Special Section: Experiments on Learning, Methods, and Voting

We experimentally investigate the evolution play in an infinitely repeated voluntary contribution mechanism (VCM). We find that in infinitely repeated VCM games: (i) average contributions in the first round are similar to those of finitely repeated VCM games; (ii) most groups have a non-monotonic trend of contribution with repetition; and (iii) contributions remain at the same level after an unexpected restart. The data provides strong support for heterogeneous subjects, which may explain the non-monotonic trend of average contributions. This trend is caused by one category of subjects who expect others to contribute in period t as they did in period t − 1.