Race and Income Distribution: Evidence from the USA, Brazil and South Africa

• Published date: 01 February 2014
• Author: Carlos Gradín
• DOI: http://doi.org/10.1111/rode.12070
• Date: 01 February 2014

Race and Income Distribution: Evidence from the

USA, Brazil and South Africa

Carlos Gradín*

Abstract

The aim of this paper is to provide some empirical evidence about black–white differentials in the distribu-

tion of income and wellbeing in three different countries: Brazil, the USA and South Africa. In all cases,

people of African descent are in a variety of ways socially disadvantaged compared with the relatively more

affluent whites. We investigate the extent of these gaps in comparative perspective, and analyze to what

degree they are associated with differences in the observed characteristics of races, such as where they live,

the types of household they have, or their performance in the labor market. We undertake this analysis

with the Oaxaca–Blinder decomposition at the means and with a propensity score approach at the entire

distribution. Our results show how the factors underlying the racial divide vary across countries and income

quantiles.

1. Introduction

Undoubtedly, in several countries in the world, a large and socially disadvantaged

black population is found cohabitating with a more affluent group of whites. Histori-

cal reasons, however, and the magnitude of these socioeconomic gaps by race may

differ in each context. The aim of this paper is, first, to document these current racial

inequalities in the USA, Brazil and South Africa, in terms of equivalized household

incomes in a comparative perspective, and then, to assess to what extent they are

associated with the poor endowments of African descendants in each country in terms

of their geographical location and demographic characteristics, such as the number of

children or single mothers, their education attainment, or their labor market perfor-

mance (characteristics effect). Alternatively, they might be the result of these charac-

teristics making them less effective in providing earnings to their households

(coefficients effect).

More specifically, we analyzed the magnitude of the average household income dif-

ferential between races in each country and, after estimating household income

regressions, we decomposed this gap into characteristics and coefficients effects fol-

lowing the well known Oaxaca (1973)–Blinder (1973) approach. This decomposition

was undertaken at two different levels: at the aggregate level, we estimated the joint

contribution of all characteristics for each country respectively; while at the detailed

level we identified the individual contribution of each set of characteristics. We

further analyzed how the racial differential by race in household incomes and its

determinants vary across income quantiles of the distribution using a propensity-

based reweighting DiNardo–Fortin–Lemieux approach. This latter approach also

allowed us to identify the factors underlying the over-representation of blacks among

* Gradín: Facultade de CC. Económicas, Universidade de Vigo, Campus Lagoas-Marcosende s/n, 36.310

Vigo, Spain. Tel: +34-986813527; Fax: +34-986812401; E-mail: cgradin@uvigo.es. The author acknowledges

financial support from the Spanish Ministerio de Educación y Ciencia (Grant ECO2010-21668-C03-03/

ECON) and Xunta de Galicia (Grant 10SEC300023PR) as well as comments from an anonymous referee.

Review of Development Economics, 18(1), 73–92, 2014

DOI:10.1111/rode.12070

© 2014 John Wiley & Sons Ltd

the poor, as well as their under-representation at middle and higher income levels.

For that we analyzed the racial differential in poverty measures, as well as in densities

along the income scale. Using these two approaches we identified those factors which

are more strongly associated with lower income among blacks as compared with

whites in each country, then showing what policies are expected to have a higher

impact on reducing racial inequalities, an issue which is of undoubted interest for

policymakers and analysts interested in the racial divide.

2. Data

In order to undertake the comparative analysis, we will use microdata from most rep-

resentative household surveys in each country with national coverage of (mostly)

non-institutionalized population, providing information on main households and indi-

vidual characteristics, including income and self-reported race/ethnic group. In the

case of Brazil, we use the 2007 release of the National Household Survey (Pesquisa

Nacional por Amostra de Domicílios, PNAD) from Instituto Brasileiro de Geografia e

Estatística. Respondents are asked to self-categorize their skin color or race into one

of five groups: indígena (indigenous), branca (white), preta (black), amarela (Asian)

and parda (of mixed race). For most of the analysis we pooled blacks and people of

mixed race into a single group (African Brazilians), since people of African descent

might choose either of these categories owing to the social stigma attached to black-

ness (Telles, 2002). The data used for the analysis in the case of the USA come from

the “Current Population Survey” (CPS), Annual Social and Economic March Sup-

plement, conducted by the US Census Bureau. In this survey, people are asked to

answer questions about their race(s) from six distinct groups: white, black, American

Indian or Alaskan Native, Asian, Native Hawaiin other Pacific Islander, and Other

race. Further, this survey inquires whether or not the origin of each person is Spanish,

Hispanic, or Latino. On the basis of these questions, we broke up the population into

five non-overlapping groups: non-Hispanic whites (those who only declared this race),

blacks or African Americans (identifying themselves as Black, either alone or in com-

bination with other races, regardless of whether they identify or not as having His-

panic origin), non-black Hispanics or Latinos, Asian Americans (who further did not

identify themselves as being Black or Hispanic), and others, even if we will focus the

main analysis on the first two groups. Finally, for the case of South Africa, we use the

2005/06 release of the Income and Expenditure Survey (IES) conducted by Statistics

South Africa (Stats SA) between September 2005 and August 2006. Respondents to

this survey report their ethnic group choosing between white, black, colored (of

mixed race), Indian or Asian, and other race. For the same reasons as in Brazil, in

most of our analysis blacks and colored will be combined in the same group of

African descents.

Individual income used throughout this paper was obtained by dividing the total

amount for his/her household annualized disposable income measured in local cur-

rency (US dollars, South African Rands and Brazilian Reals) by the square root of

the number of cohabiting members. In doing this, we take into account the existence

of economies of scale derived from living together and sharing expenses in a standard

and tractable way, allowing comparability across countries following Buhmann et al.

(1988).1For the sake of comparability among income distributions across countries,

income will be also measured relative to the corresponding median of the adjusted

distribution in each country.

74 Carlos Gradín

© 2014 John Wiley & Sons Ltd