Broken Promises : An Experiment

We test whether promises per se are effective in enhancing cooperative behavior in a form of a trust game. In Charness & Dufwenberg (2006) we found considerable effectiveness for free-form personalized pre-play statements-of-intent ("promises"), in support of a theory of belief-dependent guilt aversion. However, we were not able to reject an alternative explanation based on a belief-independent cost-of-lying. We now adapt our old design and replace the free-form messages with an opportunity for a bare promise-only message. If both forms of promises are equally effective, this would be consistent with a cost-of-lying explanation. However, in sharp contrast to previous results, we find that these bare promise-only messages lead to behavior that is much the same as when no messages are feasible. Further, beliefs are unaffected, in contrast with the change in beliefs we found with personalized promises. This provides evidence for belief-dependent guilt aversion over cost-of-lying