Not Just Babble: A Voluntary Contribution Experiment with Iterative Numerical Messages," working paper

When subjects can make non-binding announcements of possible contributions to a public good numerically, there is no effect on average level of contributions in a public goods experiment relative to play without announcements. But a detailed analysis of this experiment shows that pre-play announcements increased the variance of achieved cooperation among groups, leading both to more highly cooperative groups and to more thoroughly uncooperative groups than in a treatment without announcements. We also add a treatment in which subjects can select a statement of (non-binding) “promise ” to contribute a certain amount and we find that even though subjects were instructed that promise statements were not binding, the ability to issue them significantly increased both contributions and earnings in a treatment that includes costly punishment opportunities, although not in a treatment without punishment