On moral hazard and persistent private information

I examine a simple model of dynamic moral hazard in which the agent has persistent private information. I show that despite the complexity of the framework, the problem has a simple solution that can be found using standard methods. The incentives at the optimal contract can be captured using two state variables: the agent's continuation value and his information rent. The optimal contract uses a combination of nonnegative payments and inefficient liquidation threat to provide the agent incentives. In the beginning, the inefficient liquidation threat is severe, but the expected length of the relationship long, such that the agent's information rent is high. Over time, the information rent decays and continuation value increases as function of the past outcomes. Depending on the past performance, these two processes meet and liquidation at a fixed threshold becomes optimal. In particular, early weak performance leads to a permanent distortion that cannot be undone by performing well in the future.