WPS4504
Policy ReseaRch WoRking PaPeR 4504
Inequality in Latin America:
Determinants and Consequences
J. Humberto Lopez
Guillermo Perry
The World Bank
Latin America and the Caribbean Region
Office of the Regional Chief Economist
Febuary 2008
Policy ReseaRch WoRking PaPeR 4504
Abstract
Latin America is together with Sub-Saharan Africa the discuss channels through which inequality can affect
most unequal region of the world. This paper documents growth and output volatility. On the whole, the analysis
recent inequality trends in the Latin American region, suggests a two-pronged approach to reduce inequality
going beyond traditional measures of income inequality. in the region that combines policies aimed at improving
The paper also reviews some of the explanations that have the distribution of assets (especially education) with
been put forward to understand the current situation, elements aimed at improving the capacity of the state to
and discusses why reducing income inequality should be redistribute income through taxes and transfers.
an important policy priority. In particular, the authors
This paper--a product of the Office of the Regional Chief Economist, Latin America and the Caribbean Region--is
part of a larger effort in the department to understand the determinants of income inequality in Latin America. Policy
Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at
hlopez@worldbank.org.
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Produced by the Research Support Team
Inequality in Latin America:
Determinants and Consequences
J. Humberto Lopez and Guillermo Perry*
The World Bank
*Paper prepared for the conference "Paradigma y Opciones de Desarrollo en Latin America", Santiago de
Chile, Junio 2007. We would like to thank Maria Fernanda Rosales for excellent assistance. The views
expressed here are ours only and do not necessarily reflect those of the World Bank, its executive directors,
or the countries they represent.
I. Introduction
With an average gini coefficient of .52, Latin America stands out as one of the two1 most
unequal regions of the world. In addition to the fact that citizens dislike such high levels
of inequality (according to the Latinobarometro, in 2001 almost 90 percent of the
population of the region considered the distribution of income in the region to be unfair
or very unfair), there are good economic reasons for policy makers to be concerned with
this situation.
First, for a given average income per capita level, higher inequality implies higher
poverty levels. Moreover, beyond this qualitative assertion, the impact of inequality on
poverty is quite significant from a quantitative point of view. A simple simulation,
assuming a log normal distribution of income per capita,2 suggests that if Latin America
had the inequality levels prevalent in Europe, the poverty rate (headcount, 2$PPP) would
be closer to 12 percent than to the current estimate of 25 percent, at current levels of
average income per capita.
Second, not only high inequality leads to higher poverty levels at current income levels,
but it constitutes a barrier to poverty reduction. There are a number of studies
(Bourguignon, 2003; Ravallion, 1997, 2004; Lopez and Serven 2006a; Perry et al, 2006)
that show that the growth elasticity of poverty reduction is lower (in absolute value) in
countries with high levels of income inequality. In other words, countries with higher
inequality levels require a faster growth rate to achieve the same poverty reduction than
countries with low inequality. Once again the differences between high and low
inequality countries are sizeable. The results in Ravallion (2004) indicate that depending
on a country's initial gini coefficient, the growth elasticity of poverty could range from
almost -5 (very low inequality countries) to -.5 (very high inequality countries). Thus the
growth elasticity of poverty can be multiplied by a factor of 10 as a result of lower
inequality. More specifically, a country like Brazil would need to grow at close to 5
percent per annum �if inequality levels remain constant- to achieve the same reduction in
poverty levels than Poland could achieve by growing at just 2 percent in per capita terms.
A third reason to be concerned about high inequality is that it appears that countries with
higher inequality and poverty levels tend to grow less (see, among others, Alesina and
Rodrick 1994; Perotti 1996; Lopez and Serven, 2006; Perry et al, 2006). True, there are
also studies that find that inequality leads to faster growth (Li and Zou, 1998; Forbes
2000) and studies that find no relationship at all (Barro, 2000), but in general this is
usually the case of studies that focus on the short term impact of changes in inequality on
growth.
What are the reasons put forward in the literature to explain these findings? The
economic literature suggests several potential channels. There is, first, the political
economy argument (Alesina and Rodrick, 1994) by which the median voter of a highly
1The other being Sub Saharan Africa.
2More specifically to make this simulation we are assuming that per-capita income follows a log-normal
distribution (see Lopez and Serven, 2006 for details).
2
unequal economy may have a tendency to push for higher redistributive public
expenditures and transfers and higher taxes (assumed to negatively affect capital
accumulation) to finance the additional spending. There is also the so-called
sociopolitical instability approach (Alesina and Perotti, 1996) by which individuals in
highly unequal societies will have incentives to engage in activities outside legal markets,
such as crime and violence. For example, using a large panel of international homicide
and robbery rates, Fajnzylber, Lederman and Loayza (2002) show that countries with
higher inequality levels tend to have higher crime levels: on average a 1 percentage
increase in the Gini coefficient appears to increase crime rates by between 1 and 4
percent. Finally, there are economic arguments linked to the existence of credit
constraints (e.g. Galor and Zeira, 1993), by which such constraints coupled with fixed
costs and indivisibilities can prevent poorer individuals from investing in education or
physical capital.
Thus, there are good economic reasons, in addition to equity reasons, to be concerned
about inequality. This paper, tries to contribute to this debate along several dimensions.
First, it presents in Section II a review of the current situation of income inequality in the
region. It also explores how inequality has evolved over the past few years and discusses
some of the limits of standard indicators. Second, it discusses in Section III the reasons
that the literature has put forward to understand the high levels of inequality in the region,
namely an unequal distribution of assets and the inability of the state to correct income
inequality through taxes and transfers. And third, it reflects in Section IV about the
channels that may make inequality lead to lower growth rates and presents some new
empirical results on the impact of inequality on output volatility.
II. Inequality in Latin America
II.1 Inequality in Latin America
As noted above, the average gini coefficient in Latin America is .52.3 However this
average hides significant regional variation. In fact, there are countries like Bolivia, Haiti
or Jamaica with gini coefficients around .6.4 At the other extreme, we can find two
Caribbean countries (Trinidad and Tobago and Guyana) with a gini coefficient of .42 and
Venezuela and Uruguay with close to .45 (Panel A, figure 1).
By any standard these levels of income inequality are extremely high. Panel B reproduces
the previous chart but now accompanied by the gini coefficients of all the developing
countries for which the World Bank povcal database5 reports data. Inspection of this
panel indicates that Latin America has the highest levels of inequality. For example, the
average gini index in Sub-Saharan Africa (SSA) is .47. In the other developing regions it
is much lower and ranges from .34 in Europe and Central Asia (ECA) to .38 in East Asia
and the Pacific. Further, it is also worth noting that the most equal Latin American
3The median gini is slightly higher and equals .54.
4The data for Latin America was kindly provided by Leo Gasparini and it is based on the last available
survey for each country.
5The povcal database can be accessed at http://iresearch.worldbank.org/PovcalNet/jsp/index.jsp
3
country has a higher Gini than the most unequal of developed countries (Portugal), where
the Gini coefficient is below .4.
True, it could be that there are important biases in these statistics. Most of the figures
from Latin America come from income data while those in other part of the world are
consumption based6 and this is important because gini indices based on consumption data
tend to be substantially lower than gini indices based on income data. In fact, the analysis
in the World Development Report (2006) on Equity and Development, suggests that gini
coefficients can differ by 10 percentage points depending on whether one uses income or
consumption. For example, the income based gini coefficient for Nicaragua is around .54
but that based on consumption data is about .42. Similarly in Peru which has an income-
based gini slightly below .55 and a consumption-based gini that is below .45.
Yet, even if these differences held for all the countries, it is still worth noting that the
only region that would have inequality levels above those found in Latin America is Sub-
Saharan Africa.
Figure 1. Inequality in Latin America. Gini indices (%)
Panel A. Inequality in Latin America
65
60
xe 55
ind
50
Gini
45
40
M A N C M I
BOL HTI JA GTM BR PRY PA ECU COL PER CHL NI HND DO MEX BLZ SLV ARG CR LCA URY VEN GUY TTO
Panel B. Latin America in the global context
80
70
LAC
60 SSA MENA
SA
50 EAP ECA
indexin 40
Gi 30
20
10
0
Source: Own calculations based on Gasparini et al. (2007) and povcal data.
6The exceptions are the developed countries and a handful (about 10 percent) of developing countries.
4
II.2 How have the poor fared over the past years? A look at income levels
We have seen in the previous sub-section that Latin America's income inequality is high.
However, what are the recent trends in terms of the evolution of income inequality? To
address this issue Figure 2 plots the average growth rate in the incomes of the poor and
the average growth rate for the population as a whole. Panel A focuses on the extreme
poor (i.e. those with income levels below US$17 per person per day) and panel B on the
moderate poor (i.e. those with income levels below US$2 per person per day). To
construct these figures we have relied on household surveys for 18 countries. The first
year of the spell falls typically in the early 1990s (the average initial year for the 18
countries is 1992), and the last year in early 2000s (the average final year for the 18
countries is 2002).
The figures also plot the regression slope. If the slope relating these two variables were
equal to 1, it would indicate that on average the income of the poor has been increasing at
the same pace as the income for the average individual. If on the other hand, the slope is
smaller (larger) than 1, it would indicate that the income of the poor is increasing slower
(faster) than the income of the average individual and hence that inequality is increasing
(declining).
Figure 2. Growth in the incomes of the poor vs growth
$1 Poverty line $2 Poverty line
10 8
or y = 0.736x - 0.2239 or y = 0.7886x - 0.3432 6
poeht 5 poeht 4
2
ofe 0 ofe 0
om -15 -10 -5 0 5 10 om -15 -10 -5 -2 0 5 10
inc -5 inc
htwo -4
-10 htwo -6
Gr Gr -8
-15 -10
Growth Growth
Source: Authors' calculations
Inspection of these two figures suggests that during the 1990s the Latin American poor
have been benefiting from growth less than the average individual. Panel A in figure 2
has an associated slope of .74; the slope in Panel B is slightly higher (.78) but it still
indicates that the incomes of the poor have increased less than proportionally than those
of the non poor.8 There are two ways to look at this data. On the one hand, it is difficult
to defend that the Latin American poor have not benefited from growth during the 1990s.
