Doctoral Dissertation Research: Cooperation in Stochastic Games: A Prisoner's Dilemma Experiment

New York University, New York NY

Investigators

Abstract

One question that economists and social scientists have contemplated for some time is what are the proper prerequisites for societal cooperation when private incentives conflict with social ones? One answer to this question comes from the theory of repeated games, the study of decision making amongst members of a group (the game) where the decision is made over and over again. More precisely, the specific game associated with this answer is the repeated prisoner's dilemma. In the prisoner's dilemma, cooperation is socially beneficial but it is in the private best interest of each player to not cooperate and thus, when only played once, no cooperation occurs. However, when the prisoner's dilemma is repeated, cooperation can be sustained by linking future choices to current choices so that the gain that one would obtain by deviating from cooperation in the present period is offset by a loss in future periods. Specifically, the gain is deterred by the credible threat of mutual non-cooperation by all players in the future. Repeated games, however, are a drastic simplification of the real world. Realistically, while many economic situations feature repeated interaction, the nature of the interaction changes through time. This feature is captured by stochastic games, a generalization of repeated games in which the game played each period changes and is determined, at least partially, by chance. This project is a laboratory experiment in the field of stochastic games. In particular, the experiment utilizes a repeated prisoner's dilemma in which the payoffs of the prisoner?s dilemma are stochastically determined each period. Most importantly, given the link between future choices and current choices that is established to sustain cooperation, altering the nature of future choices is a substantial step towards understanding how people behave in these environments. From a broader perspective, the main economic example for the stochastic prisoner?s dilemma in the experiment is collusion among firms over the business cycle. In this environment, firms desire collusion, cooperation on non-competitive prices or quantities, and compete over many periods. The demand fluctuates because of booms and busts in the economy, which changes the exact interaction they face each period. This experiment can potentially guide antitrust law by providing insight into conditions where firms may find it easier to collude.