Estimating Reaction Functions in Experimental Duopoly Markets

It is well known that, in general, there are a multitude of supergame equilibria in noncooperative duopoly markets, suggesting the inability (without severe restrictions) of theory alone to determine the 'best' strategy in a repeated game context. Axelrod's prisoner's dilemma simulation tournaments have led to the view of the somewhat cooperative 'tit-fortat' approach as an attractive strategy, in particular compared to an alternative strategy of 'always defecting' (choosing at all times the single period Nash solution). In this paper, we use data obtained from two independent posted-offer duopoly experiments to investigate the actual dynamic reaction functions of participants. Neither of the above seems to be commonly employed as a pure strategy.We also provide some support for the Axelrod view of 'tit-for-tat' as the most profitable strategy.