Indeed, save a couple of exceptions the income of the poor has increased when average
7All international poverty lines are expressed in US$ adjusted for purchasing power parity differences.
8It is likely that these results are affected by the presence of important outliers. Yet, re-estimation of the
slopes using robust estimates to outliers tends to lower the slopes even further to between .5-.6 in the case
of the US$1 a day and to .6-.7 in the case of the US$2 a day.
5
income has increased and declined when growth has declined. Yet, on the other hand the
data suggest that they may have not benefited as much as the non poor.
Trends, of course, have varied by country as the variance in Figure 2 indicates. In
particular, while during the 1990s the Gini has increased in about two-thirds of the
countries (especially in some of the previously less unequal countries, such as Argentina
or Costa Rica), there are also countries that have experienced a marked decline in
inequality. For example, in Brazil (one of the most unequal countries of the region) the
Gini coefficient fell by 3 percentage points between 1990 and 2003. The country with the
most dramatic decline in inequality over this period was Mexico: between 1992 and 2002
the Gini coefficient declined by 4 percentage points.
Figure 3. Latin America: Change in Gini coefficients in the 90's (%)
8
6
4
2
0
-2
-4
-6
L G N Y L U D N R Y L V C X
CO AR CRI VE UR BO EC HN PA PE PR JAM CH SL NI BRA ME
Source: own calculations using data in Gasparini et al. (2007)
II.3 How have the poor really fared over the past years? A look at price levels
When we compute the evolution of income for a particular group of the population (as in
figure 2 above) there is a need to deflate the nominal income levels of each individual,
typically using the consumer price index or other suitable deflator. If all the households
in the economy faced the same inflation levels, this action should not introduce any bias
in the analysis.
However, rich and poor families consume different baskets of goods and the inflation
rates of these baskets can differ greatly. Go�i, Lopez, and Serven (2005) and Perry et al
(2006) show that using the aggregate CPI can greatly mislead actual trends and policies.
First, tax brackets, pensions, social transfers, and minimum wages are often indexed to
the CPI and using an inappropriate aggregate index can lead to engineering real transfers
among income classes that were not intended. Further, our picture of the evolution of
inequality (and hence poverty) can be sharply distorted by assuming that deflators are
similar across income classes, either by working with un-deflated nominal baskets of
6
goods, or by using aggregate deflators, and contaminating inference about the
relationship between these variables and growth or policy.9
Figure 4. Annual inflation by percentile
Brazil (1988-1996) Colom bia (1997-2003)
734.0 11.6
733.0
732.0 11.4
731.0
11.2
730.0
729.0 11.0
728.0
727.0 10.8
726.0
10.6
725.0
724.0 10.4
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Pih 1996 Weighted (L) Pi Pih 1997 Weighted (L) Pi
Mexico (1996-2002) Peru (2001-2003)
12.8 2.0
12.7 1.9
12.6 1.8
12.5 1.7
12.4 1.6
12.3 1.5
12.2 1.4
12.1 1.3
12.0 1.2
11.9 1.1
11.8 1.0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Pih 2002 Weighted (L) Pi Pih 1995 Weighted (L) Pi
Source: Go�i, Lopez and Serven (2005)
To explore these issues Figure 4 shows the "actual" inflation rate suffered by different
viniventiles of the income distribution for Brazil (1988-1996), Colombia (1997-2003),
Mexico (1996-2002) and Peru (2001-2003). These estimates have been constructed on
the basis of the consumption baskets of different population groups and on the different
evolution of the prices of the different components of those baskets. The Figure also plots
the average inflation rate (the horizontal bar).10
9 A previous literature has addressed this issue for some economies: Busso et al. (2000) estimate the
difference between the standard cpi-index and a democratic version for Argentina over the period 1989-
1998; Ruiz-Castillo et. al. (2002) study the Spanish experience while Hobijn and Lagakos (2003) have
analyzed the effective inflation rate suffered by different groups in U.S, just to mention some of them.
Beyond literature, Hong Kong Census and Statistics Department computes, in addition to an overall cpi,
consumer price indexes by income bracket and Colombia has begun to follow this steps. Within this strand,
our interest is to dig in LAC data to see if differences in consumption patterns among quintiles and
differences in price variations among products are strong enough to validate cpi's disarticulation
10Go�i, Lopez, and Serven (2005) show that these patterns persist even after adjusting for quality change
bias and after recomputing Paasche indices to control for potential substitution effects.
7
There are a couple of interesting elements that emerge from this figure. First, as it can be
seen the two lines cross between the 80th and 90th percentiles, an indication that the
reported inflation rate in practice tends to correspond to the consumption basket of the
very rich. In other words, observed inflation rates can offer limited information when
interest center on the welfare of the poor. Second, on a more positive note, the curve
measuring the inflation rate of the different viniventiles of the distribution trends up very
markedly. This is a reflection that for these spells the inflation rate suffered by the poor
has been consistently lower than the average inflation rate.11
The implications of these findings are far reaching. To begin, the message that emerges
from figure 2 may exaggerate the relative loss of welfare of the Latin American poor.
True, the income of the non poor may have increased faster but the prices they have faced
have also increased faster. Second, it is possible that concerns about the negative
distributional impacts of reforms have probably been overstated. Third, incorrect
deflation potentially confuses the relationship between different types of growth
strategies and their impact on poverty.
For instance, liberalizations, devaluations etc. all have, by their design, the goal of
changing relative prices of goods within the economy. When we ask what the impact of,
for instance, trade liberalization is on the poor, then, we need to ask not only what the
impact is on the side of real incomes, but also on the basket of goods that they consume.
NAFTA's liberalization of trade in corn in Mexico might have led to lower prices that
negatively impact the income of poor corn producers. But we must also take into account
the fact that the cost of maize, a key element in the consumption basket of the poor fell,
and hence the CPI of the poor fell relative to that of the well-off which, as shown above,
is what the national CPI measures. The poor are in fact better off than using the national
CPI would suggest.
II.4 Inequality and mobility
So far we have been talking about contemporaneous measures of income inequality. Yet,
it could be argued that measures of income inequality (such as the gini coefficient)
provide a very limited picture of the fairness of the income distribution. For example,
income inequality can measure differences in opportunities (undesired from a social point
of view) but also rewards to differences in effort or risk taking aversion by the different
members of society (desired from a social point of view). Moreover, high inequality
combined with equality of opportunities can be good for growth, because it would
provide individuals with the incentives to put effort, be innovative, and take risks, all
elements conductive to faster growth. On the contrary high inequality with low mobility
will provide few incentives to work. If you are born poor (rich) and you have few
chances to escape poverty (become poor) then you will find few reasons to work hard and
take risks. In other words, standard income inequality indicators provide just a snapshot
at a point in time and do not consider lifetime dynamics (i.e. in absence of additional
11 Go�i, Lopez, and Serven (2005) analysis covers 9 spells and find that there is only one example where
prices exert a negative contribution on nominal inequality.
8
information it may be difficult to reach any conclusion regarding the desirability of
attacking income inequality from the point of view of growth).
These measurement problems are illustrated in figure 5. In both panels we have ordered a
hypothetical population by per capita income in two periods of time (t1 and t2), and the
arrows follows the individual across time. Also, in both cases, the dispersion of income
(i.e. income inequality) is the same in time t1 and time t2. That is, the Gini coefficient
would remain unchanged between t1 and t2. Yet, the picture that emerges from Panels A
and B is completely different. Panel A indicates that those at the top (bottom) of the
ladder in period t1 are also at the top (bottom) in period t2. On the contrary, Panel B
suggests significant income mobility.
Figure 5. Inequality vs. mobility
Panel A. Inequality w/o mobility Panel B. Inequality with mobility
t1 t2 t1 t2
x x x x
x x x x
x x x x
income x x x x
x x x x
capiatreP x x x x
x x x x
x x x x
What do we know about social mobility/equity of opportunities in Latin America? There
are very few studies on the topic because of data limitations: a study on mobility would
require a panel that follows a number of households over a long period of time. Yet, some
authors have made an attempt at measuring social mobility using alternative indicators.
Panel A of figure 6 plots Andersen's index of social mobility (see Andersen, 2001) for a
number of Latin American countries. Briefly, this index tries to determine the importance
of family background in the schooling gap, defined as the disparity between the years of
education that a children in the household would have completed if she/he entered school
at normal school starting age (6 or 7 depending on the countries) and advanced one grade
each year, and the actual years of education. In other words, the schooling gap measures
the number of years of missing education and can be taken as a simple indicator of future
opportunities. When household factors are very important in determining educational
gaps Andersen index will approach zero. When on the contrary household factors play a
limited role, then the index will approach 1.
Panel A of Figure 6 suggests that Chile, Argentina, Uruguay, Peru, and Mexico are
countries with relatively high social mobility. At the other extreme, we have Guatemala,
Brazil, and Ecuador which would have (also in relative terms) low social mobility.
Although the work of Andersen does not allow for a comparison of social mobility in
Latin America and in the developed countries, the existing evidence indicates that this is
lower in the region.
9
For example, Panel B of figure 6 plots the correlation coefficient between parents' and
children's schooling tabulated by Behrman, Birdsall, and Szekely (1999). It indicates that
these correlations are much higher in Latin America (between .4 and .6) than in the US
(.2). Similarly, Panel C of the same figure reports the elasticity of children's income
relative to their father computed by Grawe (2002) for a handful of countries reveling that
for the countries for which we have information, this elasticity is larger among the Latin
Americans than among the Europeans or the US.
