Collusive Pricing With Incomplete Information

Fudenberg and Maskin (1986) find that any feasible and individually rational payoff can be supported by subgame perfect equilibrium strategies in an infinitely repeated game with complete information, if the discount rate is low enough. An analogous result holds for the case of finitely repeated games (Benoi t and Krishna (1985)).;These results imply that, if a cartel forms, it will have no difficulty maintaining its collusive agreement. The first part of the thesis investigates when agents choose to collude given the benefits of collusion (cooperative payoffs dominate non-cooperative payoffs) and its cost (agents risk government prosecution). We choose the context of a simple bidding model. Buyers at a first price sealed-bid auction decide whether to collude and decide on a bidding strategy. The government can decide to investigate the bidders based on the price fetched by the object. The sequential equilibrium of this one-shot game is semi-separating. Bidders choose to collude with some positive probability. A high winning bid implies that the bidders were acting non-cooperatively; a low winning bid could have been submitted by a cartel or by non-cooperative buyers. The probability of collusion is monotonically decreasing in the number of players.;Given that the results mentioned above say that any feasible payoff can be attained in equilibrium, they do not restrict the set of achievable collusive outcomes. In the second part of the thesis, we ask whether the presence of adverse selection narrows the set of collusive outcomes and payoffs in an infinitely repeated game. We examine a Bertrand duopoly where firms have private information about their costs. Firms communicate their cost to decide on a collusive pricing strategy. The communication is modeled using a direct mechanism. The mechanism prescribes a stationary price strategy based of the cost reports. A class of incentive compatible mechanisms is characterized in terms of price outcomes. The proposed model narrows the set of ex ante payoffs and outcomes with respect to the complete information context. The model, however, replicates a conclusion of the complete information models by predicting that any price can be observed in equilibrium when the firms are identical