Fairness in Extended Dictator Game Experiments

We test the robustness of behavior in dictator games by offering allocators the choice to play an unattractive lottery. With this lottery option, mean transfers from allocators to recipients substantially decline, partly because many allocators now keep the entire endowment for themselves (without playing the lottery). In our standard dictator game, the median transfer amounts to 41% of the dictators' endowment. Once the lottery option is present, the median transfer falls to zero. Introducing an additional unattractive choice thus leads subjects to violate the weak axiom of revealed preference (WARP).