The Big Robber Game

We present a novel design measuring a correlate of social preferences in a high-stakes setting. In the Big Robber Game, a "robber" can obtain large personal gains by appropriating the gains of a large group of "victims" as seen in recent corporate scandals. We observed that more than half of all robbers take as much as possible. At the same time, participants displayed standard, prosocial behavior in the Dictator, Ultimatum, and Trust games. That is, prosocial behavior in the small is compatible with highly selfish actions in the large, and the essence of corporate scandals can be reproduced in the laboratory even with a standard student sample. We show that this apparent contradiction is actually consistent with received social-preference models. In agreement with this view, in the experiment more selfish robbers also behaved more selfishly in other games and in a donation question. We conclude that social preferences are compatible with rampant selfishness in high-impact decisions affecting a large group.