"We provide a theoretical framework for understanding when an official angles for a bribe, when a client pays, and the payoffs to the client's decision. We test this framework using a new data set on bribery of Peruvian public officials by households. The theory predicts that bribery is more attractive to both parties when the client is richer, and we find empirically that both bribery incidence and value are increasing in household income. However, 65% of the relation between bribery incidence and income is explained by greater use of officials by high-income households, and by their use of more corrupt types of official. Compared to a client dealing with an honest official, a client who pays a bribe has a similar probability of concluding her business, while a client who refuses to bribe has a probability 16 percentage points lower. This indicates that service improvements in response to a bribe merely offset service reductions associated with angling for a bribe, and that clients refusing to bribe are punished. We use these and other results to argue that bribery is not a regressive tax"--National Bureau of Economic Research web site.