Brazil in black and white? Race categories, the census, and the study of inequality

Abstract Many scholars advocate the adoption of a black-and-white lens for the analysis of racial inequality in Brazil. Drawing on a nationally representative dataset that includes race questions in multiple formats, we evaluate how removal of the 'brown' category from the census or other social surveys would likely affect: (1) the descriptive picture of Brazil's racial composition; and (2) estimates of income inequality between and within racial categories. We find that a forced binary question format results in a whiter and more racially unequal picture of Brazil through the movement of many higher income mixed-race respondents into the white category. We also find that regardless of question format, racial inequality in income accounts for relatively little of Brazil's overall income inequality. We discuss implications for public policy debates in Brazil, and for the broader scientific and political challenges of ethnic and racial data collection and analysis. Keywords: Racial classificationracial inequalitycensusstatisticsBrazilaffirmative action Notes 1. Some researchers note that it is problematic to collapse browns into a collective black category because many Brazilians who self-identify as brown are of mixed European and indigenous ancestry, not African (Guimarães 2001). 2. The terms afro-descendente and Afro-Brazilian are somewhat problematic labels for the sum of 'nonwhites' because in Brazil – in contrast to the United States – many self-identified whites claim some African ancestry (Rocha and Rosemberg 2007). Whites' tendency to acknowledge some African ancestry stems from Brazil's national origin myth, which asserts that the Brazilian population was formed through the fusion of three 'races': Europeans, African, and Indians. (Beyond founding myths, genetic testing putatively demonstrates that many whites do have considerable African ancestry Santos et al. 2009.) 3. Rocha and Rosemberg (2007) suggest the popular meanings of these two terms are currently in flux, and may be diverging. 4. An empirical comparison between Theil and Gini indexes in Brazil does not show significant differences in their estimates (Ferreira, Leite and Litchfield Citation2006, p. 7). 5. For example, Nascimento and Nascimento (2001, p. 125) write, 'while official statistics put the sum of pretos and pardos at 48 per cent, estimates that take into account their distortion by the whitening ideal are closer to 70 or 80 per cent'.