Who Is Punishing Corrupt Politicians Voters or the Central Government?

In 2003 the Brazilian central government (CG) launched an anti-corruption program. Since then municipalities have been randomly selected to be audited on a monthly basis. Evidence in the literature suggests that the probability of re-election of an incumbent mayor decreases as the number of reported corruption violations rises before the municipal elections. By exploiting the exogenous variation in the timing of the release of the audit reports and the Brazilian institutional scheme, this paper sheds light on the mechanisms through which the Brazilian anti-corruption program functions. After the release of the audit reports, municipalities where more than two corruption violations were reported receive 26% fewer transfers from the CG. Total expenditure on infrastructure is also reduced. While the CG increases the amount of transfers to municipalities where the mayor is both affiliated with the partys president and found to be honest, it helps politically aligned municipalities with high levels of released corruption to move through the punishment process more quickly. The effects of the dissemination of corruption information on the probability of re-election of incumbent mayors seem to gradually disappear with time. Yet, when these effects have completely faded and voters have time to feel the consequences of receiving fewer transfers, the probability of re-election of corrupt politicians decreases.