Figure 6. Mobility and inequality in Latin America
Panel A. Mobility in Latin America Panel B. Educational Mobility
CHL
ARG USA
URY PRY
PAN
PER URY
MEX JAM
PRY CHL
PAN VEN
DOM
VEN PER
DO HND
SLV COL
HND CRI
BOL
COL ARG
CRI MEX
NIC ECU
ECU BRA
GTM
BOL NIC
BRA SLV
GTM
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
0.70 0.75 0.80 0.85 0.90 0.95
Education: Intergenerational correlation
Social mobility index
Panel C. Elasticity of son's income to father's Panel D. Mobility and income inequality
0.95
1.2 y = -0.4198x + 1.0514
1 x
0.90
0.8 Indey
0.6 lit
bi 0.85
0.4
0.2 Molaic 0.80
So
0
r
do ur l n y
Pe UK US
Nepa staik daa an 0.75
rm
Ecua Malaysia Pa Can 0.4 0.45 0.5 0.55 0.6 0.65
Ge Gini coefficient
Source: Panel A is from Andersen (2001). Panel B is from Behrman, Gaviria and Szekely (2001). Panel C
is from Grawe (2002). Panel D is own calculations.
One key question is whether there is any relation between this mobility index and the
level of income inequality. In fact, there is evidence (see Aaberge et al, 2004) indicating
that high income inequality tends to be associated with low income mobility, something
that in turn would contribute to low social mobility and less equality of opportunities. We
explore this element in Panel D of Figure 6.12 It suggests that indeed there is a marked
12 Andersen (2001) presents a similar plot but finds a very modest correlation (-.12). The main difference
between his and our calculations is that he uses adjusted gini indices to correct for lack of comparability,
whereas ours are computed using comparable data and hence are not subject to adjustment.
10
correlation between inequality and mobility (i.e. more income inequality is correlated
with less social mobility). In other words, while in principle there may be examples of
countries with relatively high social mobility and high income inequality (e.g. Chile), on
the whole it seems that lack of mobility and inequality tend to move together.
II.5 Inequality and demographics: Could things become worse before getting better?
The introduction of mobility and time horizons in the analysis of income inequality
enriches significantly the analysis. For example, as argued by Deaton and Paxton (1994)
the observed cross sectional inequality in a given period of time, is just the average
distribution of the income of successive cohorts, which in principle should show very
different distributions of income. For example, if per capita income y evolves according
to the following simple law of motion:
yt = yt + vit
-1 (1)
where v is a random shock independent of yt-1 that has variance , then if follows that
2
the variance of the income of a person of age T is given by T . In other words, 2
inequality should increase with the age of the cohort. As a result, countries with older
populations will have a tendency to have higher levels of income inequality.
Figure 7. Inequality and demographics
Panel A. S.D. of incomes by age cohort Panel B. Inequality vs. age of population
Costa Rica 2004
Costa Rica 2004 20
y = -0.2687x - 0.0331
1.1
15
1
de 10
0.9 ustj 5
Ad
0.8 0
-15 -10 -5 0 5 10
-5
0.7
-10
0.6
20s 30s 40s 50s >60 -15
Adjusted Share of Young
Source: Perry et al. (2006).
Panel A of Figure 7 lends support to the hypothesis that inequality increases with the age
of the cohort. More specifically, this panel presents the standard deviation for five Costa
Rican population cohorts in 2004. Inspection of this figure indicates that the standard
deviation of income increases steadily with age and the cohort of those in their 60s have a
standard deviation of income that is almost twice as large as that of the cohort in their
20s. Moreover, these effects are not small.
11
More generally, Panel B of Figure 7 presents the cross national partial correlation13 of the
Gini coefficient and share of people below age 14. This panel indicates that the
correlation between these two variables is negative (and significantly different from
zero). That is, younger societies have lower Gini coefficients.
What does this mean in practice? Well, if Latin America had the demographic structure
of aging Europe, its Gini coefficients could be higher by 4 percentage points. Moreover,
in some of the comparatively young countries such as Bolivia, Guatemala or Honduras
the Gini coefficient could be up to 7 percentage points higher. Thus everything else equal
and to the extent that the region starts aging, it is possible that income inequality levels
become worse before getting better.
II. 6 Beyond income inequality
Income measures are highly correlated with most aspects of household welfare. However,
it must be recognized that in some important dimensions of welfare, such as health and
life expectancy, there has been significant progress and convergence among income
groups. This is illustrated in figure 8 where we plot the distribution of income (Panel A)
and the distribution of life expectancy (Panel B) across Brazilian municipalities in 1970
and 2000. Inspection of this figure indicates that whereas on the income front there has
been some increased dispersion and the emergence of a bimodal distribution, this has not
been the case with life expectancy, which if anything has experienced a decline in
dispersion (i.e. a decline in inequality). In other words, regional distribution trends in
some welfare indicators may be improving, and this would suggest an important role for
some policies that fight poverty independently of those aimed to growth.
Figure 8. The distribution of municipal incomes and life expectancy in Brazil
Panel A. Income Panel B. Life expectancy
0.45 0.40
0.40 Distribution 1970 0.35
0.35 0.30 Distribution 2000
0.30 Distribution 2000 0.25Distribution 1970
0.25
0.20
0.20
0.15 0.15
0.10 0.10
0.05 0.05
0.00 0.00
-4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 -1 0 1 2 3 4
Country Relative Income Country Relative Life
Source: Perry et al. (2006)
13We control for level of development.
12
III. Why is inequality so high in Latin America?
III.1 Asset inequality in Latin America
The existing differences in terms of development and more concretely in inequality
between the region and the developed world did not appear overnight. In fact, they are
likely to be the result of historical processes that in go back at least to the colonial period.
Several authors (see among others Engerman and Sokoloff, 2000 and Acemoglu, Johnson
and Robinson, 2001) have argued that in order to understand the high inequality levels
observed in Latin America today it is important to understand the institutional framework
created by the colonial powers which allowed a small group of elites to protect the large
rents they were enjoying and excluded most of the population from access to land,
education and political power.
Figure 9 Composition of the population in the New World
100
80
60
40
20
0
1570 1650 1825 1935 1570 1650 1825 1935 1570 1650 1825 1935
Spanish America Brazil US & Canada
White Black Indigenous
Source: Engerman y Sokoloff (1997)
Sokoloff and Robinson (2005) synthesize much of this literature. They argue that the
combination of high settler mortality rates and the availability of rich factor endowments
(mineral riches and indigenous labor in Mexico, Mesoamerica and the Andes; and land
adequate for sugar plantations in Brazil, the Caribbean and the South of the US14, coupled
with the availability of slave labor imports) determined a colonization strategy that led to
highly exclusionary institutions in much of Latin America and the Caribbean, where a
large fraction of the population (of indigenous or African origin) remained for a long time
excluded from access to land, education and political power. In contrast, the lack of such
initial endowments and lower settler mortality rates led in the north of North America to
the progressive dominance of large settlements of Europeans (see Figure 9) and to the
establishment of more inclusive institutions.
14The South of the US evolved similar institutions than most of the Caribbean Bassin, but there was rapid
convergence with the North after the South lost the civil war
13
These authors show the high persistence of such institutional traits. As a consequence,
higher inequality of land holdings, access to education and political power characterized
most of Latin America and the Caribbean through the nineteen and early twentieth
century, as indicated in Table 1. As this Table shows, both levels of education and
indicators of political participation evolved much slower in those countries in which large
indigenous populations and mineral riches existed at the time of European colonization,
or in which plantations based on slave labor could evolve, than on those countries in
which these conditions were not present and European mortality rates were lower
(Argentina, Chile 15 and specially, the US).
Table 1. Education and Voice (Latin America and the US)
Literacy rates (%) Political institutions: Lack of secrecy in balloting
1860/1870 1890/1900 1920/1925 1940/1950 1840-80 1881-1920 1921-40
Bolivia 17
Colombia 32 62 No No
Ecuador Yes No No
Guatemala 11.3 15 20
M�xico 22 36 48.4 Yes No No
Per� 38 Yes No
Brazil 15.8 14.8 - 25.6 30 57 Yes Yes
Jamaica 16.3 32 67.9
Argentina 23.8 45.6 � 52.0 73 1896: Yes/ 1916: No No
Chile 18 � 25.7 30.3 - 43.0 66 76 No No No
Costa Rica 23.6 � 33.0 64 Yes Yes No
Canad� 82.5 1897: Yes/ 1878: No No No
E.U Blancos Norte 96.9
E.U Total 80 86.7 92.3 No No No
Source: Engerman and Sokoloff (2002)
Still today, asset inequality plays an important role in the persistence of high income
inequality. In fact, whereas the gini coefficient for operational holdings of agricultural
land is estimated at .81 for Latin America (see Deininger and Olinto, 2000), in other
regions it tends to hover around .60.16 More importantly, whereas the gini coefficient of
the distribution of years of education in Latin America is around .42, in the developed
countries it is closer to .27. See Panel A of Figure 10.
This high inequality in educational levels is particularly dramatic in Latin America
because of the low intergenerational educational mobility illustrated in Section II 5.
Indeed, differences in educational achievement among children from higher and lower
income percentiles in the region are striking, as illustrated in Panel B of Figure 10 for
Argentina, a country where the ample availability of schools and teachers do not appear
to restrict access to kids from lower social strata.
In turn, differences in education are today the most important predictor of differences in
income levels among households in Latin American countries (see Perry et al, 2006,
Chapter 8). Thus low educational mobility constitutes a major channel of reproduction of
15Also Costa Rica
16After Latin America the region with the highest land inequality would be the Middle East and North
Africa with a Gini of .67.
14
high income inequality, leading to the low intergenerational social mobility shown in
Section II.5. It should not come as a surprise, then, that we find a significant correlation
among educational and income Gini's in the region.
