Implementing randomized allocation rules with outcome-contingent transfers

Abstract

We study a mechanism design problem where the allocation rule is randomized and transfers are contingent on outcomes. In this problem, an agent reports his private information, and an exogenous randomized allocation rule assigns an outcome based on the report. A planner designs an outcome-contingent transfer to incentivize the agent to report truthfully. We say that the allocation rule is implementable if such transfers exist. For this implementation problem, we derive two sufficient and necessary conditions. Each has a geometric interpretation. Moreover, when the allocation rule is implementable, we construct transfers that implement the allocation rule.