The roles of level-k and team reasoning in solving coordination games

Level-k and team reasoning theories, among others, have been used to explain experimental evidence on coordination games. Both theories succeed in explaining some results and both fail in explaining other results. Sometimes it is impossible to discriminate between them. For this reason we propose an experiment with pie games, similar to the ones used by Crawford et al. (2008). We observe subjects playing a series of coordination games, with different configurations of equality and Pareto-dominance, for which it is possible to provide clear predictions derived from both team reasoning and a particular cognitive hierarchy model: level-k theory. In line with previous experimental results, we find that each theory fails to predict observed behaviour in some games. However, because of the design of our experiment, we can go deeper into the matter. Our results show that Pareto dominance, fairness and uniqueness are good predictors for coordination choices. Secondly, we find mixed evidence about level-k and team reasoning theories. In particular team reasoning theory fails to predict choices when they picks out a solution which is Pareto dominated and not compensated by grater equality; Level-k theory fails in games in which it predicts the choice of one of not unique slices, and the unique choice is more equal than the alternative choices. This could represent a step forward to investigate the presence of team reasoning or level-k in coordinating behaviour