| dc.description.abstract | While corruption is documented to have high social costs, the mechanisms that enable and deter corruption are not entirely understood. This paper takes advantage of the introduction of a random-audits program in Brazil, for which the only practical effect is information disclosure, to assess the effects of monitoring on corruption deterrence: because auditors were entitled to investigate transfers that occurred prior to the time of audit, we are able to compare incumbents’ decisions when they did not know they could be audited with their subsequent decisions, when they knew they had a much higher probability of being exposed. We depart from a simple theoretical model to derive empirically testable predictions, which are contingent upon local constraints and embezzlement opportunities, of the effects of the program on the prevalence of corruption. The results are that corruption dramatically dropped after the program was introduced; nevertheless, the decrease in corruption has not led to higher compliance or better health outcomes. Politicians seem to have adapted, and irregularities have only moved to where it is less straightforward to identify corruption. | |
