A behavioral theory of equilibrium selection

Theories about unique equilibrium selection are often rejected in experimental investigations. We drop the idea of selecting a single prominent equilibrium but suggest the coexistence of different beliefs about "appropriate" equilibrium or non-equilibrium play. Our main selection criterion is efficiency applied to all or only to "fair" equilibria. This assumption is applied to 16 Binary Threshold Public Good games where at least k of four homogeneous or heterogeneous players have to incur fixed costs in order to produce a public good. The case k=4 is the Stag Hunt game which is most often used to test equilibrium selection. Our finite mixture model applies with the same parameters (shares of populations, altruism parameters) to the four thresholds k=1,2,3,4. The estimated shares of populations are similar in four treatments with identical or different cost/benefit ratios of the players. Our results for k=4 clearly contradict selection by Risk Dominance and Global Games. In the two (almost) symmetric treatments the Harsanyi/Selten selection explains 40% of the decisions.