How do voters react when they learn that their politicians have violated the law? A central expectation of representative democracy is the idea that voters hold their politicians accountable. If they misbehave while in office, they will be voted out of office. One might expect accountability to be especially effective in Brazil, where decades of political corruption have led to successful activism to strengthen sanctions against malfeasant officeholders.
Survey experiments in which respondents are asked to choose between corrupt and non-corrupt hypothetical candidates routinely bear out this argument. Evidence from field experiments, however, is more mixed, and no study had been done to conduct both survey and field experiments to evaluate the same intervention. In developing democracies where politicians often operate extensive patronage networks, the electoral durability of incumbents in the face of scandal might be stronger.
The authors conducted a field experiment in Pernambuco, Brazil, to test whether voters punish politicians that are accused of malfeasance. The authors randomly sampled 3,200 voters two weeks prior to the 2016 municipal elections. A segment of respondents were randomly assigned to receive fliers with information about whether the incumbent mayor was in compliance with government laws and regulations. Other respondents received no information at all. The authors then conducted a follow-up survey in which they were able to contact 2,577 of the originally sampled voters to ask which candidate they voted for. A survey vignette experiment was also embedded within this post-electoral survey and asked whether respondents would vote for a hypothetical candidate widely praised for economic growth during his tenure. Respondents in the treatment condition were also informed that this candidate was implicated in a budgetary scandal, as an auditing institution rejected the mayor’s budget. In order to ensure comparability between the vignette and field experiments, the authors only examined vignette treatment effects for respondents who live in municipalities where the mayor’s accounts were rejected but who never received a flier with this information. The authors then compare the proportion of respondents voting for the incumbent across those assigned to the treatment and control groups.
The authors find that informing voters about either the approval or the rejection of their mayor’s accounts has no significant effect on voting behavior. However, in the vignette survey experiment embedded in the field experiment, the authors find that informing voters that a candidate’s accounts are rejected makes them significantly less likely to vote for that candidate, by a margin of 44 percentage points. These findings show that voters respond to norms against corruption in the abstract, but that the response to norms does not result in change in behavior at the polls.
The authors find significant evidence that a norm of anti-corruption exists among voters in Brazil. However, they find no evidence that this norm translates to action, even when voters possess the relevant information. One implication of this finding is that achieving higher rates of accountability in developing democracies cannot be achieved through informational campaigns alone. This research suggests that those interested in motivating higher rates of democratic accountability should couple their informational with grander interventions—undermining networks that sustain vote-buying schemes, for example.