An Experimental Comparison of the Fairness Models by Bolton and Ockenfels and by Fehr and Schmidt
by
Dirk Engelmann
Humboldt-Universitaet zu Berlin
Coauthors: Martin Strobel (Humboldt-Universitaet zu Berlin)

We present an experiment to compare the two fairness theories by Bolton and Ockenfels [ERC, AER, forthcoming] and by Fehr and Schmidt [F&S, QJE, 1999]. An impressive number of experimental results can be explained by both theories. However, if one wants to compare the predictive power of these theories, most of the experiments that are interpreted retrospectively are not helpful, since for most experiments, both theories make equal or very similar predictions. Both models rely on inequality aversion to explain deviations from selfish behavior. The fundamental difference between ERC and F&S is that the motivation function of ERC is maximal if the own payoff is equal to the average payoff in the population while F&S assumes a utility function that is decreasing in the differences between the own payoff and each of the other players' payoff. Thus while according to ERC a subject would be equally happy if all subjects received the same payoff or if some were rich and some were poor as long as she received the average, according to F&S she would clearly prefer the first situation. To obtain explicitly opposite predictions by the two theories we chose a very simple game that focuses on their fundamental difference. A subject had to choose an allocation of money between two other subjects, called persons in the experiment. The person who had the choice received a fixed intermediate payoff and chose between three different allocations between a person who received in all allocations more than her and a person who always received less. These allocations were such that whenever she chose one with an average payoff for the other two persons closer to her's, both individual payoffs became more distant from her's. ERC predicts that she chooses the allocation that is most unequal between the other two persons. The opposite allocation is predicted by F&S to be chosen. To prevent interference of preferences for efficiency with the objective of our experiment, we designed two treatments, one in which following the prediction by ERC leads to a maximization of total payoff, one where maximization of total payoff is in line with the F&S prediction. All subjects had to choose an allocation, while groups of three were randomly formed later on. Subjects knew that their decision would only matter if they were assigned to be person 2 in their group and that their payoff was fixed in that case, such that their decision could never influence their own payoff. In the treatment where the prediction of F&S leads to a maximization of total payoff the results clearly confirm this prediction. In the other treatment subjects chose in about equal proportions the two extreme allocations and some chose the intermediate allocation. Hence in our simple decision task the performance of F&S is much better than that of ERC, although both theories ignore the importance that subjects assign to efficiency.

Date received: May 2, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Mathematical Conference Abstracts. Document # caez-90.