Abstract

Previous experimental research suggests that a non-negligible fraction of subjects do not resort to free-riding in the linear public goods game but rather are willing to contribute as long as other group members do so. We extend the classic design by Fischbacher et al. (2001) by letting subjects to condition their contributions on the minimum, median, average or maximum contribution of the other group members, as well as on their full contribution profile. We find that conditional cooperation is a stable phenomenon that is not affected by the choice of a conditioning statistic. Presence of the self-serving bias, however, depends on the choice of the statistic. Subjects tend to reciprocate the average rather than the other statistics. Finally, we provide guidelines on how the total contribution level can be increased by varying the amount and type of information available to subjects.