Coordination and Self-Organization in Minority Games: Experimental Evidence

This work presents experimental results on a coordination game in which agents must repeatedly choose between two sides, and a positive fixed payoff is assigned only to agents who pick the minoritarian side. The game presents a variety of asymmetric pure strategy equilibria, and a unique symmetric mixed-strategy equilibrium in which agents randomize between the two sides at every stage. The game reflects some essential features of those economic situations in which positive rewards are assigned to individuals who behave in opposition to the modal behavior in a population. We conduct laboratory experiments in which stationary groups of five players play the game for 100 periods, and manipulate two treatment variables: the amount of "memory" M that players have regarding the game history (i.e., the length of the string of past outcomes that players can see on the screen while choosing), and the amount of information about other players' past choices: in the aggregate information treatment, players only know which side was the minority side at each period, while in the full information treatment players have information regarding the entire distribution of choices in the group at each round. We first analyze aggregate results in terms of both "allocative" and "informational" efficiency. We then analyze individual behavior in the game as compared to the theoretical benchmark provided by the mixed strategy equilibrium solution. Our results show that, first, both allocative and informational efficiency are higher on average than the benchmark value corresponding to the mixed strategy equilibrium in all treatments, suggesting that a quite remarkable degree of coordination is achieved; second, providing players with full information about other players' choice distribution does not appear to improve efficiency significantly. At the individual level, a substantial portion of subjects exhibit `inertial' behavior, i.e., the tendency to replicate their previous round choice with a higher frequency than the one prescribed by randomizing behavior, and such inertia seems to be enhanced rather than decreased by a full information treatment.