DEMOCRACY: The Puzzling Whiteness of Brazilian Politicans

You are here

Berkeley Review of Latin American Studies, Fall 2012

City council chamber of Salvador, one of Brazil’s most diverse cities. (Photo by Valter Pontes/Coperphoto.)

City council chamber of Salvador, one of Brazil’s most diverse cities.
(Photo by Valter Pontes/Coperphoto.)

- Read this article on ISSUU
- Download as printed (.pdf)
- More about this event


The Puzzling Whiteness of Brazilian Politics

by Jean Spencer

Is Brazil really a racial democracy? The idea of racial democracy, originally put forth by the Brazilian sociologist Gilberto Freyre in the 1930s, holds that racial discrimination is much more moderate in Brazil than in countries like the United States, due in part to widespread racial mixing. If Brazil is truly a racial democracy, however, why are the city council members in both Salvador and Rio de Janeiro significantly whiter than their electorates? Thad Dunning, an associate professor of Political Science at Yale University, designed a study to discover the reason for this lack of descriptive democracy.

The first problem Dunning faced was a basic one: defining terms. In Brazil, black, white, and brown are in the eye of the beholder. To get “a quick and dirty” baseline for how different politicians are perceived, he conducted an internet survey where participants were asked to assess the race of a random sample of elected officials and unelected candidates using several different scales. In one, candidates were evaluated on a zero-to-10 scale with zero being the lightest and 10 being the darkest; in another, respondents located candidates in one of multiple color categories; and in a third, participants were asked to place the candidates in one of the five categories used by the Brazilian census: branco (white), pardo (brown), preto (black), amarelo (yellow), and indigena (indigenous). In general, Dunning found that there was a good match between the results of the scales, with the pardo category generating the most heterogeneous responses. Comparing the codings of politicians with census data on residents of Salvador and Rio, he also found that whites were heavily overrepresented on the city councils of both cities, just as he had suspected. 

But why? Dunning considered three main possibilities: whites hold racist attitudes toward other groups; black and brown voters have internalized disparaging attitudes about their own groups; or voter preferences are more influenced by class than race. To test these hypotheses, Dunning ran an experiment designed to tease out voters’ underlying racial biases. He hired black and white actors to create videos that followed the same format as the free hour of coverage that Brazilian television gives to candidates for city council. In order to compensate for differences in the personal appeal of individual “candidates,” he hired six black and six white actors for each city. 

The researchers made a series of videotapes. In one set, the actors gave identical speeches, but in some videos, they wore jeans and in others a suit. In the second set of videos, the actors assumed a particular class identity and mode of dress and gave a speech that emphasized that identity — either “working class” or “elite.” Subjects were then randomly assigned one of the videos and asked to identify the candidate’s race and evaluate his characteristics such as intelligence, trustworthiness, and likeability as well as whether they would vote for him based on the speech they had just viewed. Subjects were also asked to identify their own racial group. The researchers then evaluated how subjects responded to candidates of the same race as themselves compared to those of different races and also how they responded to candidates of the same class compared to those from different classes.

Interestingly, while the race of the faux candidates was perceived as intended about 75 percent of the time, roughly 25 percent of the time candidates who the researchers had identified as black or white were seen as being brown. This finding was especially likely when people were assigned a “white” candidate, a fact that underscores the difficulty of drawing clear racial lines in Brazil.

An even more interesting result was that there were no conclusive results. Dunning found that there were “no effects of race and class that survive[d] standard adjustments for multiple statistical comparisons.” This null result is especially surprising given that Dunning has carried out similarly designed studies in other countries, including Mali and India, where he found that voters do favor candidates from their own ethnic group.

 Comparison of the racial distribution of the citizens of Rio and  Salvador and that of their city council members, based on census  data from the Brazilian Institute of Geography and Statistics (IBGE). (Graphs courtesy of Thad Dunning.)

 Comparison of the racial distribution of the citizens of Rio and  Salvador and that of their city council members, based on census  data from the Brazilian Institute of Geography and Statistics (IBGE). (Graphs courtesy of Thad Dunning.)


Comparison of the racial distribution of the citizens of Rio and  Salvador and that of their city council members, based on census
data from the Brazilian Institute of Geography and Statistics (IBGE).
(Graphs courtesy of Thad Dunning.)

Finding no strong race or class effects on voting, Dunning was left without a clear explanation for what he termed “the puzzling whiteness of Brazilian politicians.” Working together with Natália Bueno, a Ph.D. student in Political Science at Yale, Dunning followed up his initial study by doing a preliminary analysis of three alternative hypotheses that might explain the lack of descriptive representation in Brazil: party influence, barriers to candidate entry, and unequal candidate resources. 

Brazil uses an open-list, proportional representation system. The number of seats allocated to each party depends on its share of the total vote. Within each party, individual seats are given to the candidates who receive the most votes. Since candidates from the same party are competing with one another, this type of system is thought to weaken party control over candidates, which would seem to undermine the party influence hypothesis. Dunning and Bueno investigated another peculiarity of the Brazilian electoral system, however, that they thought might give the party more power over the candidates. In Brazil, voters traditionally do not select a name from a list; rather, they write in the candidates’ number. Dunning noted that parties seemed to have influence over the numbers assigned to individual candidates, and “prominent people tended to have easy-to-remember codes.” He also found that having repeated digits or digits that appeared in sequence correlated with better election results. However, there was “no significant difference for race with regard to the quality of [the candidates’] numbers.” In other words, “whites don’t seem to be getting better numbers.” Given these findings, it doesn’t seem likely that parties are suppressing non-white candidates in the polls.

The second hypothesis that Dunning and Bueno looked into was whether there are barriers that keep non-whites out of politics. They looked at non-elected candidates to see if there were fewer blacks and browns among the total candidate population as well as among those eventually elected to office. What they found was the opposite: whites are roughly proportionally represented (Salvador) or even underrepresented (Rio) among the candidates but overrepresented among the winners. 

Dunning and Bueno’s final hypothesis was that a candidate’s resources helped determine the electoral result. So far, this line of inquiry has been the most promising. In Brazil, candidates are required to declare their assets. In surveying these records, Dunning and Bueno found that candidates elected to office had significantly higher mean total assets, at 432,423 Brazilian reais (R$), than those who were not elected, whose average assets totaled R$188,649. This disparity may provide a clue to the gap in descriptive representation, since whites in general tend to have more resources: on average, whites have a net worth of R$440,790 compared to R$247,378 for nonwhites.

At the end of his presentation, Dunning joked that his findings would be perfect for submission to the mythical Journal of Null Results. Still, his research does pose some interesting questions about democracy in Brazil. First of all, it shows some support for Brazil’s claim to racial democracy. As Dunning said during the question-and- answer session, “Maybe the myth of racial democracy is right.” His follow-up work also points to areas that could be fruitful for future research, including the impact of candidates’ relative wealth on election outcomes. Until such work is done, however, the cause of the racial disconnect between the population and the political class in Brazil remains a mystery.

Thad Dunning is a professor of Political Science at Yale University. He spoke for CLAS on May 2, 2012.

Jean Spencer is the outreach and publications coordinator at the Center for Latin American Studies and a graduate student at UC Berkeley’s Goldman School of Public Policy.

City Hall, Rio de Janeiro.
City Hall, Rio de Janeiro.


Professor Thad Dunning discusses Race, Class and Representation in Brazil, from the CLAS YouTube channel.

You are here