Abstract

This paper reports the results of experiments constructed to test the effects of demand information (knowledge of the demand function or schedule) and monitoring (knowledge of rivals' actions) on collusion in infinitely repeated games. In addition, the treatments encompass the assumptions of two highly influential theories (Green and Porter [1984] and Rotemberg and Saloner [1986]), thereby allowing for an internal validity test of each theory. Given the large number of theoretical results and the multiplicity of equilibria in infinitely repeated games, this last exercise is important towards identifying the empirically plausible equilibria from the set of theoretically possible equilibria. Results indicate that monitoring may be a more critical factor than demand information in facilitating collusion. The Rotemberg and Saloner theory appears to be well supported by the data, whereas the equilibrium devised by Green and Porter does not emerge even in favorable environments; the latter result appears to be due to the uncertain environment set forth by the model's assumptions.