Why is Inequality Back on the Agenda?

By
Ravi Kanbur and Nora Lustig1
Version: April 21, 1999
(Draft: not to be quoted without permission)

1. Introduction
2. Efficiency and Equity
3. Growth and Distribution
4. Recent Changes in Inequality
5. The Complex Patterns of Inequality Change
6. Inequality Between Countries
7. Conclusion

ABSTRACT

This paper investigates the reasons why inequality, and distribution more generally, have come to the fore in
the development discourse at the turn of the century, after a period of relative neglect in the 1980s. The paper
considers, in particular the analysis of (a) efficiency and equity, (b) growth and distribution, (c) recent changes in
inequality, (d) recent work on the complex patterns of inequality change in developing countries, and (e) inequality
between countries. All of these different strands of analyses have ensured that inequality will be prominent on the
development agenda in the decade to come.

JEL Clasification: I3 Welfare and Poverty; O1: Economic Development; O4: Economic Growth; O57: Comparative
Studies of Countries;

1
Paper prepared for the Annual Bank Conference on Development Economics, World Bank, Washington, D. C., April 28-30, 1999.
Ravi Kanbur is T.H Lee Professor of World Affairs and Professor of Economics at Cornell University, and Nora Lustig is Chief of the
Poverty and Inequality Advisory Unit at the Inter-American Development Bank and a Nonresident Senior Fellow of the Brookings
Institution. The authors are very grateful to Branco Milanovic for his helpful insights and to the two anonymous reviewers of the
outline for their comments. They also wish to thank Jose Antonio Mejia, Luis Tejerina, and Heather McPhail for their valuable
assistance. The views expressed here are those of the authors and do not necessarily coincide with those of the institutions they are
affiliated with.
 1

1. Introduction

At the turn of the century discussion of inequality, and distribution more generally, has
come to the fore in the development discourse, after a period of relative neglect in the 1980s. This
is true not just of the development discourse, but the discourse on economic policy more
generally. Atkinson (1997) writes of “Bringing Income Distribution in from the Cold.” A
Handbook on Income Distribution, the first ever, will come out this year. In their introduction to
this volume, Atkinson and Bourguignon (1998) write: “There was a time in the postwar period
when interest in the distribution of income had almost vanished….Today, the position is
different.”

The object of this paper is to trace some of the contours of this resurgence of interest in
distribution in the development discourse. In Section 2 we argue that the profession wide debate
on the separation of efficiency and equity has had its impact on the development literature, where
the analysis now focuses on whether and under what conditions they are substitutes or
complements. Section 3 turns to the role of the Kuznets curve in the rise and fall of interest in
distribution. Section 4 looks at changes in inequality between the 1980s and the 1990s to see if
the resurgence of interest in inequality can be found in these facts. Section 5 introduces a
literature and a type of analysis which is going to become common in the development literature
as more household surveys become available—analysis which highlights the complex micro-
patterns of distributional change in developing countries, even when overall indices of inequality
do not change very much. Section 6 considers the question of inequality between countries.
Section 7 concludes the paper by noting that the trends we have identified look set to keep
inequality at the forefront of the development agenda in the years to come.

2. Efficiency and Equity

One strand of literature which speaks to the question of why inequality is now back on the
agenda, is welfare economics. The fundamental theorems of welfare economics give precise form
to Adam Smiths’s invisible hand. The first theorem says that any equilibrium in competitive
markets is Pareto efficient; the second that, under certain conditions, any Pareto efficient
allocation can be attained as a competitive equilibrium. There is thus a tight connection between
market equilibrium allocations and efficiency. But what about distribution? One of the conditions
 2

for the second theorem is the availability of lump sum transfers through which endowments can
be redistributed. Since any Pareto efficient outcome can be described as an equilibrium of
perfectly competitive markets, these outcomes can be egalitarian or non-egalitarian ones. The key
point is that, within this framework, efficiency and equity are separable. The policy counterpart of
this intellectual framework was present, for example, in arguments to redistribute via assets while
keeping markets as free and as competitive as possible. A lucid account of this debate is to be
found in Meade (1964).

The framework developed by Arrow and Debreu and their followers, which describes the
precise conditions under which the invisible hand did its work, was refined and sharpened in the
first decades after the war. But such precision only served to highlight the severity of the
conditions required—no increasing returns, no monopolies, a complete set of markets for present
and future goods, complete insurance markets, fully available and symmetric information, lump
sum transfer instruments, etc. In the 1970s and the 1980s, the reaction set in and a number of
economists began relaxing these assumptions. For example, Joseph Stiglitz, George Akerlof,
Michael Spence and others took the implications of imperfect information head on. This line of
research questioned the efficiency of market equilibrium. Greenwald and Stiglitz ( 1986 ), for
example, showed that in the presence of imperfect information the competitive equilibrium was
not even constrained Pareto efficient.

Much of the literature in this vein is devoted to showing features of equilibria under
conditions which approximate reality better than the Arrow-Debreu model, or demonstrating
inefficiency of market equilibira. However, there is one strand which is particularly relevant for
us. This focuses on the separation of efficiency and equity and shows that, once the Arrow-
Debreu conditions are relaxed, there is no such separation. With imperfect information, lump sum
redistribution of endowments have the capacity to improve efficiency (in the sense of making at
least one person better off without making anyone else worse off) under certain conditions, or
worsen it under others (Hoff, 1994). In the absence of lump sum instruments, market
interventions may reduce efficiency, but improve equity. In any event, whether they are
substitutes or complements, efficiency and equity have to be taken together—they are not
separable.

This non-separability has been analyzed in detail in a growing number of contributions

and we note some strands in the literature. One strand goes back to the Noble prize winning paper
 3

of Mirrlees (1971), where lack of lump sum transfers instruments makes equity and efficiency
substitutes. Mirrlees (1971) models the classic conflict between progressive income taxation and
the incentives to provide effort. He shows that the resolution depends on a combination of the
degree of egalitarianism and the characteristics of labor supply. The large literature which
followed Mirrlees’s contribution (see the appreciation in Dixit and Besley 1997) helped us
understand better the nature of this tradeoff.

Yet another strand in the literature emphasizes the complementarity between equity and
efficiency. Boadway and Keen (1999) survey the literature and provide an excellent account of
the underlying economics in the following illustration from capital markets:

“Lenders’ informationally-constrained responses to the possibilities that borrowers will

default on their loans or simply abscond with the funds, are likely to lead to levels of investment
that depart from the first best. Differing models give different predictions as to whether there will
be under- or over- investment: lenders may ration investments below the first best, for example,
or borrowers may borrow excessively to take advantage of being able to default if things go
badly. In either event, however, a redistribution of wealth towards borrowers is likely to move
investment closer to the first best, so increasing aggregate income: incentive problems will be
mitigated in so far as greater wealth enables (or obliges) investors to rely more heavily on their
own funds rather than borrowed funds tainted with the lender’s mistrust.”

Bardhan, Bowles and Gintis (1999) also review and elaborate on a large and growing
literature in this vein, looking explicitly at other issues such as common property resource
problems, and socially optimal levels of risk taking. The find complementarities as well as
tradeoffs, but end with the basic conclusion that distribution and efficiency are inextricable.

