We report results from a randomized experiment designed and implemented by the Brazilian central government audit agency to test whether increased audit risk deters corruption and waste in local public procurement and improves provision of public services. We measure waste and corruption as irregularities in local public procurement and service delivery uncovered by central government auditors. Our estimates suggest that increasing audit risk by about 20 percentage points reduced the proportion of non-competitive procurement modalities adopted by local managers by about 17 percent. Higher audit risk also reduced the proportion of local procurement processes involving waste or corruption by about 20 percent. In contrast, we nd no evidence that increased audit risk affected the quality of publicly provided preventive and primary health care services, measured using client satisfaction surveys. We also nd no evidence that higher audit risk had an effect on local compliance with national guidelines of the conditional cash transfer program Bolsa Família, measured in terms of appropriate inclusion of bene ciaries into the program or their compliance with health and education conditionalities.

This paper provides regression discontinuity evidence on development impacts of intergovernmental transfers. Extra transfers in Brazil increased local government spending per capita by about 20 percent over a 4 year period with no evidence of crowding out own revenue or other revenue sources. Schooling per capita increased by about 7 percent and literacy rates by about 4 percentage points. In line with the effect on human capital, the poverty rate was reduced by about 4 percentage points. Somewhat noisier results also suggest that the reelection probability of local incumbent parties in the 1988 elections improved by about 10 percentage points. (JEL H72, H75, I21, I28, I32, I38, O15)

Can rules be used to shield public resources from political interference? The Brazilian constitution and national tax code stipulate that revenue sharing transfers to municipal governments be determined by the size of counties in terms of estimated population. In this paper I document that the population estimates which went into the transfer allocation formula for the year 1991 were manipulated, resulting in significant transfer di¤erentials over the entire 1990's. I test whether conditional on county characteristics that might account for the manipulation, center-local party alignment, party popularity and the extent of interparty fragmentation at the county level are correlated with estimated populations in 1991. Results suggest that revenue sharing transfers were targeted at right-wing national deputies in electorally fragmented counties as well as aligned local executives.