Understanding Corruption

The goal of the paper is to understand causes and consequences of corruption. To that end we estimate a dynamic model of decisions by politicians. We consider a model with the following features. First, individuals in the economy have preferences over a private good and a public good. Second, the private good can be bought in the market at a price p. The public good, however, is produced by means of a production function which is municipality specific. It depends on the inputs from the private sector, inputs from the public sector, and the ability of the mayor governing the municipality. Third, mayors make two types of decisions. They decide which fraction of public funds received from the central government to invest in the production of public consumption and which fraction to steal. They then decide how to allocate the available resources between private consumption and savings. Fourth, at the end of a term the probability that the current mayor is reelected depends on the following variables: the amount of public good produced by the municipality during the term; whether a mayor was audited and conditional on being audited on the amount stolen; the campaign contributions for the mayor relative to the ones for the challengers; an error term which captures the residual randomness. Finally, the economy is populated by mayors who have different preferences for public consumption relative to private consumption. Municipalities with mayors that care more about public consumption relative to private consumption enjoy higher levels of public consumption. These five features imply that in the model mayors invest in public consumption for two reasons: because they care about public consumption; because it increases the probability of reelection and therefore of having control over public funds. The model is estimated using corruption data from municipalities in Brazil and Puerto Rico.