The effect of nonlinear utility on behaviour in repeated prisoner’s dilemmas

The present study focuses on the effect of agents’ utility on their cooperation in indefinitely repeated two-person prisoner’s dilemma games (PDs). A game-theoretical analysis suggests that conditions for cooperation in the PDs improve with concavity of utility, with increasing risk aversion, and with increasing loss aversion in the case of PDs with both negative and positive outcomes. The hypotheses were tested in an experiment in which participants played a number of (different) indefinitely repeated PDs and their utility was elicited. The results provide no evidence that concave utility promotes cooperation in the PDs, but we do show that cooperation is promoted by risk aversion and possibly by loss aversion.