Action revision, information and collusion in an experimental duopoly market

We report on an experiment designed to study a dynamic model of quantity competition where firms continuously revise their production targets prior to the play of the "one-shot" game. We investigate how the observability of rival firm's plans and the technology for implementation of revised actions affect market competitiveness. Under a real-time revision game where payoffs are determined only by the quantities prepared at the end, play converges to the Cournot-Nash output when rival's plans are unobservable. If plans cannot be hidden from competitors, choices are even more competitive than the static Nash equilibrium, thereby showing a negative value of information with lower profits. With stochastic revision, where opportunities to revise arrive according to a Poisson process and the quantities selected at the last opportunity are implemented, collusion is much frequent. This shows, more generally, that cooperation can be observed even when individuals interact only once