A Principal-Agent Model of Sequential Testing

This paper analyzes the optimal provision of incentives in a sequential testing context. In every period the agent can acquire costly information that is relevant to the principal's decision. Neither the agent's effort nor the realizations of his signals are observable. First, we assume that the principal and the agent are symmetrically informed at the time of contracting. We construct the optimal mechanism and show that the agent is indifferent in every period between performing the test and sending an uninformative message which continues the relationship. Furthermore, in the first period the agent is indifferent between carrying out his task and sending an uninformative message which ends the relationship immediately. We then characterize the optimal mechanisms when the agent has superior information at the outset of the relationship. The principal prefers to offer different contracts if and only if the agent types are sufficiently diverse. Finally, all agent types benefit from their initial private information