Cooperation, punishment, and group change in multilevel public goods experiments

Peer punishment is regarded as an important element in sustaining human cooperation for public good provision. Many behavioral experiments have shown that public good provision is higher if cooperation norms can be enforced by peer punishment. However, these experiments predominantly focus on single-group public goods, in which people have to choose between their private interests and the interests of their group. In many societal problems, people are involved in multilevel public goods problems, where multiple local groups are nested within a larger global group. We study experimentally how punishment affects cooperation and norms in multilevel public goods games. In our games, two local groups are nested within a larger global group. Participants have to choose between not contributing, contributing locally, and contributing globally. Local contributions would lead to a polarized outcome where two separate local public goods are provided, whereas global contributions would lead to a unified global good that benefits all. Moreover, we study whether cooperation and punishment patterns depend on the type of public good participants are initially exposed to: single-group or multilevel. Participants either begin in a single-group public goods game and then shift to a multilevel public goods game or vice versa. We find that punishment is less effective in multilevel public goods games than in single-group public goods games. In particular, punishment only promotes cooperation in multilevel public goods games if people have prior experience with solving single-group public goods games. Our results refine the boundary conditions for the effectiveness of punishment and suggest that ‘starting small’ by first solving single-group public goods problems is necessary for successful multilevel public good provision.