Making Corruption Harder: Asymmetric Information, Collusion, and Crime∗

We model the investigation of criminal activity as a principal-agent-monitor prob-lem in which the agent can corrupt the monitor and side-contract to destroy evidence. Building on insights from Laffont and Martimort (1997) we study whether the princi-pal can benefit from endogenously creating asymmetric information between the agent and the monitor. We show that the principal can potentially obtain large benefits from randomizing the incentives given to the monitor (and letting those serve as the monitor’s private information), but that the optimality of random incentives depends on pre-existing patterns of private information. This motivates us to develop a data-driven framework for policy evaluation using only unverified report data