Figure 10: Educational Gini in LAC countries vs. others
Panel A. Gini coefficient of years of education
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
LAC Africa Asia Eastern Developed
Europe countries
Panel B. Years of education: rich vs. poor
Argentina
45%
40% 30% poorest
35% 30% richest
30%
25%
20%
15%
10%
5%
0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18+
years of education
Panel C: educational and income gini correlation
60 Brazil
Bolivia
Paraguay
Colombia
58 Chile Panam�
Honduras Guatemala
56
54 El Salvador
niige 52 Argentina Peru
moc 50 Costa RicaMexico
Venezuela
48
In
46 Uruguay Nicaragua
Jamaica
44
42
40
10 20 30 40 50 60
Education gini
Source: Panel A. De Ferranti et al. (2004). Panel B, Perry et al. (2006).
15
The differences in educational attainment between the poor and the rich lead to much
higher differences in incomes due to the convexity of returns to education, as illustrated
in Figure 11. This Figure shows, indeed, that the returns to education increase
significantly after finishing secondary school �a condition rarely met by children from
lower quintile households. This fact not only helps explain the strong persistence of high
income inequality given the low observed educational mobility, but actually �in
conjunction with the presence of credit constraints- contributes to explain the low
educational mobility itself. Poor parents, who face the need to keep children in school for
an extended period of time to actually reap the benefits of their investment in education,
also face very high opportunity costs from income forgone from children work, specially
during periods of adverse income shocks. The high desertion rates observed among
children from poor households should hence not come as a surprise. Actually, this is the
main economic rationale of the Conditional Cash Transfers programs that have become
so popular in many countries of Latin America in recent years. By lifting the credit
constraint, such transfers help effectively reduce the opportunity cost of keeping children
in school for poor households that enter the program.
To complicate matters further, Figure 11 shows that educational returns for the children
of the poor are lower than for children of the rich. This maybe due to a variety of factors:
lower quality of schools, lower availability of assets that are complementary to education
in income generation (land, public infrastructure, credit), discrimination in labor markets
and unobservable factors (linked to differential access to pre schooling facilities, in
nutrition levels, etc). Such differential (and probably more uncertain) returns also help
explain the high difference in educational outcomes among children of the poor and the
rich, adding to the complexity of the problem and of its solutions. Of course, in many
countries in the region there are in addition severe limitations in the supply of schools
and teachers in poor neighborhoods, especially in rural areas.
Figure 11. Nicaragua. Returns to education by level.
1.7
yra 1.5
imrp 1.3 20% poorest 20% richest
telepmoc 1.1
0.9
in 0.7
toe 0.5
tivlaerega 0.3
0.1
Complete Some Complete Some Complete
W
primary secondary Secondary university university
Source: Perry et al. (2006)
16
The effect of the high degree of persistence of educational differences on low social
mobility is further exacerbated by the high degree of matching among educational groups
in the formation of new households. For example, Figure 12 plots the Gini coefficient
against marital sorting coefficients (defined as Pearson correlation coefficients for years
of schooling between husbands and wives). Two basic messages emerge from this figure.
First, there is a strong relationship between the two variables. In fact, the correlation
coefficient between marital sorting and the Gini coefficient is above .6. The second
message is that the marital sorting coefficients in Latin America are unusually high (at
least relative to those in the rest of the world), something that can be taken as a symptom
of a severe social stratification problem that not only further concentrates household
incomes but reinforces the observed low social mobility.
Figure 12. Gini coefficients and marital sorting by educational levels.
Source: De Ferranti et al. (2004).
III.2 Fiscal policy and income inequality
Though such differences in educational levels and returns are a major channel of the
observed persistence of high income inequality levels, it is by no means the only one.
Differences in access to other complementary assets may partially explain differences in
returns to education among children from households located in high and low income
quintiles. Indeed, previous studies have shown that the persistence of significant
differences in "bundles" of assets (education, public infrastructure, institutions) across
regions within countries explain the persistence (and frequent divergence) of high spatial
income per capita disparities 17 . Indeed, "location" appears as a major predictor of
household income differences, together with education.
17"Beyond the City: the Rural Contribution to development", WB 2005, and "Poverty Reduction and
Growth: Virtuous and Vicious Circles", WB 2006
17
All these said, high asset inequality does not have to necessarily translate into high
disposable income inequality unless taxes and transfers do not have significant corrective
effects. In this regard, it may worth looking at the role played by the government in Latin
America and compare it with some countries like the Europeans that are well known for
having inequality as a policy concern.
Figure 13. Disposable and Market income in Latin America and Europe
Disposable Income
Panel A. Latin America Panel B. Europe
0.65 0.65
0.55 0.55
0.45 0.45
0.35 0.35
0.25 0.25
0.15 0.15
il 6 ai d
str anl any ce ndla yalIt ain 15
ourg UK
ntinae az eli C Sp ende O
Br Ch iabmlo ru
Pe miulg France eerG Ire algutr
Mexico LA Au Be Fin
Denmark Germ ndslaerh Po Sw
Arg Co EUR
Luxemb Net
Market Income
Panel C. Latin America Panel D. Europe
0.65 0.65
0.55 0.55
0.45 0.45
0.35 0.35
0.25 0.25
0.15 0.15
ani eil y e
ia 6 muilg karmn dnaln cena eec alyIt 15
ugal
Fr Spain eden UK O
gentrA Brazil Ch mblo Peru Gr andelrI
Mexico LAC Austria Be Fi
De German mbourg sdnlareth Port Sw
Co EUR
Luxe Ne
Source: Go�i, Lopez, and Serven (2008)
In a recent paper, Go�i, Lopez and Serven (2008), elaborating on a topic highlighted in
Perry et al (2006), argued that whereas in Latin America the distribution of market
income (i.e. the income before taxes and government transfers and thus a measure that is
largely determined by market rewards to the private assets and efforts of individuals, and
by the underlying distribution of those private assets) and disposable income (i.e. the
income after government cash benefits such as pensions, unemployment insurance, and
social assistance transfers have been received and direct taxes have been paid) are very
similar, in Europe this is not the case. Figure 13 (taken from Go�i, Lopez and Serven,
2006) reports the value of the gini coefficient of the distributions of disposable income
(Panels A and B) and market income (Panels C and D) in Latin America (Panels A and
18
C) and Europe (Panels B and D).18 Panels A and B indicate that Latin America is much
more unequal than Europe. In fact, the Latin American country with the lowest Gini
coefficient in this sample has inequality levels above those of the most unequal European
country (Portugal).19
The situation, however, changes significantly when we look at Panels C and D, which
show the gini coefficients of the distribution of market incomes: whereas the average gini
coefficient of market income for the Latin America sample, at .52,20 is only 2 percentage
points above that of disposable income, the Gini coefficients of the European countries
are substantially higher than those in Panel B: the average for the 15 countries in the
sample is now .46. That is, it appears that most of the difference between the two regions
in the levels of disposable income inequality are due to the different impact of taxes and
transfers: they reduce market income inequality considerably in Europe, and very little in
Latin America.
A second issue of interest in this context is whether the observed redistribution in Europe
operates through the impact of taxes or transfers. This is addressed in Figure 14 which
separates the respective impacts of cash transfers and direct taxes. Panel B shows the
difference between gross income (market income plus transfers) and market income (i.e.
panel B shows the distributional impact of transfers) in Denmark, Finland, Ireland, the
UK, and Sweden and indicates that public transfers lower the Gini coefficient of market
incomes by 12-14 percentage points, and by 10-11 percentage points in Belgium,
Germany, and Luxemburg. At the other extreme, in Portugal transfers lower the Gini
coefficient by just 6 percentage points. The average contribution of transfers for the
European sample is around 10 percentage points. In contrast, public transfers contribute
only slightly to lower inequality in Latin America, lowering the Gini coefficient by
between 1 and 2 percentage points, although in some cases (Peru) the distribution of
income is even more unequal after transfers than before transfers.
As for the distributional impact taxes, panels C and D of Figure 14 show the difference in
the gini coefficient of gross income and disposable income and indicate that the contrast
between the two regions is less dramatic. Like with transfers, taxes reduce the levels of
income inequality much more in European countries than in Latin America. For example,
direct taxation lowers the Gini coefficient of household income by 6-7 percentage points
in Austria, Belgium, and Luxemburg, and by an average 5 percentage points for the
fifteen countries in the European sample. In contrast, the average decline in the Latin
American Gini coefficients as a result of direct taxes is about 1 percentage point, with
very little variation across countries. Thus on the whole for the European countries,
transfers play a more significant role than taxes: of the 15 percentage points difference
18Due to differences in the surveys used, the gini coefficients in figure 6 do not correspond exactly to those
in figure 1.
19This result continues to hold in a broader sample of Latin American countries, because the lowest Gini in
the region (Trinidad and Tobago's) is 42.
20 Somewhat surprisingly the average gini in this Latin American sample and in the broader sample in
figure 1 is the same.
19
between the average Gini coefficients of market and disposable income across European
countries, about two-thirds (10 percentage points) are due to transfers.
Figure 14. The role of taxes and transfers in Europe and Latin
America
Difference between Gini coefficients of gross income and market income
Panel A. Latin America Panel B. Europe
0.00 0.00
-0.04 -0.04
-0.08 -0.08
-0.12 -0.12
-0.16 -0.16
a ia l
tin azilrB co 6 dn e ce dn ne 51
Chile mblo Peru miug kra anc ee ain
ed UK
Mexi LAC Austria nm Sp
Bel Finla Fr Gr elaIr Italy bour OR
Argen Co De Germany rtugaoP Sw
uxemL EU
Netherlands
Difference between Gini coefficients of disposable income and gross income
Panel C. Latin America Panel D. Europe
0.00 0.00
-0.04 -0.04
-0.08 -0.08
-0.12 -0.12
-0.16 -0.16
ani il
az eil ia o 6 d e s l 51
Ch xic Peru anl ec
an any eec dna yalIt UK
gentrA Br mblo O
Me LAC miulg
Austria Fin Fr mbour Spain
Co Be Gr Irel
Denmark Germ ugatroP edenwS
EUR
Luxe Netherland
Source: Go�i, Lopez, and Serven (2008)
Such low levels of income redistribution through the State may be a reflection of high
levels of State capture, and this in turn a reflection of high inequality levels as discussed
below. But the fact that Europe was able to "break" with a history of high inequality
during the twentieth century as indicated in Figure 15 below, suggest that such an event is
not an impossibility going forward in Latin America.
