Absract - we develop a simple framework for analyzing the optimal design of leniency programs, which allow cartel members to denoune their collusive agrements. We highlight a basic tradeoff between two opposite forces: leniency can destablize cartels, by encouraging firms to report and bring evidence to the antitrust authority, but it can also reduce the expected penalties that cartel members face. We characterize the optimal leniency rates, both before any investigation and once an investigation is opened, and show that these two leniency opportunities are particularly useful when random investagations are unfrequent and/or unlikely to succeed in the absence of self-reporting; we also compare the effectiveness of alternative rules for late informants and repeated offenders.
