Abstract While many experiments demonstrate that behavior differs from the predictions of traditional economic theory, they have not shown that economic reasoning is necessarily incorrect. Instead, these experiments illustrate that the preferences of homo economicus have been mis-specified. Modeled with social preferences, it may be rational for agents to forego material gains. Social dilemmas are examples in which punishment is not credible and yet people often pay to reprimand other participants. At the same time, we show that these people also react to changes in the price of punishing and income as if punishment was an ordinary and inferior good.