From Coordination to Double-crossing: Experiments on the Strategic Behavior of Groups

We explore the empirical relevance of coalitional deviations in strategic games. To do this, we implement a series of experiments with and without unstructured, nonbinding, preplay communication. The study is based on three-player games. One of the games has a unique strong Nash equilibrium and two coalition-proof Nash equilibria. Two of the games have a unique strong and a coalition-proof Nash equilibrium. We find clear evidence that communication affects play, but not in the way theory predicts. While in games with no communication the strong Nash equilibrium is played most often, in games with communication players move away from either strong or coalition-proof Nash equilibria. The failure of the concepts does not seem to be due to some strategy profile being more attractive a priori. Players agree to play a coalitionally dominated equilibrium even when egalitarian Pareto efficient credible deviations are available. This does not seem to be the result of a lack of understanding of the games either. Subjects learn to coordinate on self-enforcing equilibria and coalitional deviations. Indeed, we find evidence of double-crossing. Players propose deviations from which they attempt to deviate again with the help of a player not involved in the original deviation. Our results show the importance of coalitional reasoning in strategic games, but also the need to adjust theories to incentive-compatible information transmission.*