Colour violence, deadly geographies, and the meanings of “race” in Brazil

ABSTRACT

This paper analyses the relationships between “race” and violent victimization, and between “race” and support for violent practices of social control in Brazil, using nationally representative survey data. I start from the premise that “race” is a set of relational practices rather than bounded “racial groups”. I operationalize this relational understanding of “race” methodologically by triangulating three measures of “race” – self-identified “census race”, interviewer-identified skin colour, and racial composition of the municipality – in conjunction with measures of class, gender and space. I find that whiter geographic spaces have lower overall levels of violent victimization, but that interviewer-identified darker-skinned individuals are disproportionately victimized in these whiter geographic spaces. Controlling for other variables, self-identified census race is not correlated with violent victimization. I find that public support for violent practices can best be understood by considering people’s simultaneous relationships to race, gender, class and spatial categories and hierarchies.

KEYWORDS:

• Race
• colour
• violence
• public opinion
• Brazil
• space

Acknowledgements

I thank Flavia Andrade, Fernando Calderón Figueroa, Philip Goodman, Sherri Klassen, Aliza Luft, Graziella Moraes Silva, Patricia Louie, Mara Loveman, Neda Maghbouleh, Ellis Monk, Jeronimo Muniz, Akwasi Owusu Bempah, Rania Salem, Felipe Schwartzman, Dan Silver, Gail Super and Geoff Wodtke, attendants of various workshops, conferences and talks where I presented this paper, as well as the anonymous reviewers, for their helpful feedback at various stages of this paper’s writing. I thank Edward Telles and Elizabeth Zechmeister for helping me access and understand this paper’s dataset.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 The story in this paper is mostly about urban violence: 80 per cent of the Brazilian population, and 93 per cent of my survey respondents, live in urban areas.

2 I experimented with both higher and lower levels of aggregation, but collapsing categories 7 and 8 together causes a loss of variation in the results, while maintaining categories above 8 left too few cases to allow for statistical analysis.

3 The AmericasBarometer only has self-reported measures of violent victimization, which likely leads to subjective evaluation and under-reporting. The murder in the neighbourhood reported correlates well with the actual murder rate in the respondents’ municipality (.37). The measure of violent victimization, because of how it is framed (as being victim of a crime) is likely to under-estimate kinds of violence that is normalized in society, such as domestic violence and police violence. On the other hand, self-rated victimization often captures violence not found in police reports, which tend to be biased by the police's individual or corporate interests. While a triangulation between different measures of victimization would be ideal, to date, as far as I know, no dataset exists that would allow me to triangulate both different of race – as I do in this paper – and of violence, at the individual level.

4 The survey collected data on household income in 2012 and 2014 and on family income in 2010. Income data was originally available as a set of aggregate nominal income categories that are not consistent across years, which I converted into quantiles in relation to the income distribution among survey respondents for that year, choosing the approximate quantile cutoffs that could be compared across years.