Stochastic learning in co-ordination games : a simulation approach

In the presence of externalities, consumption behaviour depends on the solution of a co-ordination problem. In our paper we suggest a learning approach to the study of co-ordination in consumption contexts where agents adjust their choices on the basis of the reinforcement (payoff) they receive during the game. The results of simulations allowed us to distinguish the roles of different aspects of learning in enabling co-ordination within a population of agents. Our main results highlight: 1. the role played by the speed of learning in determining failures of the co-ordination process; 2. the effect of forgetting past experiences on the speed of the co-ordination process; 3. the role of experimentation in bringing the process of co-ordination into an efficient equilibrium.