From imitation to collusion: Long-run learning in a low-information environment

We explore the stability of imitation in a 1,200-period experimental Cournot game where subjects do not know the payoff function but see the output quantities and payoffs of each oligopolist after every period. In line with theoretical predictions and previous experimental findings, our oligopolies reach highly competitive levels within 50 periods. However, already after 100 periods quantities start to drop and, eventually fall deep into collusive territory without pausing at the Nash equilibrium. Our results demonstrate how groups of subjects can learn their way out of dysfunctional heuristics, and thus suggest the need for a new theory of how cooperation emerges.