Once the conditions giving rise to the separability of efficiency and equity are abandoned,
much depends on the detail of the situation at hand, and very different results can be obtained
with different specifications. To illustrate this, consider a famous debate of the 1980s, that on
“Adjustment and Equity”, and let us focus on Africa. It will be recalled that the macro shocks of
the late 1970s revealed major weaknesses in African economies, particularly on the trade front.
Dramatically overvalued exchange rates and a plethora of trade controls, it was argued, were
highly inefficient distortions which needed to be removed during the process of stabilization and
 4

adjustment. But what about the ditstributional consequences of this adjustment (setting to one
side the stabilization phase)?

It was argued by some that these consequences would be benign. One such argument is to
be found in Kanbur (1987). The analysis proceeds from a stylized characterization of the
adjustment as increasing the returns to activities producing internationally tradable goods and
services relative to activities producing non-tradable goods and services. The issue then became
who (in the short run) derived their income primarily from the tradable sector and which factor
(in the long run) was used more intensively in this sector. Empirical evidence (Kanbur, 1990) was
used to argue that on both these counts the outcomes were likely to be more equitable than the
status quo—essentially, households engaged in the tradeable sector (for example, small holder,
export cash crop farmers in West Africa) were poorer on average than those in the more
urbanized non-tradeable sector.

All of this was analysis of distribution at the household level—mainstream economic

analysis, and empirical data collection, rarely went into intrahousehold issues in the mid 1980s.
However, during the last fifteen years there has been considerable work on gender division within
the household (for a review, see Alderman et.al, 1995). Following on from the pioneering work of
Sen (1983), intrahousehold inequality is treated much more systematically now in the mainstream
development economics literature. For example, Haddad and Kanbur (1990) showed how taking
this into account can dramatically alter the measurement of inequality and poverty. Moreover, an
intrahousehold and gender perspective also may change the analysis of adjustment and equity.
For example, if there is gender segmentation in production, if there is not complete pooling of
household resources, and if women derive their income primarily from non-tradeable goods and
services, then the sort of adjustment analyzed above may well worsen distribution by negatively
affecting women compared to men (Kanbur and Haddad, 1994) and this may well explain why
some of the most vocal complaints against “adjustment” have come from Women’s groups in
developing countries.

To summarize, our argument is that one of the reasons inequality is back on the agenda is
that the last two decades have culminated in a triumph of the imperfect information and imperfect
markets perspective in mainstream economics. As noted above, one consequence of this is that
distribution becomes an integral part of the overall analysis of economic performance, not an add-
on to be considered once efficiency has been established. But what it also means is that a detailed,
 5

case by case, analysis is needed to determine whether distribution and allocative efficiency are
substitutes or complements. Either way, it puts inequality back on the agenda of the development
discourse.

3. Growth and Distribution

A second strand of the literature we would like to refer to, one which is more directly
related to the development economics literature, is that on growth and distribution. This literature
has gone through interesting cycles in the post war period (see Kanbur, 1998). Immediately
before and after the second world war, the concern was with demand management and then with
overall growth. Thus, for example, Rosenstein-Rodan’s (1943) analysis of development problems
focuses entirely on industrialization and growth, not on the distribution of this growth. The
development and growth models of that time (Mahalanobis, 1963) mostly have this characteristic.
However, by the mid 1950s a change was already taking place, as evidenced in the classic
contributions of Lewis (1954) and Kuznets (1955). As the opening sections of Lewis’s Nobel
prize winning paper make clear, he self-consciously saw himself as reviving a “classical
tradition” which he lamented had fallen into disuse, where growth and distribution were
organically connected—the evolution of one was intimately tied to the evolution of the other (as
Ricardo noted—“To determine the laws which regulate this distribution is the principal problem
in political economy”). The distribution Lewis emphasized, true to his classical inspiration, was
that between capital and labor, and he showed how this distribution affected savings,
accumulation of capital and hence growth.

Kuznets’s (1955) celebrated paper was concerned with the personal distribution of
income and launched the famous inverted U hypothesis—as development proceeds and per capita
income increases, inequality first increases and then decreases. The simple model of intersectoral
population shifts can be elaborated upon (see Anand and Kanbur, 1993a), but the basic point is
that growth and equity were no longer independent in this model—they were substitutes to start
with, and then complements once the turning point of the inverted-U was reached. In a widely
cited paper, Ahluwalia (1976) claimed moreover to have empirical support for the inverted-U
using data on a cross section of developing and developed countries. It is interesting to see the
widely differing policy conclusions that could be drawn from this framework. To some, the
relationship meant that there was sharp conflict between growth and equity (Fishlow, 1972),
something to be managed. To others, the eventual turning point counseled fast growth—to get
 6

over the hump and into the decreasing inequality range as quickly as possible. But notice that for
both positions, growth and distribution were inseparably joined. The need to manage the
distributional consequences of growth is aptly illustrated in the title of the book which many
would argue to be the culmination of this strand of thinking, “Redistribution with Growth,”
(Chenery et.al, 1974).

And yet, in the 1980s the pendulum swung again, and growth and distribution were
separated from each other. But the literature in this phase was different in kind to that of the
immediate post war period. One of the main reasons was that empirical support for the Kuznets
relationship proved to be weak. Using the very same data set on which Ahluwalia (1976) was
based, Anand and Kanbur (1993b) argued that there was in fact no empirical relationship to be
seen once a clean data set and appropriate econometric techniques were applied. This finding has
now been confirmed by researchers on a greatly expanded data set compiled by Deininger and
Squire (1996). In fact, Li, Squire and Zou (1998) argue that inequality does not change very much
at all in countries over time (and a fortiori, since per capita incomes do change considerably, there
is no systematic relationship between inequality and per capita income for countries over time).

It is interesting to focus for a moment on what policy implications might be drawn from
the new conventional wisdom that inequality does not change very much as per capita income
increases. One example of such a policy inference is in World Bank (1998):

“One study examined trends in the distribution of income for 45 countries for which high
quality household income or expenditure data are available over time (Li, Squire and Zou, 1998).
In 29 countries there has been no trend in either direction. The remaining 16 countries are
divided: 8 have shown rising inequality, 8 declining. If income distribution does not change much
over time, the gains in per capita income will affect different segments of society to about the
same degree. Thus, in countries with rapid growth, incomes of the poorest will rise rapidly and
the incidence of poverty will decline. But in countries with no per capita growth and a stable
income distribution, there will be no poverty reduction.”

It is worth examining this logic closely since it is to be found in many World Bank and
other writings. What it seems to be saying is that going for can be pursued without fear of
increased inequality because, when one looks at countries over time, inequality does not change
very much. But the latter is a statement about a reduced form relationship between per capita
 7

income and income distribution. It cannot give policy guidance since the instruments which
influence growth may also influence distribution. Of course there is a combination of policy
instruments that could replicate the “growth but no distribution change” scenario seen in the
outcomes for various countries. But what is that combination? To this question, such analysis
cannot give an answer. In fact, it can be argued that such analysis is dangerous because it leads to
an easy leap to a stylized assumption that, since growth is distribution neutral in the reduced
form, the same is true in the structural form. In other words, those policy instruments which lead
to growth are also distribution neutral. Recent work by Lundberg and Squire (1999) shows how
individual policy instruments can be highly distributionally non-neutral, even though some
combination of them could of course lead to distribution neutral growth.2

There is another type of analysis which risks falling into a similar trap, and this is the
“decomposition” of poverty change into a “growth” component and a redistribution component.
The analytics of this are laid out in Kakwani (1994), and it has been widely applied (eg see Datt
and Ravallion, 1992). The exercise is conducted as follows. We have two income distributions,
say for periods t and t+1. These will differ, in general, in their means and in their inequality. They
will also, in general, have different values of any given poverty index (like the head count ratio),
P(t) and P(t+1). The exercise then constructs a synthetic intermediate distribution * which has the
mean of the t+1 distribution but the spread of the t distribution (there are different technical ways
of doing this, and there are issues of “path dependence” but the basic idea should be clear). This
synthetic distribution * also has a poverty number P(*) associated with it. Then, the change from
P(t) to P(*) is referred to as the “growth component” of the overall poverty change between the
two periods, and the change from P(*) to P(t+1) as the “redistribution component”. Typically, the
first component accounts for as much as 80% of the overall change.

