The subgroup problem: When can binding voting on extractions from a common pool resource overcome the tragedy of the commons?

Using a common pool resource game protocol with voting we examine experimentally how cooperation varies with the level at which (binding) votes are aggregated. Our results are broadly in line with theoretical predictions. When players can vote on the behavior of the whole group or when leaders from each group can vote for the group as a whole, extraction levels from the common resource pool are close to the social optimum. When players extract resources individually, there is substantial overextraction. When players vote in subgroups, there is initially less overextraction but it increases over time. This suggests that in order for binding voting to overcome the tragedy of the commons in social dilemmas, it should ideally affect the group as a whole.