Flexibility and Reputation in Repeated Prisoners' Dilemma Games

We study the role that the option to terminate a relationship has on cooperation in a repeated prisoner's dilemma. While cooperation is, in theory, sustainable with or without this option, we show experimentally that cooperation rates are significantly lower with the option to terminate. Rather than punishing a defection, most subjects choose to terminate the relationship, which increases the temptation to behave opportunistically. However, we show that introducing a reputation mechanism, through which signals of past cooperative behavior are given to future matches, can substantially increase cooperation rates, in some cases to a level higher than when the option to terminate the relationship is absent. Our results show that an objective, long-lasting measure of reputation is the most effective in promoting cooperation, but short-lasting or subjective but long-lasting reputation mechanisms also increase cooperative behavior. Moreover, reputation mechanisms generally lead to stable cooperation rates over time, in contrast to the declining cooperation rates observed in their absence when subjects are free to terminate relationships. Finally we show that the option to terminate a relationship acts as a sorting mechanism between subjects who cooperate frequently and those who do not, and that reputation mechanisms further enhance this sorting effect