Now, the above calculations are perfectly acceptable as an accounting exercise—they

represent an interesting way of organizing and presenting the evolution of income distribution.
But their cutting power, and their danger, in policy analysis comes from a leap to an implicit
separability between the two components—that, somehow, the growth component could be
acomplished independently of the redistribution component. This, combined with the empirical
finding that the growth component is much larger than the redistribution component, can and
does lead into an easy slip into a classification of policy instruments into “growth” instruments

2
See also the recent work of Gallup, Radelet and Warner (1999) for a more structural approach to
distributional change.
 8

(lower tariffs, higher FDI, privatizing SOEs) and “redistribution” instruments (food subsidies,
labor based public works, progressive taxation). And yet, it is now clear that there is no
justification in economic theory for such a separation. Moreover, it should be clear that many if
not most individual policy instruments have both growth and distributional consequences (see
Lundberg and Squire, 1999, for evidence on this front). Of course, there are combinations of
instruments which can produce distribution neutral growth, but to find out this combination
(which will vary from case to case) requires, among other things, a distribution sensitive analysis
of each of the instruments.

Finally, a large and growing literature in the last decade and a half has explored the
connections between growth and distribution through an investigation of a possible causal link
between initial inequality and subsequent growth. This literature has brought together recently
developed models of endogenous growth, with an earlier generation of models of credit and other
market failures, and also incorporated insights of the new political economy. This is a large
literature (see Kanbur, 1998, Benabou, 1996), but some flavor of it can be found in Alesina and
Rodrik (1994), for example. Here output depends on capital, labor and a public good. The public
good is financed through a proportional tax and, given the endogenous growth specification,
steady state growth depends on this tax rate. Individuals differ in their endowments and hence
have differing interests—the common tax rate delivers a common benefit through expenditure on
the public good, but its costs are paid for by those with higher endowments. These differing
interests will be resolved through the political economy institutions. With majority voting the
median voter theorem can be invoked under certain conditions—as the ratio of median to mean
wealth rises, ie as the measure of inequality decreases, the voted for tax rate will be lower and
hence growth rate will be higher. It has to be said that the jury is still out on whether empirically
there is a positive or negative causal link from inequality to growth (Fishlow, 1995 and Forbes,
1998). However, the assumption of independence between the two is seriously in doubt.

In the growth and distribution literature, much has been made of the empirical regularity
of a lack of systematic correlation between per capita income and inequality at the aggregate
level, which is sometimes used as the basis for separating out growth and redistribution in policy
discourse, particularly to emphasize the primacy of “growth” instruments for poverty reduction.
But, as we have argued, such reduced form relationships have little to say on policy instruments,
and what is needed is empirical work on the distributional consequences of specific policy
instruments. Such work is now under way and it will be illustrated with specific country studies
 9

in Section 5. The key then is to find combinations of instruments that will deliver both growth
and equity and, in this task, inequality is clearly back on the agenda.

4. Recent changes in income inequality

Can at least some of the recent revival of interest in distribution be attributed to the brute
facts of inequality change between the last two decades? In table 1 we can observe the Gini
coefficients for all countries for which “reliable” data points spanning the last decade are
available.3 One result that becomes immediately apparent is that, while rising inequality is by no
means the norm, there have been very sharp upward movements in a number of countries. In
eleven countries shown in the table, the Gini coefficient has increased between five and nine
percentage points; in seven countries, between ten and nineteen percentage points; and in two
countries, by more that twenty (!) percentage points. These changes occurred in a span of a
decade or less. Clearly, monitoring the evolution of the Gini coefficient is no longer as unexciting
as “watching the grass grow.”

While it is true that all of the very large increases occurred in transition economies,
upward movements are witnessed in countries with very different structural characteristics.
Inequality has risen in countries that were traditionally more egalitarian (Thailand, for example)
or very unequal (Mexico, for example); in advanced nations (the United States and the UK, for
example) and in poor countries alike (Panama and Ethiopia, for example); and, in long-standing
market economies (Hong Kong, for example) or countries in transition (China and Russia, for
example). The other important fact is that there is no systematic relationship between the
evolution of inequality and growth performance: output growth was positive in 16 of the 37
countries where inequality increased and in 9 of the 14 countries where inequality declined.
Inequality has risen in both expanding economies (for example, Australia) and stagnant ones (for
example, Lithuania). As shown in Figure 1, there is no systematic relationship between trends in
inequality and growth.

3
We put “reliable” in quotations because, in general, available inequality measures for a particular country
do not necessarily provide an accurate description of the country’s degree of inequality. One source of the
problem is the underreporting of income (and expenditures) that occurs more frequently at the lower and
higher ends of the income spectrum. As a result, an inequality measure for the same country, using the
same survey and for the same year can vary by several percentage points depending on whether the data on
income was corrected for underreporting and on the method followed to do the correction. For an
illustration of this see Lustig and Szekely’s (1998) paper on Mexico.
 10

Observers and analysts have attempted to explain the increases in inequality as stemming
from the same global forces of skilled-biased technological change and, for the higher income
countries, of increased trade with cheap countries whose abundant factor is unskilled (cheap)
labor. However, an equally important result observed in table 1 is that for a very large number of
equally diverse countries inequality has remained practically unchanged (Colombia, India, Japan,
Morocco, and Sweden, for example) and it has even fallen in a few (for example, Bangladesh,
Canada, Honduras and Tunisia). What accounts for the difference in outcomes?

A. B. Atkinson (1998) addresses this question for the case of advanced industrialized
countries. Noting that the latter featured a divergent pattern in income inequality during the 1980s
and early 1990s, he posits that in addition to differences in government policies, social norms
may have played a role.4 While the Gini coefficient rose by more than 40 percent in the U.K.,
over 10 percent in Japan and Germany (in the early nineties), and close to 10 percent in the
United States, it fell slightly for Canada, and more pronouncedly for Italy, and, particularly,
France.5 One fact in support of this sociological theory of the determination of labor remuneration
is that all the countries (in Atkinson’s study) for which the Gini rose were governed by center-
right or outward rightwing governments while those for which the Gini declined were governed
by center-left social democratic government. The question that remains in place is in what
direction did the causality go. Did countries with more egalitarian social norms vote for more
leftist governments? Or, once the more right-wing governments were voted in, did inequality
increase as a result of the influence on social norms by non-egalitarian leaders?

