Cooperation norms in multiple-stage punishment

Carpenter and Matthews (2009) examine the cooperation norms determining people's punishment behavior in a social-dilemma game. Their findings are striking: absolute norms outperform the relative norms commonly regarded as the determinants of punishment. Using multiple punishment stages and self-contained episodes of interaction, we disentangle the effects of retaliation and norm-related punishment. An additional treatment provides data on the norms bystanders use in judging punishment actions. Our results partly confirm the findings of Carpenter and Matthews: only for the punishment-related decisions in the first iteration is the absolute norm outperformed by the self-referential norm set by the punisher's own contribution. For the decisions in all later iterations, as well as for bystanders' support in all iterations, the absolute norm organizes our data best. In contrast to the study by Carpenter and Matthews, we find an absolute norm of 3=4 of players' endowments to be both consistent across decisions and relatively stable over time.