Frequent monitoring in dynamic contracts

Abstract

I consider a dynamic principal-agent model in which the agent does not observe the quality of his output and the principal chooses how much to monitor the agent. Monitoring improves production outcomes but can reduce the agent's incentive to work if it uncovers bad performance and leads to punishing the agent. The optimal monitoring intensity is path dependent: an agent who performed poorly in the past is monitored less going forward to reduce the risk of early termination. Conversely, an agent who performed well is monitored more as his accumulated promised compensation provides a buffer against bad monitoring outcomes and reduces the associated penalties by deferring the pay-for-performance risk to states of the world in which termination costs are low.