Abstract

When decisions are made before roles are assigned, the Dictator Game is strategically equivalent to a linear Public Goods Game. This suggests that, when played between individuals with the same income, the prosocial behavior observed may be attributed at least in part to reciprocal altruism. Dictators transfer money only because they believe Recipients would transfer money as well, if roles were reversed. By contrast, when the game is played between individuals with different background income, the generosity of the rich towards the poor is more easily attributed to pure, non-reciprocal altruism. We test this hypothesis by eliciting conditional preferences for giving in a Dictator Game in two treatments. In the first students are matched with other students, while in the second students are matched with subjects living in a refugee camp in Uganda. We find that our predictions are only partially borne out by the data. Whether giving is directed to a person with similar or lower socioeconomic status, most subjects reveal conditionally altruistic preferences. Unconditional altruism is virtually absent in both treatments. These counter-intuitive results have important implications for the experimental elicitation of social preferences.