The Role of Demand Information and Monitoring on Tacit Collusion

This paper studies infinitely repeated games with stochastic demand in an experimental environment. The lab experiments are constructed to test the effects of demand information (knowledge of next period’s demand realization) and monitoring (knowledge of rivals’ actions) on collusion. Two treatments are motivated by the corresponding assumptions in the Green and Porter (1984) and Rotemberg and Saloner (1986) models: the former assumes imperfect knowledge of next period’s demand realization and imperfect monitoring of rivals’ actions whereas the latter assumes perfect knowledge of next period’s demand realization and perfect monitoring. A third, intermediate, treatment allows for imperfect demand information (as in Green and Porter) but perfect monitoring (as in Rotemberg and Saloner). Results indicate that for a sufficiently high discount rate, demand information seems to facilitate collusion more than monitoring, especially as subjects gain experience. While the data exhibits behavioral patterns consistent the theoretical predictions of the two motivating models, a robust finding is that subjects in the Rotemberg and Saloner treatment cooperate in the way predicted by this theory: collusion falls dramatically in anticipation of unusually large demand but returns to high levels otherwise. Our results suggest that tacit collusion could arise in stochastic environments and that it could be fairly elaborate