Department of Economics

We test whether promises per se are effective in enhancing cooperative behavior in a form of trust game. In a new treatment, rather than permitting free-form messages, we instead allow only a bare promise-only message to be sent (or not). We find that bare promises are much less effective in achieving good social outcomes than free-form messages; in fact, bare promise-only messages lead to behavior that is much the same as when no messages are feasible. Our design also permits us to test the predictions of guilt aversion against the predictions of lying aversion. Our experimental results provide evidence that mainly supports the guilt-aversion predictions, but we also find some support for the presence of lying aversion.