Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Antisocial punishment in public good games, i.e., punishment of individuals who contributed the same or more than their punisher, varies substantially across cultures. We exploit the data of Herrmann et al. (2008) and implement an empirical strategy based on a finite mixture model to uncover the heterogeneity behind this variation. After we exclude 17.7% Type N subjects who never punish, estimating the finite mixture model reveals two distinct punisher types among the remaining subjects: 30.3% Type AF subjects who engage in antisocial punishment as well as free rider punishment and 52.0% Type F subjects who engage exclusively in free rider punishment. Moreover, we find that in cultures with high levels of antisocial punishment, Type AF subjects are more frequent. Despite the parsimony of this classification, the number of Type AF in a group predicts mean earnings in that group and enhances our understanding of the large variation in the effectiveness of peer punishment across cultures.