| The common focus of the three studies in this dissertation is the tension between cooperative, efficiency-enhancing behaviors and the narrowly self-interested equilibrium behavior predicted by standard game theory. In recent years, the tenets of the latter have been challenged by empirical findings that indicate that humans are more cooperative than traditional normative theory predicts. Basic concepts and predictions of game theory and past research on cooperative tendencies are reviewed. The first experimental study here involves a two-person one-shot extensive-form trust game. The major finding is that the Mach-IV personality test can help screen reciprocators from non-reciprocators. In the second study, subjects in a standard voluntary Contribution Mechanism (VCM) are rematched at each trial. If rematched at random, contributions decline over rounds as in many other VCM experiments. If, however, subjects, without their knowledge, are rearranged into groups according to the amount of their contribution, contribution rates remain higher because cooperators do not get discouraged by free riders. This finding is accounted for by the goodwill accounting framework of McCabe and Smith (1999). In the third experimental study, a standard VCM is embedded in intergroup conflict as groups compete for a fixed prize that is divided among the members of the winning group, either equally or proportionally to individual members' contributions. A group's wining probability depends on how its public good contributions compare to the contributions of the competing group. Given specific parameterization, non-contribution in such a two-level interaction is no longer the dominant strategy, even though it would be in the corresponding one-level ACM. Further, equilibrium contributions are higher under proportional than under equal prize division. On the aggregate level, the data support equilibrium predictions, even though there is some evidence of over-contribution under group competition and individual contribution patterns are very idiosyncratic. A simple reinforcement learning model (See Roth and Erev, 1995) accounts well for the change in aggregate data over trials. The implications of all three studies are that cooperativeness can be enhanced by the means of screening, social structure and incentive schemes. |