In the case of the transition economies of Eastern Europe and the former Soviet Union, it
is conceivable that government policy made an important difference. One interesting fact is that
in three out of the only four countries where income inequality practically did not change
between the period right before and the years immediately after the dismantling of the
Communist regime, the governments pursued what Milanovic (1998) called "populist” policies.
In Hungary, Poland and Slovenia, the Gini for 1993-95 hovered around levels below .25, and they

4
It is better to not refer to them as OECD countries since the “club” now includes NIC’s such as Mexico,
South Korea and Turkey.
5
France is quite remarkable in this respect. With a Gini of Latin American standards—i.e., close to .5--in
the early 1960s, it fell to .35 in the mid-1980s and it continued to decline afterwards through the end of the
decade. Data for the 1990s, however, is not available.
 11

are among the most egalitarian countries of the previous Soviet bloc.6 In these countries wages,
cash social transfers and non-wage private income rose. Furthermore, while wage inequality
contributed to increase inequality, the cash social transfers were equalizing. Milanovic argues that
in those countries the governments attempted to cushion the population from the sharp
contractions in GDP. On the other end, the countries that saw the sharpest increases in inequality
such as Russia and the Ukraine were under governments which Milanovic calls “non-
compensators.”7 Nonetheless, it is important to note that not all countries with such governments
experienced the sharp increases in inequality observed in those two.

In the end, the answer to the question of why outcomes are so different in countries that in
appearance have similar structural characteristics and are subject to similar global forces may lie
in the following explanation. The size distribution of income is the result of a number of complex
forces which sometimes move in the same direction but sometimes countervail each other even to
the point that their effects cancel each other out. Some of these forces stem from the evolution of
the distribution of endowments and their market returns; the latter, in turn, are affected by trends
in technology, openness, and decisions affecting the rate of labor market participation and
occupational choice. Other forces are related to family formation decisions such as the choice of
partner and the number of children. Yet other forces are related to tax and transfer policy and how
it affects after-tax income levels (or what is sometimes called “unearned” income). A closer
examination at these various forces and how they reinforce or counteract each other shall be the
topic of our next section.

5. The Complex Patterns of Inequality Change

We wish to argue that one of the reasons why inequality is back on the agenda is that in
recent years, based on increasingly more widely available time series of household surveys for
developing countries, analysts have gone beneath the surface of the overall Gini coefficient to
explore the micro structures of inequality evolution. Such detailed investigations have revealed
interesting insights into income distribution processes, and have raised issues about alternative
entry points for policy. In this section we illustrate these new developments with specific case
studies.

6
Inequality in Poland began to rise after 1992 which is why Poland appears among the countries with
rising inequality in table 2.
 12

In what follows we shall highlight how changes in returns, endowments and preferences may
contribute to the evolution of the size distribution of income by drawing on three case studies:
Brazil, Mexico and Taiwan. All three studies use the micro-simulation decomposition
methodology proposed by Bourguignon, Fournier and Gurgand (1998) for Taiwan. Briefly, this
methodology decomposes the observed trends in the distribution of income in returns or price
effects, endowment effects, labor participation and occupational choice effects, and --whenever
data permits-- unearned income effects. Its advantage over other existing decomposition
methodologies is that it allows one to simulate changes on the entire distribution. Second, it
provides a much richer texture of the forces behind observed trends in inequality than the
standard group decomposition or decomposition by source exercises.8

Brazil
Using the Theil index, Ferreira and Paes de Barros (1998) find that household per capita
income inequality in urban Brazil fell from 0.88 in 1976 to 0.68 in 1996 (the Gini fell from 0.62
to 0.59 over the same period). However, the authors found that - controlling for age and gender -
the earnings-education profile for both wage-earners and self-employed workers became
substantially more convex over that period, implying an increase in the returns to education over
most pairwise comparisons. This places Brazil in the same category as many other places -
including Taiwan and Mexico - where steeper returns to education were also observed. The
returns to experience (measured by age) were essentially unchanged. There was a reduction in the
gender earnings gap (as measured by the ‘returns to being male’) for both types of worker, but
simulations indicate that this effect was not sufficient to outweigh the dispersion-increasing effect
of a steeper earnings-education profile. Overall, then, the returns or price effects were
unequalizing.

How can this be reconciled with the decline in inequality actually observed? Applying
the Bourguignon et. al. methodology, the authors find that there are three main equalizing factors
at work in Brazil over the period which, together, more than offset the unequalizing effect of
increasing returns to education. The first was changes in the educational composition of the
population. As the average years of schooling of Brazilians rose from 3.8 to 5.9 over the period,
the simulated effect of changes in educational endowments on both the individual earnings
distribution and on that of household per capita incomes was equalizing. The second effect was

7
For work on Russia see also Commander and Lee (1998).
 13

demographic in nature, and can be seen – at least in part – as a secondary effect of the rightward
shift in the distribution of education: higher levels of schooling (particularly for women)
contributed to a noticeable reduction in family sizes (largely through a decline in the number of
children). The average household size declined from 4.3 to 3.5 persons, and the dependency ratio
also fell. This decline was more pronounced for poorer households, leading to a disproportionate
relative increase in their per capita incomes and a reduction in inequality.

The third effect was a reduction in the variance of returns to ‘unobserved’ characteristics,
which include skills unrelated to education, regional location, race, and the firm size in which one
works. This effect suggests a reduction in the degree of segmentation in the Brazilian labor
market during the 1976 to 1996 period, as well as a possible decline in regional inequalities.
These trends are compatible with some of the earlier evidence from both static and dynamic
decompositions of scalar inequality measures, reported by Ferreira and Litchfield (1998). They
also highlight the continuing importance of investigating the evolution of discrimination and
segmentation in the Brazilian labor market in a more systematic way.

Mexico
Income inequality in Mexico rose sharply between 1984 and 1994: the Gini coefficient
increased from .491 to .549, respectively.9 As the Lorenz curves present no crossings, the
increase is unambiguous.10 Applying the micro-simulation decomposition methodology referred
to above, Bouillon, Legovini and Lustig (1998) attempt to identify which factors lie behind the
rise in inequality using a reduced-form household income regression model. The results of this
exercise revealed that the widening gap in the “returns” to education explain close to fifty percent
of the observed increase in inequality, while the “returns” to regional location around 24 percent
with the South alone accounting for 15 percent of the latter.11 Hence, most of the rising inequality
in Mexico should be ascribed to increasing disparities in returns. Endowment effects—that is, the
distribution of skills, etc.—account for about a fourth of the increase in inequality. In contrast to
the case of Brazil (above) and Taiwan (below), the unequalizing effect of the widening gap in the

8
For a detailed description of this method and its application see the proposal by Bourguignon, Ferreira
and Lustig (1997).
9
Bouillon, Legovini and Lustig (1999), Figure 1. The rise in inequality is robust to the inequality measure
and to adjustments to income to account for economies of scale in the household and underreporting.
10
Lustig and Szekely (1998).
11
Because the decomposition method is applied to a reduced form household income model, strictly
speaking the estimated coefficients are not returns. They capture not only the market returns to the
corresponding characteristics but also a whole range of endogenous decisions such as labor force
participation and occupational choice.
 14

returns to skill was not compensated by a more equal distribution of (observable and
unobservable) skills.

Several explanations have been put forward to account for the observed sharp increase in
the returns to education. One revolves around institutional changes in the labor market such as
reductions in the minimum wage, the decreasing strength of trade unions, and the declining share
of state-owned enterprises. However, the distribution of the real wages, for example, does not
reveal any truncation around the minimum wage, which suggests that the latter was not binding
during the period under consideration.12 In addition, there is no evidence that the distribution of
union wages is significantly different from non-union wages, once differences in education levels
are accounted for. However, the average wages for unionized workers relative to those for non-
unionized workers declined over the period13, a trend that should make the distribution of wage
income more equal.

The other explanation for the rising returns to education is that the demand for labor has
become more skill-biased. Using Mexico’s employment surveys, Cragg and Epelbaum (1995)
concluded that the increase in wage inequality can be attributed to the rapid increase in wages for
more educated and experienced workers, while the wage increase was minimal among less skilled
workers. The authors concluded that this trend is caused by a shift in demand, skewed in favor of
higher skills more than a uniform increase in demand faced with different elasticities of labor
supply by skill.

