Asymmetric Information and Cooperation

Abstract

This thesis investigates the theory of cooperative behavior in the presence of asymmetric information. Traditionally, the core has been a powerful and much used solution concept to describe cooperative outcomes. In settings where agents have some private information, it may be appropriate to include the opportunity for communication in the development of the core. I study the relationship of various core solution concepts with prevalent noncooperative solution concepts for environments with asymmetric information. The core definitions examined vary by the level of communication assumed. In Chapter 2, I investigate the welfare properties of market equilibria. I demonstrate that appropriate communication restrictions can be placed on the core (and efficiency) in order to obtain first and second welfare theorems. In Chapter 3, I discuss the Bayesian implementation of core solutions. If full communication is assumed, Palfrey and Srivastava (1987) have shown that the core is not Bayesian implementable: a game cannot be constructed that has only core allocations as its equilibria. I demonstrate that communication restrictions on the core are sufficient to obtain positive Bayesian implementation results in the environment studied by Palfrey and Srivastava. In other words, a game can be constructed that entices noncooperative players to choose strategies that are cooperative under limited communication. In Chapter 4, I examine cooperation between bidders in private value, sealed bid auctions. I assume that bidders can overcome their one period temptation to break any collusive agreement, and that they attempt to formulate a collusive mechanism. However, each bidder's valuations are still his own, private information. If he is not given the proper incentives, he may lie about his values in order to increase his profits. Therefore, any collusive mechanism must be incentive compatible and is likely to be, at a minimum, interim efficient. I demonstrate that the theory provides some predictions about the set of collusive mechanisms chosen by bidders and that, when moving to a setting where multiple objects are for sale, the set of feasible collusive mechanisms grows. When multiple objects are for sale, there exist incentive compatible mechanisms that are preferred by all bidders to the only incentive compatible mechanisms in the single object case. Laboratory experiments indicate that these predictions are often consistent with actual behavior. However, deviations by some bidders suggest some weaknesses in this approach.