20
Figure 15. Historical inequality trends (UK, France, Spain)
0.2
0.15
1%pote 0.1
omcnI0.05
0
1910 1920 1930 1940 1950 1960 1970 1980 1990
UK FR SP
Source: Perry et al. (2006)
IV. Why is reducing inequality important?
As noted in the introduction there are several reasons why policy makers should be
concerned with such levels of high inequality, including that high inequality appears to
lower long-run growth. In this section we review three of them.
IV.1 Inequality and crime
First, there is now significant evidence that high crime and violence levels lower growth
prospects (see Alessina and Perotti, 1996 for macro evidence of how crime levels lower
growth) More recently, Alaimo et al. 2007 have presented micro evidence based on
10,000+ Latin American firms suggesting a negative impact of crime on firm
productivity. In turn, as shown by Fajnzylber, Lederman and Loayza (1998) income
inequality appears to be a major determinant of crime and violence levels (see Panel A,
Figure 16 for the correlation between the Gini coefficient and the % of firms that find
crime to be a major constraint to growth and Panel B for the correlation between income
Ginis and homicide rates. Nor surprisingly then, the percentage of firms that find crime
and violence to be a major constraint to growth in Latin America is much larger than in
other regions (Panel C, Figure 16).
21
Figure 16: Crime and Violence as a constraint to growth
Panel A. Inequality and crime as a barrier to growth
)
(% 90
80
70
growth 60
to
50
rrieraba 40
30
ase 20
10
m
0
Cri
25 30 35 40 45 50 55 60 65
Gini index (%)
Panel B. Inequality and crime
00.0 100
100r 80
pe 60
rate 40
20
micidesoH 0
40 45 50 55 60
Gini index (%)
Panel C. Crime as a barrier to growth
30
osknihtohws 25
20
15
mrif 10
of 5
%
0
LAC EAP SSA MENA ECA
Note: Crime as a barrier to growth indicates the % of firms that find crime
and violence to be a major constraint to growth in each country.
Source: World Bank Enterprise surveys
22
IV.2 Inequality, poverty and growth
Second, it has been argued that inequality, coupled with credit constraints, reduce
physical and human capital accumulation (see, for example, Galor and Zeira, 1993). As
shown by Perry et al 2006 and Lopez and Serven, 2006, this is really an argument about
why poverty (rather than inequality) can be a drag on growth. Lopez and Serven, 2006
show, indeed, that when poverty is included in cross country panel regressions as an
additional independent variable it has a significant negative effect on growth and the
negative effect of inequality losses its statistical significance. Further, they show that, as
proposed by theory, such an effect takes place mostly through a negative effect on
investment in countries with low financial deepening (see Table 2). As high inequality
leads to higher poverty levels �for the same level of average income per capita-, this
should be seen as an indirect negative effect of high inequality on growth.
Table 2: The impact of poverty on investment
Model 1 Model 2
GFCF GCF GFCF GCF
(1) (2) (3) (4)
Investment (t-1) 0.658 0.652 0.721 0.653
t-stat 11.15 19.05 16.36 24.34
Income (in logs) (t-1) -0.009 -0.012 -0.005 -0.005
t-stat -1.58 -2.29 -1.55 -1.61
Growth (t) 0.539 0.550 0.524 0.620
t-stat 8.87 9.28 14.59 14.39
PPP (t-1) -0.010 -0.014 -0.004 0.000
t-stat -1.66 -1.84 -0.81 -0.06
Terms of Trade (t) 0.064 0.132 0.079 0.071
t-stat 1.60 3.02 3.97 3.02
P0 ($2) (t-1) -0.079 -0.105
t-stat -1.88 -2.74
P0 HFD ($2) (t-1) 0.031 0.016
t-stat 0.90 0.47
P0 LFD ($2) (t-1) -0.055 -0.057
t-stat -2.03 -2.52
# Observations 338 345 308 311
# Countries 108 108 103 103
Hansen Test p-value 0.29 0.34 0.47 0.28
AR(2) p-value 0.33 0.30 0.36 0.43
Notes: The table reports regression results with investment (i.e., gross fixed capital formation �GFCF- or
gross capital formation �GCF-) as dependent variable; and lagged investment, the lagged per capita income
(in logs), the income growth rate, a lagged measure of market distortion (given by the price of investment
goods), the terms of trade, and headcount poverty (US$2 poverty line). The models in columns (3) and (4)
use the same controls but separate the poverty data according to whether the country in question has a high
23
level of financial deepening (above the median sample) or not. All regressions include a constant. The
regressions are calculated using system GMM estimators and allowing the instrument set to start with
lagged levels at t-1. Robust t-statistics are reported below the coefficients.
One additional potential channel through which poverty can negatively affect growth is
through education. As noted in the previous section, the existing microeconomic
evidence suggests that poor people have less incentives to get educated than richer
people. Yet, to the best of our knowledge this hypothesis has not been tested in a cross
national context. We now explore it on the basis of the following empirical model
Educit = Educit + Xit + pit +i +it
-1 (2)
where Educ is the secondary net enrollment rate, X is a set of control variables to be
discussed shortly, p is the poverty headcount (using a poverty line of US$2 a day) i is a
country-specific effect, andit is an i.i.d error term. Our parameter of interest in (2) is .
If poverty acts as a barrier to increases in human capital, then we would expect to find
<0.
As for the control set in X we include variables that capture both the availability of school
resources and family factors.21
To measure availability of resources we rely on two variables. One is the share of GDP
spent by the public sector on education and (logged) per capita GDP. We would expect
that both variables carry a positive sign in the education equation: countries that spent
more could be expected to have higher enrollment rates.
To measure family factors we use the fertility rate and the infant mortality rate. The first
of these variables is a proxy for the average number of children in a household and hence
for the individual time that parents can dedicate to each children. That is, implicitly we
are assuming that there is a trade-off between child quantity and child quality. For
example, both Leibowtz (1974) and Hanushek (1992) find that children's educational
attainment and family size are negatively correlated. The second of our variables, the
infant mortality rate, would aim at capturing health status in the early years of a
children's life, a factor that has been found to also correlate with academic achievement
(Glewwe, Jacoby and King, 2001). Thus we would expect both the fertility rate and the
infant mortality rate to carry a negative sign.
Table 3 presents the results for different specifications depending on the different
controls included. First thing to note is that the Sargan test of overidentifying restrictions
and the test for second order serial correlation do not indicate any problem with the
selected specifications. As for the parameters of the control variables, in general they
have the expected sign. The only exception is column (6) where per capita income
appears with a negative parameter (although not statistically significant at the 10
21There are several studies that argue that family background and socioeconomic factors are more
important determinants of student achievement that school resources.
24
percent). In all the other cases, the fertility rate and the infant mortality rates have
negative parameters whereas public spending and income positive.
Moving now to assess the role of poverty, table 3 indicates that poverty levels indeed
reduce enrolment rates. Depending on the specification being considered we find that a 1
percentage change in the poverty rate would translate into a decline in secondary
enrolment rates ranging between .5 and 1.5 percentage points. That is, beyond being
statistically significant, the magnitude of these estimates indicates that they are also
economically significant. Thus this analysis indicates an additional channel through
which poverty can affect growth.
Table 3: The impact of poverty on education
Dependent variable is the secondary_net_enrollment rate
(1) (2) (3) (4) (5) (6) (7)
Persistence 1.004 0.560 0.383 0.435 0.533 0.524 0.332
(lagged dependent) 13.67 14.20 10.22 15.28 4.86 22.15 7.93
Public spending on education 3.495 2.974 3.212 2.883 2.853 3.912 2.217
(% of GDP) 2.31 4.98 10.51 13.68 4.71 10.38 7.17
Fertility rate -5.562 -3.014 -6.622 -2.431
(births per woman) -7.93 -3.33 -8.38 -1.67
Infant Mortality -0.169 -0.214
(per thousand live births) -3.37 -2.46
Income 1.881 -3.878 3.557
(log GDP per capita) 7.51 -1.60 2.29
Poverty -1.467 -1.121 -0.602 -1.392 -0.878 -0.521
(headcount) -8.33 -8.37 -3.61 -8.27 -6.35 -2.35
Observations 163 105 105 99 105 105 99
Countries 73 55 55 53 55 55 53
Sargan test. p-val 0.18 0.24 0.35 0.78 0.20 0.67 0.98
Second order correlation. p-val 0.58 0.25 0.13 0.12 0.26 0.13 1.40
Note: The table reports the results of regressing the secondary net enrolment rates on the variables in the
first column. The data is in non-overlapping 5-year averages. Estimation method is GMM and the
instrument set is the same in all the specifications and includes the lagged values of the dependent
varaibles. The null hypotheses in the Sargan test of overidentifying restrictions and the test for second order
serial correlation are (i) no correlation between the residuals and the instrument set and (ii) no second order
correlation in the residuals (i.e. a large p-value indicates that there is no evidence against the null
hypothesis).
IV.3 Inequality and output volatility
In the remainder of this section we focus on a third rather unexplored channel: on how
inequality can raise output volatility, which in turn is known to reduce growth. In fact, the
last few years have witnessed a renewed interest in the relationship between
macroeconomic volatility and economic growth (among others see Ramey and Ramey,
1995; Martin and Rogers, 2000; Fatas, 2002; Wolf, 2003; Hnatkovska and Loayza,
2004). To a large extent the main conclusion of these papers is that volatility and long-
run economic growth tend to be negatively related.
