Structure d'interactions et problème de coordination : une approche expérimentale

We report the results of an experiment that examines play in a 50 periods repetition of a two-player coordination game (Stag Hunt game), which admits two pure strategy Nash equilibria that are Pareto-ranked : a payoff-dominant equilibrium and a risk-dominant equilibrium. We consider a 2x3 factorial design, with two different matching rules -global and local interaction- and three sizes for the basin of attraction of the risk-dominant equilibrium. Under global interaction, each player can be matched in each period with any player in the population. Under local interaction, each player can be matched only with his two closest neighbours. Our results confirm earlier experimental results obtained under global interaction : i) non-equilibrium outcomes are rarely observed, ii) the first period modal choice is the payoff-dominant strategy, and iii) the first period play determines strongly the steady-state which will be reached, as it generally lies in the same basin of attraction as the initial state. The two first stylised facts are also observed under local interaction but unlike the dynamics observed under global interaction, the convention towards which choices converge is strongly dependent on the size of the basin of attraction of the risk-dominant equilibrium. We observe convergence towards the payoff dominant convention when the two basins of attraction are of the same size and towards the risk dominant convention for the largest of the three basins of attraction of the risk dominant equilibrium. We found no evidence for a faster convergence under local interaction than under global interaction.