The skill(education)-biased shift in demand for labor could be the result of several
factors. Two in particular have been analyzed in the literature: trade liberalization and technical
change. Hanson and Harrison (1995), for example, found that 23 percent of the increase in
relative wages for skilled workers during the period 1986-1990 could be attributed to the
reduction in tariffs and the elimination of import license requirements. Revenga (1995) finds that
reductions in import quota coverage and tariff levels were associated with moderate falls in firm-
level employment. Revenga’s analysis suggests that there has been a slight shift in the skill mix in
favor of skilled labor (non-production workers in her sample). It also indicates that employment
and wages for unskilled labor (production workers in her sample) are significantly more
responsive to reductions in protection levels, which the author attributes to the fact that unskilled

12
Láchler (1997), p. 7.
13
OECD Economic Surveys 1996-1997: Mexico
 15

workers were more heavily concentrated in industries that underwent large reductions in
protection.

Tan and Batra (1997) find that investments in technology as measured by investments in
research and development and labor training, as well as export-orientation have a large impact on
the size-wage distributions for skilled workers and a relatively smaller effect on wages paid to the
unskilled. For them, this asymmetric impact of technology is consistent with skill-biased
technical change. This view is supported by the fact that there is evidence that in Mexico foreign-
owned exporting enterprises, which operate in the most export-oriented sectors, and those that use
technology of other enterprises pay higher wages to skilled workers. As these enterprises are in a
more favorable competitive position in the world economy, they have potential to increase
production and increase in numbers as the integration process progresses and consolidates,
especially in connection with the Free Trade Agreement. Even if Mexico invests in educational
and training institutions, it may not be able to do it at a pace sufficiently rapid to counter the
trends in labor demand. Demand and remuneration for skilled workers might thus continue to
rise, and accordingly, the wage gap between skilled and unskilled labor might widen even further.

Taiwan
Taiwan is noted for having a low and stable level of inequality. The Gini coefficient, for
example, hovered around .30 for approximately the past thirty years. A closer look at the
evolution of income distribution, however, reveals that this is by no means the result of the
absence of changes in earning patterns, demographic structure, or labor choices. On the contrary,
it is the outcome of a number of complex forces with often opposing effects but which ended up
compensating each other to the point that the net effect was almost imperceptible.14

While it is true that overall inequality practically did not change, the distribution of
primary (before taxes and transfers) adult equivalent income became more unequal, implying that
government policy was effective at countervailing other forces. On the other hand, the
distribution of income of individual earnings became more equal. Does this mean that Taiwan
was spared from the widening gap in returns to education observed in many other countries? If
not, what forces operated to compensate what in other countries has been at the bottom of rising
overall inequality? What factors have been at play to make the distribution of primary income
more unequal in spite of lower earnings inequality?
 16

In their study, Bourguignon, Fournier and Gurgand (1998) find that Taiwan has
experienced the same trend in rising returns to education as it has been for countries throughout
the world. Interestingly, this occurred despite of the incredibly rapid expansion in the supply of
labor with more years of education. For example, average years of schooling in Taiwan increased
from 6 to 9.5 years between 1970 and 1995. However, the unequalizing effect on the distribution
of earnings implied by higher returns to education was more than offset by three other factors in
order of importance: i. a decline in the variance of unobserved characteristics; ii. a change in the
level and distribution of schooling among wage earners; and, iii. the fact that more women chose
to participate in the labor force and some men dropped out. Because of the initial gap in earnings
between men and women, these changes in labor force participation were equalizing. The net
result of all these forces was a lower inequality in the distribution of individual earnings.

At the level of the household--i.e., when earners were paired with other earners, or
earnings were paired with other sources of income--the result was that overall primary (adult
equivalent) income inequality increased over time. One interesting observation is that changes in
labor force participation and occupational choice were unequalizing. This is the result of two
phenomena. First, the negative effect of the husband’s income on married women’s participation
in the labor force weakened over time: that is, more women married to higher income earners
chose to participate in the labor force compared to the past. Also, while the fact that women
entering the labor force were more educated had an equalizing effect on the distribution of
earnings, this trend made the distribution of income more unequal at the level of the household.
This is so because better educated women tend to live in households with higher incomes to begin
with. On the other hand, the equalizing effect of the expansion of education was weaker at the
household level because of its negative impact on labor force participation of younger cohorts,
particularly those coming from the lower end of the income spectrum.

What the three studies reveal is that behind apparently straightforward trends, there are
major structural forces at play. Some of these forces can be directly affected by policy, such as
the average schooling and the distribution of the stock education—but it takes time--. Other
forces are strictly market outcomes or stem from individual preferences in terms of labor force
participation and the choice of partner or the number and spacing of children. The influence of
policy measures on the latter is likely to be more ambiguous. Finally, it is important to mention

14
The results discussed here are from the study of Bourguignon, Fournier and Gurgand (1998).
 17

once more that as the case of Taiwan has shown, tax and transfer policies can counteract the
observed increases in primary income inequality.

6. Inequality between countries

The final reason we consider for why inequality is back on the agenda is the evolution of
inequality between countries. At a very simple level, in table 2 we present the coefficient of
variation of per capita GDP across countries since 1980. By this measure it can be observed that
inequality across nations has been steadily on the rise. This widening gap, seen in the historical
context, has led to a burgeoning literature which has, at least indirectly, put inequality back on the
agenda.

If the world behaved as predicted by the simple neoclassical growth model, the per capita
incomes of countries with the same preferences (i.e., saving rate), technologies, government
policies, and population growth, would eventually converge. This framework implies that the
differences in the countries’ steady state income per capita could be explained in full by
differences in the latter and how far they are from their steady state equilibrium. Initial conditions
do not matter and transitory shocks do not alter the long run equilibrium. If, on the other hand, the
world is characterized by multiple, locally stable, equilibria, convergence may not ensue.
Different initial conditions or transitory shocks could push countries with even the same
preferences, technologies, etc., into permanently different long term equilibria.

Whichever model is correct, the question is why are some countries kept in the low-
income club, and can something be done to reverse this? Although no clear consensus has
emerged whether conditional convergence is validated by the empirical analysis, there is enough
evidence in support of the view that the world seems to be converging towards two clubs: the rich
15
and the poor countries. More precisely, a sort of conditional convergence—i.e., an inverse
relationship between output per capita growth and income per capita level--appears to apply to
countries with incomes per worker above the world average. In contrast, those on the opposite
end of the spectrum seem caught in a “poverty trap.” Historically, there has been massive
 18

divergence in absolute and relative per capita incomes with steady and nearly equal growth rates
for the leaders in the long run while the currently poor countries always had low growth rates.16
This phenomenon explains the difficulty of finding strong tendencies towards divergence or
convergence for the whole sample of countries.17

A number of explanations have been put forward in the literature as to why such “poverty
traps” may arise. In the extensions of the simple neoclassical framework this could reflect
differences in steady state equilibrium resulting from low investments in human capital and
political instability.18 It could also be the result of a protracted transition due to a series of large
adverse external shocks. Ros (1998) tries to address this issue by grouping the sample of
countries used in Mankiw, Romer and Weil (1992) in growth performance categories, and
estimating their steady state incomes per capita. Subsequently, the relative position of countries in
their steady states is compared with the actual income gaps. Since the former are much narrower
than the latter, the differences would seem to arise from the fact that poorer countries are further
away from their steady state. However, this would imply—according to the neoclassical model—
that the poorer countries should be growing faster, something that is not validated by their actual
performance. (Ros, pp. 48-51)

