EconCS Whiteboard Talk Series

Robust Incentive Contracting

I will cover two related papers that study a robust version of the principal-agent model, a workhorse of information economics. In the basic version of this model, an agent can produce output, which is valuable for the principal, but only by exerting effort. The principal can incentivize effort by paying the agent as a function of the output he produces; her problem is to choose this function optimally.

I will first give some general background on the principal-agent problem and its applications, and then will present the robust version (the first paper). In this version, the principal is uncertain about exactly what the agent can do, and uses a worst-case criterion to evaluate possible incentive contracts. Unlike the traditional model, the robust version quite generally predicts that the optimum is a linear contract, in which the agent is paid a constant fraction of the output he produces.

In the second paper, I will apply this same robust modeling framework to a problem in which the agent's effort produces valuable information, rather than producing output directly. This combines the principal-agent problem with a mechanism design problem, since incentives must be given both for exerting effort and for truthfully reporting the information obtained. Linear contracts have a natural formulation in this model, but the robust optimum turns out to be similar to, but not exactly, a linear contract.