ANONYMITY VERSUS PUNISHMENT IN ULTIMATUM BARGAINING

Previous investigations have shown that laboratory play of the ultimatum game differs from the perfect equilibrium prediction. The anonymity hypothesis attributes this to a distortion of subject objectives caused by the act of experimental observation. An alternative hypothesis attributes the phenomenon to the willingness of some subjects to punish those who treat them ''unfairly,'' independent of any experimenter influence. We test these hypotheses. In a control cell, 30% of play is in equilibrium. In a second cell, played under strong experimenter-subject anonymity conditions, 46% of play is in equilibrium. A third cell, in which subject capacity to punish has been removed, exhibits equilibrium play approaching 100%. The evidence supports the conclusion that the punishment hypothesis explains much more of the deviation from perfect equilibrium than does the anonymity hypothesis. (C) 1995 Academic Press, Inc.