Although this exercise is not conclusive, it does support the view that economic growth
may be better characterized by multiple equilibria: that is, a world were initial conditions matter
and transitional shocks have long-run implications. In such a world, poverty traps stem from
countries being pushed—by either initial conditions or shocks—into a low-level equilibrium. In
the development literature, particularly in the early one, several simple models of low-level
equilibrium traps were put forward. One is the insufficient savings theory first introduced by
Leibenstein (1957) and subsequently formalized by King and Rebelo (1993). The scenario
predicted by this approach is that at low levels of the capital-labor ratio income per capita is
barely enough for subsistence, and savings fall below depreciation. The country is “stuck” in a
low-level equilibrium trap. Thus, although the country may be filled with good investment
opportunities, it is too poor to take advantage of them. Foreign investment or foreign aid would

15
On the formation of a two-club world see Quah (1993), Ros (1998).
16
Pritchett (1997).
17
Of course, the observed patterns in the relationship between growth and income per capita may be
reflecting gaps across countries in their position vis-à-vis their steady state. The simple neoclassical
framework is perfectly consistent with a country which is closer to its low steady state income per capita
level may grow more slowly than a richer country which is much further away from its steady state
equilibrium.
 19

take care of such a problem, though. The only factor that could counter this, then, would be
political risk (deterring private capital flows) or international misbehavior (deterring foreign aid).
As mentioned above, Barro (1991, 1997) finds that when in addition to controlling for human
capital differences, one controls for the role of political risk, convergence—albeit slow—seems to
ensue. Ros (1998), however, finds an interesting result: if countries are classified by income level,
at the low end of the spectrum, differences in the rule of law index (used by Barro) appear too
small compared with the large differences in growth rates observed among the countries in this
group.

There are many ways in which the debate on the seemingly inexorable growth of
inequality between countries is impinging on the development discourse. In fact, the multiple
equilibria and club convergence perspectives do have strong implication for policy. The club
convergence view gives credence to the view that policy matters, that a country’s evolution is not
preordained but can be altered by intervention—which leads to the far from resolved debate on
what exactly these interventions ought to be. In so far as initial conditions include low levels of
physical and human capital, these perspective also support the arguments for foreign aid to help
poor countries break out of their low level traps.

The most recent literature on aid, however, has thrown doubts on the historical record of
aid in actually helping poorer countries to grow. Burnside and Dollar (1997) and World Bank
(1998), and a host of other recent writings (eg Kanbur, 1999) point to the dismal record of aid in
breaking the poverty traps identified in the growth literature. The reason, it would seem, is that
resources by themselves are not enough. How they are used is crucial. Burnside and Dollar
(1997) show that aid promotes growth only when it flows into “good” policy environments—
otherwise, it is at best ineffective. This is, in fact, the empirical counterpart to the general
theoretical proposition that in a second best world simply expanding the resource base may in fact
do no good or may even immiserize a society. What matters, primarily, is domestic policy—and
recall the complementarity between redistribution and efficiency in this case. Thus the evolution
of inequality between nations is not independent of the inequality within each nation.

18
Barro (1991, 1997); Mankiw, Romer and Weil (1992).
 20

7. Conclusion

It could of course be argued that inequality never went away—it was always on the
agenda in some form or another. The untidy way in which academic analysis advances means that
one can always find discussion of a topic during the period it is claimed to have fallen off the
radar screen. However, we believe that the turn of the century has seen a resurgence of interest in
inequality, and distribution more generally. This resurgence is the culmination of several trends in
the analytical, empirical and policy literature.

First, the separation of distribution and allocative efficiency, an immediate implication of

post war attempts to formalize Adam Smith’s invisible hand, no longer holds sway. As the real
world features of imperfect information and imperfect markets have been recognized and
incorporated into the analytical framework, a more complicated terrain has emerged where equity
and efficiency are integrally connected, and detailed analysis is required to discover these
connections.

Second, the empirical regularities between inequality and growth are increasingly being
seen for what they are—regularities which have little in the way of specific policy implications.
The current consensus is that there is no systematic empirical relationship between inequality and
growth—but this type of “separation” only has policy implications if it is further assumed that
there are separable sets of “growth” and “redistribution” policy instruments. Once specific policy
instruments are considered, and this is what policy makers are interested in, inequality is back on
the agenda.

Third, aggregate measures of inequality have changed dramatically for some countries in
the last decade. This has brought inequality back on the agenda for these countries, of course, but
also more generally in terms of the implications for other countries. Now, it is true that inequality
has not changed much for other countries, or has in fact declined in some. At the very least,
however, there is the issue of why countries facing broadly similar global circumstance have had
such widely different experiences on inequality change. What is it about the domestic
circumstances that has led to these divergences of outcome? Such a question puts inequality back
on the agenda.
 21

Fourth, whether aggregate inequality changed or not, recent empirical work is showing
how the evolution of overall inequality is the outcome of a complex pattern of forces beneath the
surface. These include the evolution of markets, assets and institutions, overlaid on basic
demographic shifts. Understanding these forces is the real challenge which takes us beyond the
simple lack of a correlation between aggregate inequality and per capita income. It requires new
techniques and opens up new avenues of policy analysis.

Fifth, global divergence has put the issue of inequality between nations back on the
agenda. In the last decade this has led to a rich literature on what determines relative growth rates
and whether growing inequality between nations is inevitable. This debate is also linked to the
role of foreign aid in overcoming the poverty traps which the literature has identified.

The five strands of the literature we have identified look set to define the research and
policy agenda in the first decade of the next century. They have already led to fruitful insights and
significant policy implications, even beyond the realm of distributional analysis. Yes, inequality
is definitely back on the agenda.
 22

References

Ahluwalia, M. (1976), Inequality, Poverty and Development, Journal of Development Economics,

3: 307-342.

Alderman, H., P-A. Chiappori, L. Haddad, J. Hoddinott and R. Kanbur (1995), Unitary Versus
Collective Models of the Household: Is it Time to Shift the Burden of Proof? World Bank
Research Observer, 10: 1-19.

Alesina, A. and D. Rodrik (1994), DistrbutivePolitics and Economic Growth, Quarterly Journal
of Economics, 109: 465-490.

Anand, S. and R. Kanbur (1993a), The Kuznets Process and the Inequality-Development
Relationship, Journal of Development Economics, 40: 25-52.

Anand, S. and R. Kanbur (1993b), Inequality and Development: A Critique, Journal of

Development Economics, 41: 19-43.

Atkinson, A. B. (1997), Bringing Income Distribution in from the Cold, Economic Journal 107:
297-321

Atkinson, A. B. (1998), Equity Issues in a Globalizing World: the Experience of OECD

Countries, International Monetary Fund, Conference on Economic Policy and Equity, June 8-9,
Washington, D.C.

Atkinson, A. B. and F. Bourguignon (1998), Introduction to Handbook of Income Distribution,

North Holland, forthcoming.

Bardhan, P., S. Bowles and H. Gintis (1999), Wealth Inequality, Credit Constraints, and
Economic Performance, in A. B. Atkinson and Francois Bourguignon (eds.) Handbook of Income
Distribution, North Holland.

Barro, R. J. (1991), Economic Growth in a Cross Section of Countries, Quarterly Journal of

Economics, 106: 407-44.

Barro, R. J. (1997). Determinants of Economic Growth: a Cross-Country Empirical Study, MIT

press.

