may in fact be nonlinear.
He argued that both growth and measured inequality can be represented as
inverted U-shaped functions of the percentage of high-skilled people in the popu-
lation (which represented the stage in the development process). This can explain
why the growth-income inequality relationship is very difficult to predict, since
the sign of the association depends on where each particular economy lies, in
terms of its percentage of high-skilled people, relative to the turning points of its
growth and measured inequality functions.

IV. Final remarks


Rather than attempting to create a new theoretical model, we have conducted a
critique of the neoclassical models in the literature. While our model has sought
to include elements that better represent reality, it has highlighted the complica-
tions arising from a more realistic model. In order to work towards a solution, we
had to adopt some simplifications. Although the closed-form solution is not
explicitly stated, we achieved some general implications, and find that higher
inequality would lead to lower economic growth and inefficient outcomes.

11
There are several caveats regarding the findings in our report. As Forbes [2000]
observes, data quality, period length and estimation technique all influence the
sign and significance of the coefficient on inequality. For the purposes of this
report, we have worked with datasets collected by other researchers rather than
primary data. Additionally, we highlight that correlation does not imply causa-
tion, and more research is needed to determine important issues such as whether
economic growth in developing countries results in a more unequal distribution of
income, and whether countries with unequal income distributions experience
slower economic growth [Deininger and Squire, 1997], as well as to expound on
the links between human capital, inequality and growth.
We propose that the determinant that slows down economic growth is not the
income distribution itself, but the reaction from the governments to inequality.
One argument is that the mechanism deployed by government to alleviate
inequality used to be taxes, which resulted in a strong negative relationship
between inequality and growth during the 1960s to 1980s. On the other hand, as
governments shifted to alternative measures to combat poverty since the 1980s,
this negative relation may have weakened over time2. At the same time, we find
evidence that education levels are likely to have a strong role in determining the
relationship between income inequality and growth. Future research lies in
exploring the dynamic interaction of income inequality and growth while taking
into account education levels and the stage of development, as well as a more in-
depth analysis of the measures used by governments to alleviate inequality, as
well as improved empirical estimation using panel data.

2. This information can be found in different sources such as IMF, WB and WTO. Although, the trend
is not true for all the countries involved in our study.

12
References
[1] Alesina, Alberto, and Dani Rodrik, “Distributive Politics and Economic
Growth,” Quarterly Journal of Economics, CIX (May 1994), 465-490.
[2] Barro, Robert, “Economic Growth in a Cross section of Countries”, Quarterly
Journal of Economics, CVI (1991), 407-444.
[3] Barro, Robert, “Inquality and Growth in a Panel of Countries”, Journal of Eco-
nomic Growth, VIII (2000), 5-32.
[4] Barro, Robert, and Jong-Wha Lee, “International data on educational attain-
ment: updates and implications”, Oxford Economic Papers, III (2001), 541-563.
[5] Barro, Robert and Jong-Wha Lee, “A New Data Set of Educational Attain-
ment in the World, 1950-2010”. NBER Working Paper No. 15902, 2010.
[6] Barro, Robert, and Xavier Sala i Martin, “Public Finance in Models of Eco-
nomic Growth”, NBER Working Paper No. 3362, 1990.
[7] Benhabib, Jess and Mark Spiegel, “The Role of Human Capital in Economic
Development: Evidence from Aggregate Cross-Country Data”, Journal of Mone-
tary Economics, XXXIV (1994), 143-173.
[8] Bertola, Giuseppe, “Shares and Savings in Endogenous Growth,” American
Economic Review, LXXXIII (Dec 1993), 1184-1198.
[9] Chenery, H. and T.N. Srinivasan, Handbook of Development Economics,
Volume II, Chapter 19 (Elsevier Science Publishers B.V., 1989).
[10] Deininger, Klaus and Lyn Squire, “A New Data Set Measuring Income
Inequality,” The World Bank Economic Review , X (1996), 565-91.
[11] Deininger, Klaus and Lyn Squire, “Economic Growth and Income Inequality:
Reexamining the Links”, Finance and Development (Mar 1997), 38-41.
[12] Forbes, Kristin J., “A Reassessment of the Relationship Between Inequality
and Growth”, American Economic Review , XC (Sep 2000), 869-887.
[13] Heston, Alan, Robert Summers and Bettina Aten, Penn World Table Version
6.3, Center for International Comparisons of Production, Income and Prices at
the University of Pennsylvania, August 2009.
[14] Jones, Charles I., Introduction to Economic Growth, Chapter 6, 2nd edition
(New York, NY: W.W. Norton & Company, Inc., 2002).
[15] Li, Hongyi and Heng-fu Zou, “Income Inequality is not Harmful for Growth:
Theory and Evidence”, Review of Development Economics, II (1998), 318-334.
[16] Lucas, Robert E., “On the Mechanics of Economic Development”, Journal of
Monetary Economics, XXII (1988), 3-42.

