On the Credibility of Punishment in Repeated Social Dilemma Games

Various experimental studies have shown that the availability of a punishment option can increase the prevalence of cooperative behaviour in repeated social dilemmas. A punishment option should only matter if it is a credible threat. We investigate if the degree of credibility depends on standard strategic equilibrium considerations (i.e. SPNE or NE logic) or stems from a non-strategic motivation such as reciprocity. We find that for punishment to be credible non-strategic motivations are sufficient and that subgame perfection does not further improve credibility.