Identifying the Reasons for Coordination Failure in a Laboratory Experiment

In this paper, we use a laboratory experiment to investigate the effect of absence of common knowledge on the outcomes of coordination games. We introduce cognitive types into a pure coordination game in which there is no common knowledge about the distribution of cognitive types. In our experiment, around 76% of the subjects managed to coordinate on the payoff-dominant equilibrium despite the absence of common knowledge. However, around 9% of the players had first-order beliefs that lead to coordination failure and another 9% exhibited coordination failure due to higher-order beliefs. Furthermore, we compare our results with predictions of different models of higher-order beliefs, commonly used in the literature.