Project selection with partially verifiable information

Abstract

We study a principal–agent project selection problem with asymmetric information. The principal must choose exactly one of $N$ projects, each defined by the utility it provides to the principal and to the agent. The agent knows all the utilities, and the principal can commit to a mechanism (without transfers) that maps the agent’s report about the utilities to a chosen project. Unlike the typical literature, which assumes the agent can lie arbitrarily, we examine the principal’s problem under partial verifiability constraints. We characterize the class of truthful mechanisms under a family of partial verifiability constraints and study the principal’s problem for the specific cases of no-overselling and no-underselling. Our results suggest significant benefits for the principal from identifying or inducing such partial verifiability constraints, while also highlighting the simple mechanisms that perform well.