25
One possible explanation behind this finding is that if volatility is viewed as a measure of
risk, then other things being equal, countries with higher volatility will have a tendency to
under-invest or undertake inefficient investment projects (see Bertola and Caballero,
1994) and therefore grow less. A similar argument is made by Krebs, Krishna and
Maloney (2005) focusing on the impact of risk on education (rather than on physical
capital) and of education on growth. Another potential explanation is that there may be
asymmetries in the process of capital or knowledge accumulation. If the negative effects
of recessions on learning by doing are larger than the positive effects of expansions, then
we would also expect that high volatility leads to lower growth (Martin and Rogers,
1997). Similarly, if firm entry and exit rates differ dramatically in good and bad times
along the business cycle, then volatility would also lead to lower investment and growth.
This can be the case if for example there are important fixed costs (such as establishing
an important sales network) associated to market entry.
Thus a natural question that arises is whether inequality contributes to income volatility.
If so, this would be uncovering another potential channel by which inequality may have a
negative impact on economic growth. Moreover, recent work by Calderon and Levi-
Yeyati (2007) suggests that periods of economic turmoil tend to be associated with
deteriorations in the income distribution. In other words, it could that higher volatility
leads to higher inequality so that there is a potential for a vicious circle in which high
inequality and high volatility reinforce each other.
Figure 17. Volatility and inequality
20
ht 15
ow
gr 10
of.d.s 5
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Gini
Figure 17 presents the scatter plot of output volatility and inequality based on panel data
for 118 countries spanning 1960-2000. Each pair represents the volatility of growth over
half a decade and the gini coefficient at the beginning of the period. The figure shows
that there is significant dispersion around the regression line, but also a positive slope
which is significant at standard levels (4.1 with a robust s.e. of 1.6). The existence of a
positive and significant correlation between the standard deviation of output and income
26
inequality gini coefficients is robust to controlling for the many apparent outliers in the
sample.22
Which reasons could be behind inequality leading to higher volatility? On the one hand,
there is the possibility that more unequal societies are not able to isolate themselves from
external shocks as more equal ones. For example, Rodrik (1998) argues that when there is
an external shock (say a sharp decline in the price of a commodity which to simplify is
the country's main export) the policy response (e.g. devaluation and fiscal retrenchment)
would be much more difficult to implement in a country with potential for social conflict
because of the distributional implications of the policies. Thus countries with high
inequality could in principle suffer more severe consequences form external shocks.
This channel could be particularly important in the Latin American context, not only
because of the high inequality levels in the region but also because Latin America has
been traditionally subject to large external disturbances from world goods and financial
markets. For example, according to de Ferranti et al.(2000), over the past two decades
Latin America has suffered terms of trade disturbances that were much larger than those
affecting industrial economies and the East Asian countries, and on par with the rest of
the developing world.
On the other hand, inequality could negatively affect the quality and volatility of policies
if policy makers reflect the view of the median voter and she/he feels that has more to
gain from policies that aim at solely redistribution even if those have a negative impact
on growth. For example, this is the theoretical result obtained by Alesina and Rodrik
(1994) who assumed that taxes are proportional to income, public spending is equally
distributed among individuals, and the median voter rules. Taking this view to the
extreme, one can imagine a situation where a populist government applies policies based
more on their acceptance by the median voter than on whether they are more or less
appropriate in a particular context (i.e. where the government makes policy mistakes
because of popular pressure).
Clearly, these are just hypotheses, which to the best of our knowledge have not been
explored empirically23. We next explore what the data have to say. To explore the role
played by inequality in the transmission of international shocks, our empirical strategy is
based on the following econometric specification:
it =it + it +itgit + Xit +i +it
-1 (3)
where is the standard deviation of GDP growth over a 5 year period, is a variable
that captures the magnitude of external shocks (such as the standard deviation of the
trading partners growth rate or the standard deviation of the terms of trade), g is the gini
22A regression estimate of the slope coefficient robust to the presence of outliers be 2.6 with an associated
s.e. of .86 (i.e. it continues to be significant).
23Rodrik (1998) explored whether external shocks affect growth more in high inequality economies but we
do not know of any study that has explored the impact on inequality on output volatitility.
27
coefficient at the beginning of the period, X is a set of control variables, i is a country-
specific effect, andit is an i.i.d error term. According to (3), output volatility depends on
past volatility, external conditions, inequality and a set of control variables to capture the
economic environment. Among these controls, we include inflation, government
spending, openness to trade, and a measure of financial deepening. We also explore the
role played by a measure of fiscal policy volatility (denoted F) and a measure of
monetary policy volatility (denoted M). As noted by Fatas and Mihov (2007) policy
volatility is probably a better indicator of macroeconomic policy than standard measures
reflecting levels of policy instruments. That is, one could view F and M as measures of
policy "quality".
Briefly, to compute F we first filter government consumption growth from business
cycle fluctuations by projecting that variable over GDP growth and a constant and then
proceed to compute the standard deviation of the residual over non-overlapping 5 year
periods. Similarly, we compute M on the basis of the residual of a regression24 of money
growth on inflation and GDP growth and a constant.25 That is, both F and M are
unrelated to the economic cycle.
In (3) our primary focus is the estimate of . If inequality leads to higher domestic
volatility when there is an external shock then we should find that > 0. Note that in this
model the relationship between output volatility and the external shock is given by:
it (4)
it = +git .
Thus if > 0, domestic output volatility would increase with the level of inequality. As
for the relationship between output volatility and inequality, it follows from (3) that:
it (5)
git = it ,
which to the extent that cannot take negative values will also be positive when > 0.
Testing the second hypothesis is much more challenging because the literature (at least
the empirical literature) on the determinants of policy quality is much scarcer. One could
think of a framework where institutions are behind policy quality and hence policy
volatility, perhaps through constraints on the executive (see Fatas and Mihov 2007), or
elements such central bank independence, but clearly this unchartered territory. Against
this background we rely on a simple econometric model given by:
24Due to the existence of important outliers the estimates are robust to the presence of outliers.
25We are aware that the measure of monetary policy is likely to be more challenging because different
countries rely on different policy instruments (i.e. exchange rates, money, interest rates) and hence any
measure we use will have important limits
28
Pit =Pit + git + Xit +i +it
-1 (6)
where P=F,M, and as above g is the gini coefficient at the beginning of the period, X is a
set of control variables, i is a country-specific effect, andit is an i.i.d error term.
Among the control variables we consider are Mt-1 (Ft-1 )in the regression for F (M ),
inflation, and external shocks. According to (6), policy volatility depends on past policy
volatility (to control for inertia), policy context (i.e. inflation, external conditions), and
inequality. Our primary focus is the estimate of in equation (1). If inequality leads to
higher policy volatility then we should find that > 0. If one the other hand, inequality
does not affect policy volatility, we should find that = 0. It could be argued that the
previous model is ignoring important elements such as institutions. While we do agree
with that point, we would note that to the extent that institutions are more or less
permanent or highly persistent (at least within the horizon of the econometric exercise)
our fixed effects model should be able to account for them.
Given the dynamic nature of equations (3) and (4) and the presence of fixed effects to
account for unobserved country heterogeneity, both equations are estimated using the
system GMM estimator.
Tables 4 and 5 report the results corresponding to equation (3) for two different measures
of the external shock (the s.d. of the trading partners growth rate in table 1 and the s.d. of
terms of trade in table 2) and different specifications depending on the control variables.
Inspection of these tables suggests 26 that indeed our parameter of interest is always
positive and significant at standard levels. That is countries with higher inequality levels
seem to suffer more from external shocks. In a number of cases is negative but when
we solve (4) to find its zero value, the obtained gini coefficient (between .2 and .3) tends
to take values that are below the actual distribution of gini coefficients across countries.27
As for the values taken by parameters corresponding to the policy controls, it has to be
noted that there are important differences depending on whether one considers as our
measure of external shock growth volatility in the trading partners or the volatility in the
changes in terms of trade. In the first case, the controls carry the expected sign and are in
most cases significant, with the exception of the monetary volatility variable. Fiscal
volatility leads to higher output volatility. A similar result is found for inflation and for
openness to trade. On the contrary, financial deepening and a large government seem to
help in mitigating the impact of external shocks. While the results for financial deepening
are encouraging because progress on this are would be win-win in the sense that it would
contribute to lower volatility and to faster growth (see Levine, 1997), the result for the
size of the government appears a bit more problematic because it may somewhat present
a trade-off between volatility (countries with smaller governments appear to be more
26Hansen test of overidentifying restrictions and the test for second order serial correlation do not indicate
any particular problem with the specification of the models.
27We have also explored the extent to which this finding is driven by the fact that more unequal countries
are less financially developed, and thus more exposed to external shocks. However, there is no evidence of
such hypothesis in the data.
29
volatile) and growth (there is plenty of evidence indicating that in general countries with
larger governments tend to grow less; among others see Loayza, Fajnzylber and
Calderon, 2005). When, instead, we look at the results for the specification with the
volatility of the changes in the terms of trade, we find that the only variable that appears
to matter (apart from the external shock) is the volatility of fiscal policy.
Could it be that inequality is correlated with some missing variables that should belong to
the equation and that our results are affected by missing variable bias? To explore this
issue we augment equation (3) with the interaction of our external shock variable and two
potential candidates: the degree of openness of the economy and financial deepening.
This gives rise to four different models depending on the variable used to proxy the shock
and the variable used to augment the equation. Table 6 reports the results in a synthetic
way, when we exclude the direct effect of individual policies -the equivalent of equation
(2i Tables 4 and 5. The main message of this table is that our finding is robust to this
departure from our basic specification. While the interaction of our external shock
variables and the degree of openness and financial deepening are significant and have the
expected signs (positive in the first case and negative in the second one), the interaction
of the external shock variables and inequality continue to be significant. These results
hold when we include the effect of individual policies (not shown here). Thus, inequality
augments the effect of external shocks, even when controlling for the mitigating effect of
financial deepening and the augmenting effect of trade opening.
