Randomization is optimal in the robust principal-agent problem

A principal contracts with an agent, who takes a hidden action. The principal does not know all of the actions the agent can take and evaluates her payoff from any contract according to its worst-case performance. Carroll (2015) showed that there exists a linear contract that is optimal within the class of deterministic contracts. This paper shows that, whenever there is an optimal linear contract with non-zero slope, the principal can strictly increase her payoff by randomizing over deterministic, linear contracts. Hence, if the principal believes that randomization can alleviate her ambiguity aversion, then restricting attention to the study of deterministic contracts is with loss of generality.(c) 2022 Elsevier Inc. All rights reserved.