Efficient Design with Interdependent Valuations and an Informed Center

Abstract

We consider mechanism design in social choice problems in which agents' types are mutually payoff-relevant, multidimensional, and take on a continuum of possible values. If the center receives a signal that is stochastically related to the agents' types and direct returns are bounded, for any decision rule there is a balanced transfer scheme that ensures that any strategy that is not arbitrarily close to truthful is dominated by one that is. If direct returns are also continuous, truthful revelation becomes a nearly dominant strategy, all Bayes-Nash equilibrium strategies are nearly truthful, and at least one such strategy exists. If the center's information is not informative but agents' types are stochastically related, then there are balanced transfers under which truthful revelation is a Bayesian epsilon-equilibrium, again for any decision rule. Analogous results hold when agents also take mutually payoff-relevant actions in advance of any action by the center.