Bénabou, R. (1996), Inequality and Growth. In B.S. Bernanke and J. Rotemberg (eds.), NBER
Macroeconomics annual 11, MIT Press.

Boadway, R. and M. Keen (1999), Redistribution, in A. B. Atkinson and F. Bourguignon (eds.)

Handbook of Income Distribution, North Holland, forthcoming.

Bouillón, C., A. Legovini and N. Lustig (1998). Rising Inequality in Mexico: Returns to
Household Characteristics and the “Chiapas effect”. Inter-American Development Bank,
Mimeo.

Bourguignon F., M. Fournier, and M. Gurgand (1998). Distribution, Development and Education:
Taiwan, 1979-1994. World Bank, Mimeo.
 23

Bourgignon F., F. Ferreira and N. Lustig (1997). The Microeconomics of Income Distribution
Dynamics in East Asia and Latin America. IDB-World Bank, Research proposal.

Burnside, C. and D. Dollar, (1997). Aid Spurs Growth—In a Sound Policy Environment. Finance
and Development V34, n4 pp.4-7.

Chenery. H., M. Ahluwalia, C. Bell, J. Duloy and R. Jolly (1974), Redistribution with Growth,
Oxford University Press, Oxford.

Commander, S. and U. Lee, (1998). How does Public Policy Affect The Income Distribution?
Evidence from Russia, 1992-1996. Mimeo

Cragg, M and M. Epelbaum (1995). Why Has Wage Dispersion Grown in Mexico? Is it the
Incidence of Reforms or the Growing demand for Skills? Journal of Development Economics
Vol. 51, No. 1.

Datt, G. and M. Ravallion (1992), Growth and Redistribution Components of Changes in Poverty
Measures: A Decomposition with Applications to Brazil and India in the 1980s, Journal of
Development Economics, 38: 275-295.

Deininger, K. and L. Squire (1996). A new Data Set Measuring Income Inequality. World Bank
Economic Review 10: 565-591.

Dixit, A. and T. Besley (1997), James Mirrlees' Contributions to the Theory of Information and
Incentives, Scandinavian Journal of Economics, 99: 207-235

Deininger, K. and L. Squire (1996), A new Data Set Measuring Income Inequality, World Bank
Economic Review, 10: 565-592.

Ferreira, F and J. Litchfield (1998): “Education or Inflation?: The Roles of Structural Factors and
Macroeconomic Instability in Explaining Brazilian Inequality in the 1980s”; LSE STICERD
DARP Discussion Paper No. 41 (July).

Ferreira, F. and R. Paes de Barros (1998). Climbing a Moving Mountain: Explaining the Decline
in Income Inequality in Brazil from 1976 to 1996. Mimeo.

Fishlow, A (1972), Brazilian Size Distribution of Income, American Economic Review, 62: 391-
402.

Fishlow, A (1995), Inequality, Poverty and Growth: Where do we Stand? In M. Bruno and B.
Pleskovic (eds.) Proceedings of the 1995 Annual World Bank Conference on Development
Economics, World Bank, Washington, D. C.

Forbes K. (1998). A Reassessment of the Relationship Between Inequality and Growth, Working
Paper, MIT.

Gallup, J. L., S. Radelet, A. Warner (1998), Economic Growth and the Income of the Poor,
Mimeo, Harvard Institute for International Development.
 24

Greenwald, B. and J. E. Stiglitz (1986), Externalities in Economies with Imperfect Information

and Incomplete Markets, Quarterly Journal of Economics, 101:229-264

Haddad. L. and R. Kanbur (1990), How Serious is the Neglect of Intrahousehold Inequality?
Economic Journal, 100: 866-881.

Hanson, G., and Harrison A (1995). Trade, Technology and Wage Inequality in Mexico. National
Bureau of Economic Research (NBER) Working Paper No. 5110, May.

Hoff, K. (1994), The Second Theorem of the Second Best, Journal of Public Economics, 45:
223-242.

Kakwani, N. (1994), Poverty and Economic Growth, with Application to Cote d’Ivoire, Review of
Income and Wealth, 39: 121-139

Kanbur, R. (1987), Measurement and Alleviation of Poverty, IMF Staff Papers, 36: 60-85.

Kanbur, R. (1990), Poverty and the Social Dimensions of Adjustment in Cote d’Ivoire, Social
Dimensions of Adjustment Paper no. 5, World Bank, Washington, D. C.

Kanbur, R. (1998), Income Distribution and Development, in A.B. Atkinson and F. Bourguignon
(eds.) Handbook of Income Distribution, North Holland, forthcoming.

Kanbur, R. (1999), A Framework for Thinking Through Reduced Aid Dependence in Africa,
Working Paper no. WP 99-06, ARME, Cornell University.

Kanbur, R. and L. Haddad (1994), Are Better off Household More Unequal or Less Unequal?
Oxford economic Papers, 46: 445-458.

King, R.G., and S. T. Rebelo (1993). Transitional Dynamics and Economic Growth, in the
Neoclassical Mode, American Economic Review 83: 908-31.

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

Láchler, U. (1997). Education and Earnings Inequality in Mexico. Washington, DC : World

Bank, Mexico Country Department.

Lewis, W. A. (1954), Economic Development with Unlimited Supplies of Labour, Manchester

School of Economic and Social Studies, 22: 139-181.

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

Luxembourg Income Study, August 1998. Income Distribution Measures Based on LIS Data,
website: http://lissy.ceps.lu/inequality.htm

Londoño, J. L. and M. Szekely (1997). Persistent Poverty and Excess Inequality: Latin America
During 1970-1995. OCE Working Paper Series No. 358. Washington, D.C.: Inter-American
Development Bank.
 25

Lundberg, M. and L. Squire (1999), Inequality and Growth; Lessons for Policy, mimeo, World
Bank, Washington, D. C.

Lustig, N. and M. Szekely (1997). México: evolución económica, pobreza y desigualdad.

Technical Study, Sustainable Development Department. Washington, D.C.: Inter-American
Development Bank.

Mahalanobis, P. (1963) The Approach of Operational Research to Planning India, Asia Press,
Bombay.

Mankiw, G., D. Romer, and D. Weil (1992), A Contribution to the Empirics of Economic
Growth, Quarterly Journal of Economics 107: 407-37.

Meade, J. E. (1964), Efficiency, Equality, and the Ownership of Property, Allen and Unwin,
London.

Milanovic, B. (1998). Income, Inequality and Poverty During the Transition from Planned to
Market Economy. Washington, D.C.: The World Bank.

Mirrlees, J. (1971) An Exploration in the Theory of Optimum Income Taxation, Review of

Economic Studies, 38: 175-208.

Organization for Economic Co-operation and Development. OECD Economic Surveys 1997-
1998. Mexico

Rosenstein-Rodan, P. (1943), Problems of Industrialization in Eastern and Southern Europe,

Economic Journal, 53: 202-211.

Revenga, A. (1995). Employment and Wage Effects of Trade Liberalization: The case of Mexican
Manufacturing. Washington, D.C.: World Bank. Mimeo.

Ros, J. (1998). Increasing Returns, Development Traps, and Economic Growth. University of
Notre Dame, Draft.

Sen, A. K. (1983), Resources, Values and Development, Blackwell, Oxford.

Tan H. and Batra G. (1997). Technology and Firm Size Wage Differentials in Colombia, Mexico,
and Taiwan (China). World Bank Economic Review 11 59-83.

