Inequality Aversion and Reciprocity in Moonlighting Games

Dirk Engelmann,Martin Strobel

doi:10.3390/g1040459

Abstract

We study behavior in a moonlighting game with unequal initial endowments. In this game, predictions for second-mover behavior based on inequality aversion are in contrast to reciprocity. We find that inequality aversion explains only few observations. The comparison to a treatment with equal endowments supports the conclusion that behavior is better captured by intuitive notions of reciprocity than by inequality aversion. Extending the model by allowing for alternative reference points promises better performance, but leads to other problems. We conclude that the fact that inequality aversion often works as a good short-hand for reciprocity is driven by biased design choices.

Highlights

Models of inequality aversion, notably the Fehr-Schmidt model and the Bolton-Ockenfels model, have been able to organize experimental data from games involving interactions between players, such as results from ultimatum and gift-exchange games [1,2]
We find that while inequality aversion plays some role in moderating reciprocity, it has little explanatory power when it is inconsistent with reciprocity
The experiment we report on deviates from the typical experimental design choices where reciprocity, when it matters, coincides with inequality aversion

Summary

Introduction

Notably the Fehr-Schmidt model and the Bolton-Ockenfels model, have been able to organize experimental data from games involving interactions between players, such as results from ultimatum and gift-exchange games [1,2]. The action of the first mover to do nothing leaves payoffs roughly equal and reciprocating kindness (sending money) or unkindness (taking money) coincides with inequality aversion. If Player 1 takes something from Player 2, inequality aversion and a notion of reciprocity that considers sending money as kind and taking money as unkind, suggest different actions. The kindness of Player 2 depends on the expectations of Player 1 (or more precisely on Player 2’s second-order belief regarding Player 1’s expectation) This means that “kinder” actions do not necessarily imply higher transfers (or lower punishments), but they do under the plausible assumption that Player 1 expects to be treated better or equal in Treatment 2 than in Treatment 1.

Results

Result

5: The qualitative predictions derived for the Dufwenberg-Kirchsteiger and

Discussion and Concluding

Will you be able to choose more than once?

Will person 2 be informed about the choice of person 1 before making a choice?