We now move to explore whether inequality affects the quality of policy. Table 7
indicates that in general we cannot reject the null hypothesis that inequality does not
affect policy volatility. In fact, the null hypothesis never comes close to rejection.
Table 7 also indicates that there is moderate persistency in the volatility of both fiscal and
monetary policy (autoregressive parameters in the .1 to .2 range) and that both external
shocks and inflation appear to play a role in this context contributing to higher
volatility.28
On the whole, from these results one can conclude that indeed inequality contributes to
higher volatility through the external shock transmission channel, but it does not appear
to contribute through a weaker quality of implemented policies. The latter result should
not come as a surprise, given that other predictions of the median voter theory are neither
supported by the data. In fact, we already have shown in Section 3 that, contrary to what
this theory would predict, countries with higher inequality �as those in Latin America- do
not appear to engage in more redistributive policies. We can not rule out, however, that
higher inequality could lead to weaker economic institutions (eg, lower property rights
protection) and through this channel affect negatively long term growth, as proposed by
29
the median voter theory and suggested by recent events in some countries in the region.
28We have also tried with specifications that include public spending but this variable does not seem to
belong to the equation.
29On this point, see Lederman and Perry, forthcoming.
30
Table 4. Inequality and the transmission of international shocks
Dependet variable is s.d. of per capita growth rates
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Persistence 0.382 0.287 0.085 0.214 0.124 0.156 0.081 0.267 0.191 0.303
(lagged dependent) 6.40 7.88 1.61 4.92 2.90 2.06 1.76 6.11 5.33 3.67
External shock 0.318 -1.443 -1.964 -1.099 -0.515 -1.609 -1.325 -1.422 -1.164 -0.702
(s.d. of annual growth of trading partners) 4.19 2.50 3.11 2.29 1.02 2.62 2.08 2.42 2.29 1.04
External shock * inequality 5.538 6.355 3.585 1.937 5.630 4.747 5.245 4.309 4.014
(s.d. of annual growth of trading partners *gini) 3.49 3.65 2.95 1.51 3.42 2.67 3.26 3.07 2.39
Fiscal policy volatility 0.048 0.033
(s.d. of annual growth of gov. consumption) 4.72 3.41
Monetary policy volatility 0.000 0.000
(s.d. of annual growth of corrected money) 0.38 0.15
Price stability 1.027 0.294
(inflation rate) 2.13 1.12
Access to credit -0.747 0.129
(credit to the private sector to GDP) 2.86 0.47
Openness to trade 0.458 1.232
(strcuture-adjusted trade volume to GDP) 1.45 2.52
Government spending -1.203 -3.221
(Government consumption to GDP) 2.04 4.71
# of countries 110 107 104 94 92 105 102 104 106 99
# of observations 685 364 339 288 275 348 337 341 349 313
Hansen test of overident. Restrictions (p-val) 0.16 0.49 0.53 0.62 0.58 0.55 0.53 0.43 0.37 0.13
Test for second order serial correlation (p-val) 0.23 0.17 0.18 0.20 0.25 0.23 0.20 0.20 0.28 0.15
Note: The table reports the results of regressing the s.d. of a country's growth rate on the variables in the first column. The data is in non-overlapping 5-year
averages. Estimation method is GMM and the instrument set is the same in all the specifications and includes the lagged values of monetary policy volatility,
fiscal policy Volatility, price stability, access to credit, openness to trade, government spending and the gini coefficient. The external shock is treated as
exogenous. t-stat in italics. The null hypotheses in the Hansen test of overidentifying restrictions and the test for second order serial correlation are no correlation
between the residuals and the instrument set and no second order correlation (i.e. a large p-value indicates that there is no evidence against the null hypothesis).
31
Table 5. Inequality and the transmission of international shocks.
Dependet variable is s.d. of per capita growth rates
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Persistence 0.272 0.009 -0.193 0.035 -0.066 0.024 -0.064 0.008 -0.008 -0.033
(lagged dependent) 4.65 0.26 3.47 0.99 1.59 0.43 1.29 0.25 0.17 0.79
External shock 0.097 -0.382 -0.418 -0.240 -0.266 -0.318 -0.550 -0.319 -0.353 -0.230
(s.d. of annual change in terms of trade) 10.67 5.36 5.34 4.09 3.93 4.13 6.70 4.90 5.10 3.25
External shock * inequality 1.179 1.319 0.757 0.854 0.990 1.648 1.050 1.111 0.830
(s.d. of annual change in terms of trade *gini) 6.41 6.43 4.97 4.78 5.05 7.66 6.14 6.17 4.60
Fiscal policy volatility 0.043 0.038
(s.d. of annual growth of gov. consumption) 3.91 4.15
Monetary policy volatility 0.000 -0.001
(s.d. of annual growth of corrected money) 0.05 0.63
Price stability -0.057 -0.052
(inflation rate) 0.17 0.14
Access to credit 0.171 -0.248
(credit to the private sector to GDP) 0.56 0.76
Openness to trade 0.304 1.204
(strcuture-adjusted trade volume to GDP) 1.35 3.11
Government spending 0.357 -0.823
(Government consumption to GDP) 0.72 1.27
# of countries 110 107 104 94 92 105 102 104 106 99
# of observations 685 364 339 288 275 348 337 341 349 313
Hansen test of overident. Restrictions (p-val) 0.20 0.72 0.84 0.87 0.73 0.62 0.60 0.72 0.73 0.46
Test for second order serial correlation (p-val) 0.34 0.48 0.68 0.33 0.42 0.51 0.36 0.46 0.43 0.31
Note: The table reports the results of regressing the s.d. of a country's growth rate on the variables in the first column. The data is in non-overlapping 5-year
averages. Estimation method is GMM and the instrument set is the same in all the specifications and includes the lagged values of monetary policy volatility,
fiscal policy volatility, price stability, access to credit, openness to trade, government spending and the gini coefficient. The external shock is treated as
exogenous. t-stat in italics. The null hypotheses in the Hansen test of overidentifying restrictions and the test for second order serial correlation are no correlation
between the residuals and the instrument set and no second order correlation (i.e. a large p-value indicates that there is no evidence against the null hypothesis).
32
Table 6. Inequality and the transmission of international shocks: augmented models.
Dependet variable is s.d. of per capita growth rates
s.d. of annual growth s.d. of annual change
of trading partners in terms of trade
(1) (2)
External shock * inequality 6.194 1.498
3.63 8.47
External shock * credit -0.206 -0.029
1.04 0.77
External shock * inequality 5.115 0.837
3.22 4.76
External shock * trade openness 0.258 0.040
1.54 2.34
Note: The table reports the parameter and t-stat in regressions like those in column (2) of tables 4 and 5, augmented with an interaction of the external shock with
either financial deepening or trade openness.
33
Table 7. Inequality and the transmission of international shocks.
Dependet variable is fiscal policy volatilility Dependet variable is monetary policy volatilility
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Lagged fiscal policy volatility 0.206 0.202 0.049 0.068 0.212 0.450 0.404 0.086
(s.d. of annual growth of gov. consumption) 5.28 5.16 1.13 1.37 2.59 1.76 1.17 0.24
Lagged monetary policy volatility -0.194 -0.647 -0.004 0.251 0.141 0.108 -0.093 -0.148
(s.d. of annual growth of corrected money) -0.91 0.45 -0.73 7.70 7.84 4.75 -8.01 -3.98
Inequality 3.810 0.007 0.004 -1.290 -0.905 -1.341 0.118 1.000
(gini coefficient) 0.40 0.87 1.28 -0.08 -1.44 -1.52 0.23 1.40
External shock 0.600 -1.07
(s.d. of annual growth of trading partners) 2.60 -0.97
Price stability 1.581 2.352
(inflation rate) 1.73 3.38
# of countries 121 110 95 90 94 115 105 95 90 94
# of observations 551 312 245 239 238 523 272 246 240 239
Hansen test of overident. Restrictions (p-val) 0.68 0.38 0.69 0.68 0.53 0.18 0.67 0.71 0.51 0.69
Test for second order serial correlation (p-val) 0.28 0.20 0.61 0.62 0.54 0.36 0.32 0.32 0.32 0.24
Note: The table reports the results of regressing the s.d. of fiscal policy and monetary policy volatility on the variables in the first column. The data is in non-
overlapping 5-year averages. Estimation method is GMM and the instrument set is the same in all the specifications and includes the lagged values of monetary
policy volatility, fiscal policy, and the gini coefficient. The external shock is treated as exogenous. t-stat in italics. The null hypotheses in the Hansen test of
overidentifying restrictions and the test for second order serial correlation are no correlation between the residuals and the instrument set and no second order
correlation (i.e. a large p-value indicates that there is no evidence against the null hypothesis).
34
V. Conclusions
Inequality is high in Latin America however we measure it. The region has among
the highest traditional income inequality measures and it does not fare better with respect
to more dynamic indicators of social or educational mobility. Income inequality tended to
increase in most countries during the 1990s, with significant variation across countries,
though trends look somewhat better when we use the right price deflators by income
ventiviles.
Such high inequality levels must be a concern for policy makers, not just on
equity but on efficiency grounds. We illustrate three channels through which high
inequality reduces economic growth, based on previous research but also on new
evidence. First, high inequality is partly responsible for the high levels of crime and
violence in most countries of the region, which in turn affect their growth performance.
Second, high inequality leads to high poverty levels (for a given income per capita
average), and high poverty in turn is a drag on physical and human capital accumulation,
thus indirectly lowering growth rates. Third, high inequality contributes to high output
volatility (by augmenting the effect of external shocks), which in turn affects economic
growth adversely through several channels.
