Repeated Moral Hazard with Boundedly Rational Agents

In this paper we consider a situation where a number of identical myopic agents enter a long term contract with a principal. The actions of the agents cannot be observed, and the principal ooers the agents a payment scheme where payments from the principal to the agents are based on the observable outcome in the corresponding period. Whereas the principal knows the distribution of outcomes, given the level of eeort agents invest, agents do not have perfect knowledge of this distribution. They have initial beliefs about the induced distributions and adapt these beliefs dynamically using a simple updating rule. The principal tries to optimize his discounted expected payoo by taking into account the agents expected learning behavior. It is shown that an intertemporally optimal incentive scheme exists in this setup and optimal schemes for all possible constellations of agents' initial beliefs are characterized. Further, it is shown how the optimal scheme changes if the principal is uncertain about the agents initial beliefs.