To Trust or Not to Trust That is Not the Question

Abstract

We propose that there are three determinants of sender behavior in trust games: Beliefs re-garding the amounts returned, risk aversion, and reciprocity. Particularly, we are interested in the role of reciprocity because the possibility of negative expected reciprocal utility may lead to betrayal-averse sender behavior, i.e. to a situation where reciprocal subjects send less money than solely selfish ones. In our experiment, most subjects show distinct social preferences in the receiver role, but in the sender role they do not distinguish between a standard trust game with a human partner and a non-social, lottery-like setting, where a computer plays the role of the receiver. This means that the relevance of reciprocity considerations under uncertainty might be fundamentally different from those when no uncertainty is present. Furthermore, we find that sendings are mainly driven by overoptimistic beliefs. We use a modified version of the reciprocity model of Dufwenberg and Kirchsteiger [Games Econ. Beh. 47 (2004) 268] to explain our results and thereby show why reciprocal and selfish subjects almost do not differ in their sending behavior if one controls for beliefs and risk aversion. For other experimental settings, our model does predict differing behavior, which is in line with findings reported in the literature.