Some doubts about measuring self-interest using dictator experiments: the costs of anonymity

Traditionally, economists have assumed self-interest governs economic choic es. Recently, some social scientists and economists, especially those worki ng in game theoretic and experimental areas, have begun to treat self-inter est as a testable hypothesis. One important vehicle for evaluating self-int erest has been a class of experiments called 'dictator' experiments. We bel ieve that these experiments may have a flaw in their design which leads res earchers to overstate, systematically, the role of self-interest in individ uals' motivations. Double-blind experiments, designed to create conditions of privacy and anonymity, may engender doubts in subjects regarding the exi stence of pairings and the disposition of any money they share. Moreover, s ubjects may view the experiment as a game. We test these conjectures using both traditional and modified dictator experiments. (C) 2001 Elsevier Scien ce B.V. All rights reserved.