Abstract

Repeated games of cooperation share the same equilibrium selection problem as coordination games. In both settings, providing information might help players coordinating on efficient equilibria. Which equilibrium is most likely to be selected strictly depends on the type and the amount of information provided. It is then natural to ask under which conditions providing more information increases efficiency. The present thesis makes a step in answering this question. It analyzes how the presence of information regarding either the opponent, or the options that are available for choice, might change players' behavior. It focuses on two settings where increasing information might be detrimental for players: a repeated Prisoner's dilemma, and a coordination game. The first chapter develops a theoretical model in which players have limited information about the opponents' previous moves. When applied to the Trust Game, we show that by increasing the amount of information disclosed to the first player, more exploitative equilibria appear, in which that player obtains a smaller payoff. These equilibria disappear in settings in which the information the first player obtains about the second player's past behavior is limited. This is a case in which providing a player more information may reduce his payoff in equilibrium. In the second chapter, we test this latter result with a laboratory experiment, and we show that subjects do understand that different behavior might be optimal in different settings. Subjects tend to use a fully cooperative strategy more often when only minimal information is available. Moreover, subjects trying to exploit the opponent succeeded in gaining more than the mutual cooperation payoff only when the information provided to the opponent is sufficiently rich, that is when our model predicts that exploitative outcomes are equilibria. The last chapter considers the effects of introducing information about the options available for choice in a coordination game. It reports the results from a simulated crowdfunding experiment. We show that the presence of non payoff-relevant information is able to make a project focal. However, when returns from coordination are uncertain, the presence of information is instead detrimental for coordination.