Truthful Reporting, Moral Hazard and Purely Soft Information

We examine a hierarchical model where a principal hires a risk averse supervisor to monitor the e ort exerted by a productive agent. We assume that the supervisor can misreport the collected evidence without incurring any cost. We develop a corruption-proof contract which makes it sequentially rational for the supervisor to report truthfully. Crucial features of our contract are the timing at which the report is sent and the supervisor's payment scheme. In particular, the report must be sent before the outcome observation and the principal must reward the supervisor if and only if her report maximizes the conditional probability of the realized outcome. We also highlight a non-trivial interplay between corruption incentives, the signal precision and the supervisor's risk aversion.