Selfish in the end?

An investigation of consistency and stability of individual behaviour

Abstract

This paper puts three of the most prominent specifications of 'other-regarding' preferences to the experimental test, namely the theories developed by Charness and Rabin, by Fehr and Schmidt, and by Andreoni and Miller. In a series of experiments based on various dictator and prisoner's dilemma games, we try to uncover which of these concepts, or the classical selfish approach, is able to explain most of our experimental findings. The experiments are special with regard to two aspects: First, we investigate the consistency of individual behavior within and across different classes of games. Second, we analyze the stability of individual behavior over time by running the same experiments on the same subjects at several points in time. Our results demonstrate that in the first wave of experiments, all theories of other-regarding preferences explain a high share of individual decisions. Other-regarding preferences seem to wash out over time, however. In the final wave, it is the classical theory of selfish behavior that delivers the best explanation. Stable behavior over time is observed only for subjects, who behave strictly selfish. Most subjects behave consistently with regard to at least one of the theories within the same class of games, but are much less consistent across games.