This paper provides novel results for the extensive literature on dictator games: recipients do not expect dictators to behave selfishly, but instead expect the equal split division. We performed a field experiment in Baja California among a population of unexperienced subjects. Using monetary incentives we find that only 10 percent of subjects correctly guessed the expected Nash equilibrium payoff (zero). In sharp contrast, the modal subject predicts the equal split. The predictions made by dictators are notably different: 45% predicted the zero contribution and 40% expected the equal split. Surprisingly, their actions are uncorrelated with their predictions: they choose a donation in the interior of the interval. We conjecture that the equal split is the natural solution to the problem but because the dictators are involved, they also consider the chance of keeping the complete pie for themselves. Dictators solve the puzzle by passing a positive amount of money wh ich reflects the tension between fairness and self-interest. In consequence, any giving smaller than the equal split division may not be considered altruistic behavior. Only a donation larger than the 50/50 split would reflect generosity.

Financial support from MCI SEJ2007-62081/ECON) and Excelencia—Junta (P07-SEJ-0254).