Robust Incentive Contracting

Abstract

We present a simple approach to robust contracting in uncertain environments. First, we consider a basic principal-agent problem with risk-neutrality and limited liability. The principal knows some actions available to the agent, but other, unknown actions may also exist. The principal evaluates contracts by their worst-case performance, with respect to the actions that may or may not be available. It turns out that under very general conditions, linear contracts are optimal. Second, we apply the worst-case approach to a problem where the agent is an expert, whose job is to help the principal make a decision by obtaining relevant information. The expert's payment can be made contingent on his reported information and the state of nature, which is revealed ex post. Here the optimum is a "restricted investment contract": the expert can choose from a subset of the principal's decisions, and is paid proportionally to what his chosen decision would have earned in the realized state.