Optimal Threats and Efficient Mechanism Design

This paper studies Bayesian efficient mechanism design in environments where agents' utility depend on the chosen alternative even if they do not participate. A VCG mechanism maximizes the expected revenue of the planner among efficient mechanisms that induce the truthful participation of all the agents. The class of VCG mechanisms considered in this paper extends the one in Krishna and Perry (2000). In addition to a basis (whose components are critical types of the agents) the planner must choose appropriate threats to agents who do not participate. I show that the solution of the design problem can be found by a max-min choice of the critical types and threats