Modeling Implicit Collusion Using Coevolution

Many oligopolies operate as a repeated game. In such circumstances, it can be expected that profit-maximising participants may engage in implicit collusion to profitably increase spot market prices. This paper models the emergence of such implicit collusion in a stylised market model using a coevolutionary approach. Players bid supply functions made up of a finite number of linear pieces. Each player uses a genetic algorithm to find state-based strategies depending on the price and demand in the last period and the predicted demand in the next period. We consider a symmetric duopoly and demonstrate that collusive behaviour can be learned even when there is very limited information available to the participants. Moreover, we show a type of implicit collusive behaviour that occurs even though the system does not settle into a stable equilibrium. We use a wholesale electricity market, in which supply function bids are typical, as a motivating example throughout this paper.