Abstract
Costly third party punishment has been interpreted as a tool for studying the enforcement of social norms. Experiments on this topic typically involve a third party observer who can pay to decrease the payoff of one player who has behaved selfishly (or generously) toward another. We investigate whether third parties are sensitive to the number of players affected by this selfish or generous action. We allowed dictators to be selfish, fair, or generous, and unlike in other experiments, third parties could both punish and reward. Across all variations, responses followed a consistent and intuitive pattern: selfish behavior was punished while generous behavior was rewarded. Third party response was more pronounced when the dictator transfer had the nonrivalrous character of a public good, in the sense that both (a) the number of recipients increased and (b) the dictator's transfer was multiplied by a constant factor, so that the larger number of recipients did not reduce potential payoff of each recipient. Third party response did not change significantly when either of these manipulations was performed alone, suggesting a specific response to situations involving a public good.