When Do Voters Punish Corrupt Politicians? Experimental Evidence from Brazil

While there is a vast literature about the consequences of corruption, research focused on the causes of non-corrupt governments is still nascent. Electoral accountability is one important means of changing existing corruption dynamics in government, and one potential way of punishing corrupt elected officials is to give voters information about the corrupt behavior. In a randomized field experiment in a Brazilian election in 2008, we take advantage of a situation where both the incumbent and challenger had corruption convictions in order to examine whether informing voters about their convictions has an effect on voting behavior. We inform voters by distributing 187,177 fliers to individual households in the vicinity of 200 voting locations that detail the corruption convictions of both candidates in a run-off election. One hundred locations received the flier of a candidate from a center-right party, accused of using public funds to take out ads in newspapers while he was under investigation, and another one hundred locations received the flier of a left-party candidate, accused of giving a no-bid contract of R$2 million (approximately US$1 million) to a non-governmental organization (NGO) she founded. A control group of 200 polling locations received no flier. The flier for the centerright candidate had no effect on vote choice, spoiled ballots, or turnout. The flier for the other candidate, by contrast, moved 2.6 percent of the votes on average relative to the control group, had no effect on spoiled ballots, and a negative 1.2 percent average treatment effect on voter turnout. The turnout results are surprising given that Brazil has mandatory voting. The effect of the treatment also shows the issue-salience of corruption for voters, in contrast to some previous studies. In order to understand the mechanisms that lead to these different results, we conduct a survey experiment in which we obtain preand post-treatment attitudes toward the candidates, and show respondents the fliers. The survey experiment, along with descriptive survey results from the treatment and control group of the field experiment give insight into factors that motivated voting behavior in the election. The results shed light on the effects of negative campaigning, the conditions under which increased information can lead to lower voter turnout, and on varied reactions to corruption by voters.