Abstract
This paper provides evidence that audits can be an effective instrument for improving public services, depending on institutional and sectoral conditions. Leveraging randomized audits of municipal funds in Brazil, I show that audits led to better outcomes for public school students, but were not as effective in improving the provision of other public services. Using a difference-in-differences strategy, I show causal evidence that the audits led to a decrease in dropout rates in municipal elementary and middle schools by 0.41 percentage points (17%), and that this effect persisted for more than eight years. These estimates correspond to approximately one-third of the decrease in dropout rates resulting from the early years of Bolsa Família, a conditional cash transfer program partially designed to minimize student abandonment. The audits were most effective in municipalities where high levels of corruption were uncovered, and whose residents relied heavily on conditional cash transfers.