13
[17] Neusser, Klaus, “Savings, Social Security, and Bequests in an OLG Model. A
simulation Exercise for Austria”, Journal of Economics, VII (1993), 133-155.
[18] Persson, Torsten, and Guido Tabellini, “Is Inequality Harmful for Growth”,
American Economic Review, XCIV (Jun 1994), 600-621.
[19] Rehme, Gunther, “Education, Economic Growth and Measured Income
Inequality”, Economica, LXXIV (2007), 493-514
[20] Salvadori, Neri, Economic Growth and Distribution: On the Nature and
Causes of the Wealth of Nations, Chapter 4 (Cheltenham, UK: Edward Elgar
Publishing, Inc., 2006).
[21] Serrano, Carlos, “Social Security Reform Accumulation, Income Distribution,
Fiscal Policy, and Capital Accumulation”, The World Bank, Policy Research
Working Paper No. 2055, 1999.
[22] Shitovitz, Benyamin, “On Stationary 3-Period Overlapping Generations
Models,” Journal of Economic Theory, XLVI (1988), 402-408.
[23] Thibault, Emmanuel, “Existence of equilibrium in an OLG model with pro-
duction and altruistic preferences,” Economic Theory, XV (2000), 709-715.
[24] Von Weizsäcker, Robert K., A theory of earnings distribution (Cambridge,
CB: Cambridge University Press, 1993).

14
Appendix 1: Gini Coefficients
1991- 1996- 2001- 2006-
1961-1965 1966-1970 1971-1975 1976-1980 1981-1985 1986-1990
1995 2000 2005 2010
Australia - - - 39.3 37.6 41.7 35.2 - - -
Bangladesh 37.3 34.2 36.0 35.2 36.0 35.5 26.2 30.7 31.0 -
Belgium - - - 28.3 26.2 26.6 - 33.0 - -
Brazil - 57.6 61.9 57.8 61.8 59.6 59.2 58.6 56.4 55.0
Bulgaria - - - - 23.4 24.5 31.1 26.4 29.2 -
Canada 31.6 32.3 31.6 31.0 32.8 27.6 - 32.6 - -
Chile - 45.6 46.0 53.2 - - 55.2 55.4 54.9 52.0
China - - - 32.0 31.4 34.6 - - 41.5 -
Colombia - 52.0 46.0 54.5 - - 57.2 57.5 58.8 58.5
Costa Rica - - 44.4 45.0 47.0 46.1 46.3 46.6 47.2 48.9
Denmark - - - 31.0 31.0 33.2 - 24.7 - -
Dominican Rep. - - - 45.0 43.3 50.5 51.4 52.1 50.0 48.4
Finland - 31.8 27.0 30.9 30.8 26.2 - 26.9 - -
France 47.0 44.0 43.0 34.9 34.9 - 32.7 - - -
Germany 28.1 33.6 30.6 32.1 32.2 - - 28.3 - -
Greece - - - - 39.9 41.8 - 34.3 - -
Hong Kong - - 39.8 37.3 45.2 42.0 - 43.4 - -
Hungary - - - 21.5 21.0 23.3 27.9 27.3 30.0 -
India 37.7 37.0 35.8 38.7 38.1 36.3 - - 36.8 -
Indonesia - - - 42.2 39.0 39.7 - - 39.4 37.6
Ireland - - 38.7 35.7 - - - 34.3 - -
Italy - - 39.0 34.3 33.2 32.7 - 36.0 - -
Japan 34.8 35.5 34.4 33.4 35.9 35.0 24.9 - - -
Korea, Rep. 34.3 33.3 36.0 38.6 34.5 33.6 - 31.6 - -
Malaysia - 50.0 51.8 51.0 48.0 48.4 48.5 49.2 37.9 37.9
Mexico 55.5 57.7 57.9 50.0 50.6 55.0 51.9 51.9 46.1 51.6
Netherlands - - 28.6 28.1 29.1 29.6 - 30.9 - -
New Zealand - - 30.0 34.8 35.8 40.2 - 36.2 - -
Norway 37.5 36.0 37.5 31.2 31.4 33.1 - 25.8 - -
Pakistan - 36.5 38.1 38.9 39.0 38.0 30.3 33.0 31.2 31.2
Peru - - - - 49.3 49.4 44.9 46.2 52.0 50.5
Philippines - - - - 46.1 45.7 42.9 46.1 44.5 44.0
Poland - - - - 25.3 26.2 32.4 32.9 34.9 -
Portugal - - 40.6 36.8 - - - 38.5 - -
Singapore - - 41.0 40.7 42.0 39.0 - 42.5 - -
Spain - - 37.1 33.4 31.8 32.5 - 34.7 - -
Sri Lanka 47.0 37.7 35.3 42.0 45.3 36.7 32.5 35.4 41.1 -
Sweden - 33.4 27.3 32.4 31.2 32.5 - 25.0 - -
Thailand 41.3 42.6 41.7 - - - 46.2 43.2 42.5 -

Trinidad and Tobago 40.3 - - -


- - 51.0 46.1 41.7 -
Tunisia - - 50.6 49.6 49.6 46.8 41.7 40.8 - -
Turkey - 56.0 51.0 - - - 41.5 - 43.2 41.2
United Kingdom 24.3 25.1 23.3 24.9 27.1 32.3 - 36.0 - -
United States 34.6 34.1 34.4 35.2 37.3 37.8 - 40.8 - -
Venezuela - - 47.7 39.4 42.8 53.8 46.8 49.5 47.6 43.4

Average 37.8 40.3 39.9 38.1 37.4 38.0 41.2 38.3 42.7 46.2

Note: Gini coefficient is taken from the latest available date within the given period. We rely on the 1961-1990
adjusted Gini coefficients provided by Forbes [2000]. These are supplemented by recent figures provided by the
World Bank (calculated using Povcal).

You might also like