Optimal Delegation with Multiple Agents

Abstract

This paper investigates the issue of optimal delegation in the presence of multiple agents. Specifically, we analyze the design of decision rules by a principal who relies on two biased agents to inform her decision. Each agent’s type/information is private information. The principal would like to implement an action that matches the state of nature, where the state of nature is the sum of the agents’ types/information. The principal is unable to use monetary transfers, and the information the agents possess is complementary and non-overlapping. The principal is able to commit to a decision rule that maps the agents’ unverifiable reports into a single dimensional decision. The solution concept used is that of ex post implementation. We show the optimal robust mechanism can be implemented via a sequential delegation rule. According to the rule, the principal allows the first agent (chosen at random) to either choose an action from a delegation set or delegate the decision to the second agent, who in turn chooses an action from a delegation set.