Selecting Correlated Random Actions

In many markets, it can be beneficial for competing firms to coordinate their actions. However, such arrangements usually have the problem that nobody has an incentive to adhere to the agreement. In this paper we investigate a way for two companies to communicate together and agree on a coordinated strategy in such a way that both participants have an incentive to keep to the agreement. We provide a more efficient solution to the game theoretic problem solved by Dodis, Halevi and Rabin in [DHR00]: two selfish rational parties want to select one of a list of pairs, according to some probability distribution, so that one party learns the first element and the other learns the second, and neither gains any other information. In game theory terms, the problem is to achieve a correlated equilibrium without the trusted mediator. Our solution is more efficient than [DHR00] in terms of the probability distribution.