Various factors are behind the persistence of high inequality levels. Inequality of
assets, especially of human capital, is a major determinant of current income inequality.
Convexity of returns to education and high degrees of marital sorting exacerbate the
importance of the existing high concentration in educational attainment. Low educational
mobility becomes thus a critical factor behind the persistence of high income inequality.
Overcoming low educational mobility requires attention both to supply side (availability
of schools and teachers) and demand side issues: given the convexity of returns (marginal
returns become significant only after completion of secondary schooling), credit
constraints and increasingly high opportunity costs for the poor in terms of forgone
income from youth in school, wisely designed and implement conditional cash transfers
appear as a potentially key instrument for breaking this vicious circle. At the same time,
attention must be paid to factors that determine differences in returns to education across
income groups: quality of schools, access to pre schooling facilities and, most
importantly, access to complementary assets (financial services, public infrastructure,
etc). Equalizing access to assets is part of a broader Agenda of equalizing opportunities,
which would have the major advantage of contributing to both higher equality of incomes
and higher growth. The State has of course a central role in guaranteeing higher equality
of opportunities.
But equalizing access to assets and opportunities take time and the state can do
significant income redistribution in the short term, without incurring in high growth costs.
Indeed, we have shown that the extent of income redistribution through taxes and,
specially, transfers explain more than half of the observed differences in disposable
income distribution between Latin American and developed countries. Overall, public
expenditures in Latin America are much less progressive in Latin America than in OECD
countries. This is both due to the weight of big expenditure items that benefit
35
disproportionately the well to do (generalized subsidies to energy consumption, pensions
and higher education) and the still relatively minor importance of targeted transfers.
Transforming the Latin American state, so that it becomes an agent of equalization of
opportunities and efficient income redistribution, is perhaps the most significant
challenge in our regional development agenda.
References
Aaberge, R., A. Bjorklund, M. Jantti, M. Palme, P.J. Pedersen, N. Smith and T.
Wennemo. 2002. "Income inequality and income mobility in the Scandinavian countries
compared to the United States", Review of Income and Wealth 48 (2002), pp. 443�469.
Acemoglu, D., S. Johnson, and J. Robinson. 2001. "The Colonial Origins of
Comparative Development: An Empirical Investigation."American Economic Review 91:
1369�1401.
Alesina, A., and D. Perotti. 1996. "Income Distribution, Political Instability, and
Investment." European Economic Review 40 (6): 1203�28.
Alesina, A., and D. Rodrick. 1994. "Distributive Politics and Economic Growth."
Quarterly Journal of Economics 109: 465�90.
Andersen, L. 2001. "Social Mobility in Latin America: Links with Adolescent
Schooling." Washington, DC: Inter-American Development Bank.
Barro, R.. 2000. "Inequality and Growth in a Panel of Countries." Journal of
Economic Growth 5: 5�32.
Behrman, J., N. Birdsall, and M. Sz�kely. 1999. "Intergenerational Mobility in
Latin America: Deeper Markets and Better Schools Make a Difference." In New Markets,
New Opportunities? ed. N. Birdsall and C. Graham. Washington, DC: Brookings
Institution Press.
Behrman, J., A. Gaviria, and M. Sz�kely. 2001. "Intergenerational Mobility in
Latin America." Inter-American Development Bank, Washington, DC.
Bertola, G., and R. J. Caballero. 1994. "Irreversibility and Aggregate Investment."
The Review of Economic Studies 61 (2) :223--46.
Bourguignon, F 2003. "The Growth Elasticity of Poverty Reduction; Explaining
Heterogeneity across Countries and Time Periods." In Inequality and Growth: Theory
and Policy Implications, ed. T. Eicher and S. Turnovsky. Cambridge, MA: MIT Press.
Calder�n, C and E. Levy-Yeyati. 2007. "Three degrees of vulnerability: External
shocks in developing countries", World Bank, Washington, DC.
36
Deaton, A. 1985. "Panel Data from Time Series of Cross-Sections." Journal of
Econometrics 30:109�26.
Deininger, K. and P. Olinto. 2000. "Asset Distribution, Inequality and Growth".
World Bank. Policy Research Working Paper 2375.
de Ferranti, D., G. Perry, I. Gill, and L. Serv�n, with F. Ferreira, N. Ilahi, W.
Maloney, and M. Rama. 2000. Securing Our Future in a Global Economy. Washington,
DC: World Bank.
de Ferranti, D., G. Perry, F. Ferreira, and M. Walton. 2004. Inequality in Latin
America and the Caribbean: Breaking with History? Washington, DC: World Bank.
de Ferranti, D., G. Perry, D. Lederman, A. Valdes, and W. Foster. 2005. Beyond
the City: The Rural Contribution to Development. Washington, DC: World Bank.
Engerman, S., and K. Sokoloff. 2004. "Factor Endowments, Institutions, and
Differential Paths of Growth among New World Economies: A View from Economic
Historians of the United States." NBER Working Paper H0066, National Bureau of
Economic Research, Cambridge, MA
Fajnzylber, P., D. Lederman, and N. Loayza. 2002. "Inequality and Violent
Crime". Journal of Law and Economics, vol. XLV, April 2002. The University of
Chicago.
Fat�s, A. (2002). "The Effects of Business Cycles on Growth." Central Bank of
Chile Working Paper No. 156, May.
Fatas, A and I. Mihov. 2007. "Fiscal Discipline, Volatility and Growth"
forthcoming in Prudence or Abstinence? Fiscal Policy, Stabilization and Growth,
Stanford University Press.
Forbes, K. 2000. "A Reassessment of the Relationship between Inequality and
Growth." American Economic Review 90 (4): 869�87.
Galor, O., and J. Zeira. 1993. "Income Distribution and Macroeconomics."
Review of Economic Studies 60 (1): 35�52.
Glewwe, P., H. Jacoby and E. King (2001). "Early Childhood Nutrition and
Academic Achievement: A Longitudinal Analysis". Journal of Public Economics, 81:
345-368
Go�i, E., H. Lopez, and L. Serv�n (2005). "Getting Real about Inequality." World
Bank, Washington, DC.
37
Go�i, H. Lopez y L. Serv�n (2008), "Reforma Fiscal y Equidad Social en
Am�rica Latina", in J.L. Machinea y N. Serra eds., Hacia un Nuevo Pacto Social,
ECLAC-CIDOB, Santiago de Chile.
Grawe, N. 2002. "Quantile Measures of Mobility in the US and Abroad." In
Generational Income Mobility in North America and Europe, ed. Miles Corak.
Cambridge: Cambridge University Press.
Hanushek, Eric A. 1992. "The trade-off between child quantity and quality."
Journal of Political Economy 100,no.1 (February):84-117.
Hnatkovska, V. and N. Loaiza. 2004. "Volatility and growth," Policy Research
Working Paper Series 3184, The World Bank
Hobjin, B., and D. Lagakos. 2003. "Inflation Inequality in the United States."
Staff Report 173, Federal Reserve Bank of New York, New York.
Krebs, T., P. Krishna, and W. Maloney. 2005. "Income Risk and Human Capital
in LDCs.", World Bank, Washington, DC.
Leibowitz, A., 1974. Home investments in children. Journal of Political Economy
82, S111-31.
Levine, R. 1997. "Financial Development and Economic Growth: Views and
Agenda," Journal of Economic Literature, American Economic Association, vol. 35(2),
pages 688-726, June.
Li, H., and H. Zou. 1998. "Income Inequality Is Not Harmful for Growth: Theory
and Evidence." Review of Development Economics 2 (3): 318�34.
Loayza, N., P. Fajnzylber, and C. Calderon. 2005. Economic Growth in Latin
America and the Caribbean. Washington, DC: World Bank.
L�dola, A., F. Busso and F. Cerimedo (2000); "Sesgos en el IPC: el Sesgo
Plutocr�tico en Argentina," Anales de la Asociaci�n Argentina de Econom�a Pol�tica.
Lopez, H., and L. Serv�n. 2006a. "A Normal Relationship? Poverty, Growth and
Inequality." World Bank, Washington, DC.
------. 2006b. "Too Poor to Grow." World Bank, Washington, DC.
Martin, P. and C. A. Rogers (1997). "Stabilization Policy, Learning By Doing,
and Economic Growth". Oxford Economic Papers.
Martin, P. and C. A. Rogers (2000). "Long-term growth and instability".
European Economic Review 44 (2): 359-381.
38
Perotti, R. 1996. "Growth, Income Distribution and Democracy." Journal of
Economic Growth 1 (2): 149�87.
Perry, G., O. Arias, H. L�pez, W. Maloney, L. Serv�n (2006). "Poverty reduction
and growth: virtuous and vicious circles". The World Bank.
Ramey, G. and V. Ramey (1995). "Cross-Country Evidence on the Link between
Volatility and Growth". American Economic Review 85 (5): 1138-1150.
Ravallion, M. 1997. "Can High Inequality Development Countries Escape
Absolute Poverty?" Economics Letters 56 (1): 51�7.
------. 2004. "Pro-Poor Growth: A Primer." Policy Research Working Paper
3242, World Bank, Washington, DC.
Rodrik, D. 1999. "Where Did All the Growth Go? External Shocks, Social
Conflict, and Growth Collapses." Journal of Economic Growth 4 (4): 385�412.
Wolf, Holger. 2003. "Volatility: Definitions and Consequences." Chapter 1 in
"Managing Volatility and Crises: A Practitioner's Guide.". World Bank Regional Study
Ruiz-Castillo J., E. Ley, and M. Izquierdo. 2002. "Distributional Aspects of the
Quality Change Bias in the CPI: Evidence from Spain." Economic Letters 76 (1): 137�
44.
World Bank. 2006. World Development Report: Equity and Development.
Washington, DC: World Bank.
39