World Bank (1997), Taking Action to Reduce Poverty in Sub-Saharan Africa, Washington, D.C.

World Bank (1998), Assessing Aid: What Works, What Doesn’t, and Why. Oxford University
Press.
 26

TABLE 1
COMPARING THE AVERAGE GINI COEFFICIENT FOR THE 80s AND THE 90s
Countries for which the average of the Gini coefficient Countries for which the average of the Gini coefficient
of the 90s is greater than that of the 80s of the 90s is lower than that of the 80s
(Inequality increased) (Inequality decreased)
Country (1) 80s 90s Welf (2) Unit (3) Source (4) GDP p.c. (5) Country (1) 80s 90s Welf (2) Unit (3) Source (4) GDP p.c. (5)
Australia 2 9 .7 3 3 1 .7 0 ID P L IS + Bahamas 4 4 .4 2 4 2 .9 9 I H D /S -
Belarus* 2 3 .0 0 2 8 .0 0 I P M - Bangladesh 3 6 .8 6 3 4 .8 7 I P D /S +
Brazil 5 7 .6 4 6 0 .9 0 I P L /S - Canada 3 1 .5 4 2 7 .6 1 I H D /S +
Bulgaria** 2 3 .0 0 3 4 .0 0 I P M + Finland 3 0 .3 0 2 6 .1 1 I H D /S +
China* 3 0 .4 1 3 5 .0 3 I P D /S + Ghana 3 6 .3 2 3 3 .9 4 E P D /S -
Czech Rep* 1 9 .0 0 2 7 .0 0 I P M - Honduras 5 9 .9 0 5 5 .5 3 I P L /S -
Czechoslovakia 2 1 .1 3 2 4 .5 6 I P D /S … Indonesia 3 3 .4 4 3 2 .3 9 E P D /S +
Denmark 3 2 .0 7 3 3 .2 0 I H D /S + Italy 3 3 .4 2 3 2 .1 9 I H D /S +
Dominican Rep. 4 6 .8 8 4 9 .0 0 I P D /S + Jamaica 4 3 .3 5 3 9 .8 3 E P D /S +
Estonia** 2 3 .0 0 3 5 .0 0 I P M - Mauritius 3 9 .3 7 3 6 .6 9 E P D /S +
Ethiopia 4 1 .0 0 4 4 .5 0 E P WB + Philippines 4 6 .0 8 4 5 .0 0 I H D /S -
Finland 2 0 .7 0 2 2 .3 0 ID P L IS + Slovak 2 0 .0 0 1 9 .0 0 I P M -
Hong Kong 4 1 .4 9 4 5 .0 0 I H D /S + Spain 3 2 .1 0 3 0 .6 0 ID P L IS +
Hungary 2 1 .0 0 2 3 .0 0 I P M + Tunisia 4 3 .0 0 4 0 .2 4 E P D /S +
Jordan 3 8 .4 5 4 0 .6 6 E P D /S -
Kazakhstan* 2 6 .0 0 3 3 .0 0 I P M -
Kyrgyz Rep.** 2 6 .0 0 5 5 .0 0 I P M -
Latvia** 2 3 .0 0 3 1 .0 0 I P M - Countries for which the average of the Gini coefficient
Lithuania** 2 3 .0 0 3 7 .0 0 I P M - of the 90s is approximately equal to that of the 80s
Mexico 4 4 .9 3 4 7 .6 0 I P L& S + (Inequality has not changed) (6)
Moldova** 2 4 .0 0 3 6 .0 0 I P M - Country (1) 80s 90s Welf (2) Unit (3) Source (4) GDP p.c. (5)
Nigeria 3 7 .0 2 3 9 .3 1 E P D /S + Belgium 2 6 .4 3 2 6 .9 2 I H D /S +
New Zealand* 3 5 .3 1 4 0 .2 1 I H D /S + Chile 5 5 .5 5 5 6 .4 9 I P D /S +
Norway 3 1 .6 9 3 3 .3 1 I H D /S + Colombia 4 8 .3 0 4 9 .2 5 I P L /S +
Panama* 5 2 .1 0 5 7 .4 0 I P L /S - Costa Rica 4 6 .2 6 4 6 .8 3 I P L /S +
Peru 4 2 .7 5 4 4 .8 7 E P D /S - India 3 1 .4 5 3 1 .4 1 E P D /S +
Poland 2 4 .5 9 2 8 .2 6 I P D /S - Israel 3 0 .9 0 3 0 .5 0 ID P L IS +
Romania* 2 3 .0 0 2 9 .0 0 I P M - Japan 3 5 .2 0 3 5 .0 0 I H D /S +
Russia*** 2 4 .0 0 4 8 .0 0 I P M - Lesotho 5 6 .0 2 5 7 .0 0 E P D /S +
Slovenia 2 2 .0 0 2 5 .0 0 I P M … Luxembourg 2 3 .8 0 2 3 .7 0 ID P L IS +
Taiwan 2 9 .0 5 3 0 .5 4 I P D /S … Morocco 3 9 .1 9 3 9 .2 0 E P D /S +
Thailand 4 7 .2 7 5 1 .7 8 I P WB + Netherlands 2 7 .5 0 2 7 .2 0 ID P L IS +
Turkmenistan** 2 6 .0 0 3 6 .0 0 I P M - Portugal 3 6 .8 0 3 6 .2 0 I H D /S +
UK* 2 7 .3 2 3 2 .3 5 I Pe D /S + Sweden 3 1 .7 4 3 2 .4 8 I H D /S +
Ukraine*** 2 3 .0 0 4 7 .0 0 I P M - Venezuela 4 6 .0 2 4 6 .3 3 I P L /S -
USA 3 2 .5 0 3 5 .5 5 ID P L IS + Yugoslavia 3 2 .7 4 3 1 .8 8 I P D /S …
Uzbekistan* 2 8 .0 0 3 3 .0 0 I P M -

Notes:
(1) Changes: *Between 5 and 9 points; **Between 10 and 19 points; ***20 points or more. (4) Sources:
D/S: Deininger and Squire.
(2) Welf: Whether the Gini coefficient is calculated based on income or expenditure (I = Income, E = Expenditure). LIS: Luxembourg Income Study.
ID = Disposable Income (LIS own definition). L/S: Londoño and Szekely.
M: Milanovic.
(3) Unit: Whether the base unit is the person or the household L&S: Lustig and Szekely.
He = Household equivalent (households are weighted by the number of persons). WB: World Bank
Pe = Person equivalent (in addition to He, the effective number of members in the household
is assumed to be the square root of the actual members). (5) Yearly average change in GDP per capita
P = By individual. (constant PPP 1987 US$). Growth refers to the period
H = By household. between the first and the last available Gini estimates;
"+" income growth; "-" income decline;
… data not available.

(6) Not change indicates that the variation in the

Gini coefficient was less than 1.00 in either direction.
 27

TABLE 2:
COEFFICIENT OF VARIATION OF GDP PER CAPITA
Year Coefficient of Variation
1980 0.7575
1985 0.7867
1990 0.8083
1995 0.8291

1996 0.8294

The 80s 0.7789

The 90s 0.8210

Note: The total number of countries included in the calculations

is 61.

Figure 1

25

20

15

10
Change in log of Gini coefficient

-5

-10
-15 -10 -5 0 5 10 15

Change in log of GDP per capita

Notes: (1) The figure is based on 103 non-overlapping observations for over 80 countries spanning from
the late 1970s to 1996.
(2) The fitted regression line was estimated excluding the former Soviet republics and the
countries of Eastern Europe.