Paying to punish on a free ride: member and nonmember punishments in voluntary coalitions

Abstract

We explore public good experiments in which players have the opportunity to form cooperative coalitions. The decision to join a coalition is voluntary as is a member's decision to comply with the terms of the coalition. An informal enforcement mechanism is examined in which coalition members and nonmembers can impose costly punishments on one another. We find that when punishments are purely voluntary, coalition members and free-riding nonmembers are equally likely to punish noncompliant coalition members. Moreover, the extent to which nonmembers punish noncompliance does not change regardless of whether the members can credibly enforce compliance within their coalition.

Keywords:

• coalitions
• informal punishment
• public good experiments

JEL Classification:

• C92
• D71
• H41

Notes

¹The complete set of experimental instructions is available upon request.

²The first stage of our experiments has attributes similar to a threshold public good experiment with a money-back guarantee (Marks and Croson, 1998).

³Joining a coalition without binding commitments can be considered a form of ‘cheap talk’. See Charness and Grosskopt (2004) for a brief review of the effectiveness of cheap talk in economic experiments.

⁴This form of costly punishment is motivated by the use of trade sanctions to enforce compliance with international environmental agreements (Barrett, 1997).

⁵There are 60 group-level observations and 240 individual-level observations for the coalition and compliance stages and 720 individual-level observations in the punishment stage.

⁶The first two columns report results from linear regressions on group-level public good provision and participation. The second and third columns report on logit regressions on whether a subject punishes noncompliance and whether a subject punishes nonparticipation, respectively. All regressions include period fixed effects and the logit regressions include a subject-specific random effects specification of the error term. The dummy variable member to member (member to nonmember) captures the effect of being a member on the likelihood of punishing noncompliance (nonparticipation), relative to nonmembers.