The relevance of private information in mechanism design

Recent results in mechanism design show that as long as agents have correlated private information and are sufficiently risk neutral, it is possible to design mechanisms that leave agents with arbitrarily small information rents. We show that these full-rent-extraction results hinge on the implicit assumption that the agents’ beliefs uniquely determine their preferences. We present an example of the voluntary provision of a public good in which this assumption is relaxed, and consequently, even in highly correlated environments, if agents’ beliefs do not uniquely determine their preferences, then the extraction of the agents’ entire information rents is impossible.