Collective Punishment and Self Governance in Public Good Provision

The experiment is based on the voluntary contributions mechanism (VCM) game, which is widely used in the literature (Charness and Kuhn 2011). In this game of public good provision, the payoff to each member of a group, i, is as follows:
-e_i+θ (∑_1^N▒e_i )/N (12)
where  is her total endowment, and e_i is her contribution to the public good. Concerning the correlation between theory and empirical results, it is crucial to recognize that earnings in laboratory experiments act as a substitute for "utility" in the theoretical framework.Note that Equation (6) is consistent with our theoretical model above in which u(x)=x and v(x)=θx. It is assumed that θ >1/N, which implies that the socially optimal outcome is e_i=. However, if an individual is entirely selfish, he/she will choose e_i=0 when N is large enough.

The participants in our experiment will be randomly assigned to one of the following four treatment groups:
The basic control treatment, in which the subjects will be asked to make their free choices of donating to the public good.
The individual punishment treatment: Participants in this group will have the option to impose monetary punishments on other group members. They can also seek revenge or counter-punishment after being punished.
The collective punishment treatment (without communications): In this group, participants will vote on several aspects, including the desired level of initial donation, the specific members to be punished based on their actual donations, and the monetary punishment to be imposed on free riders. Participants who are punished will have the opportunity for revenge.
The collective punishment treatment (with communications): In addition to the features in Group 3, participants in this group will be allowed to communicate with each other.

In all experimental groups, there is a two minutes’ deliberation on the proper donation amount to the public good. This discussion aims to ensure a consistent donation norm across all groups, with differences primarily arising from the methods of punishment rather than donation norms themselves. At the end of the discussion, there will be a voting on this amount, and the average of the voting outcome will be informed to all group members. Kessler and Leider (2012) and Krupka, Leider, Jiang (2017) demonstrate the importance of informal agreements in enforcing social norms. However, we hypothesize that the credibility of group members donating based on the average varies among different groups, with the group engaging in collective punishment more likely to conform to the average.

2026-06-20

2026-12-31

In the case of collective punishment, it incorporates certain aspects of “democratic punishment” in a public good game (Ambrus and Greiner, 2019), where a majority vote from group members is required to punish another member. In other words, we will initially follow a similar procedure to that of “democratic punishment” as outlined in Ambrus and Greiner (2019) when designing the treatment for “collective punishment.” However, unlike “democratic punishment,” the extent of punishment in our experiment is entirely determined by the individual choices of each member within the majority group. Additionally, the concept of counter-punishment is not considered in the existing literature on “democratic punishment,” but we do account for the possibility of counter-punishment in Treatments 3 and 4 of our experiment. Therefore, in this study, all punishments are based on the principles of self-governance.
It is important to highlight that in “collective punishment,” the act of punishment is carried out individually. The collective aspect of “collective punishment” lies in the fact that all contributors to the donations engage in a collective discussion, through voting or discussions, to determine the specific members who are to be punished based on their actual donations. Additionally, they decide on the targeted level of monetary punishment that each punisher is expected to impose on free riders. In essence, “collective punishment” differs from “individual punishment” primarily in the presence of a collective voting and/or communication process prior to individual punitive actions. This coordination among potential punishers allows for better alignment and decision-making.
Individual punishment faces challenges in sustaining cooperation due to two reasons. Firstly, an individual punisher must personally bear the cost of punishment, leading to their designation as “altruistic punishers” in the literature (Bowles and Gintis, 2011). Secondly, an individual punisher runs the risk of retaliation. In contrast, collective punishment has the potential to address both of these weaknesses. In collective punishment, each punisher within the coalition faces a relatively low risk of retaliation, and there is an expectation of contribution to the punishment from everyone in the majority group. Consequently, we anticipate that Treatment 4 will result in the highest level of cooperation, followed by Treatment 3.
We plan to conduct our laboratory experiments by enlisting university students, a widely adopted approach in academic research. The participants for our experimental study will be recruited through university bulletin board systems and online posters. To ensure appropriate subject pool recruitment procedures, we will closely adhere to the guidelines outlined by Greiner (2015). All experiment sessions will be computerized using the z-Tree software package (Fischbacher, 2007).
In our experiment, each participant engages in 20 periods. At the start of each period, participants receive a fixed amount of experimental currency units (ECUs), which they can allocate to either private or public goods. Each period consists of multiple stages. The initial stage is the contribution stage where participants decide how much to contribute to the public good. Participants’ contributions to the public good are revealed to others after the donation stage. Subsequently, there are punishment stages where participants can administer punishments and counter-punishments.
All individual punishments are implemented in the form of reducing other players’ earnings while simultaneously reducing the punisher’s own earnings. In the case of collective punishment treatments, participants first observe the donations made by all players. A voting or discussion process then takes place to determine the “free riders” who will be subject to punishment. Subsequently, cooperators engage in discussions to establish the appropriate level of punishment before each member of the cooperative majority group carries out the punishments. If no player assigns a punishment during a given stage, the period concludes, and a new one begins.
Group size often plays a crucial role in interpersonal interactions (Lim, Matros, and Turocy, 2014), and it can significantly impact the effectiveness of collective punishment for various reasons. On one hand, in larger groups, the individual who contributes the least to the public good may feel more isolated and could potentially face more substantial collective punishment. Consequently, cooperation may increase within larger groups when the mechanism of collective punishment is implemented. On the other hand, as the group size expands, individuals may feel less connected to the community (Ellickson, 2001). Hence, the overall impact of increasing group size is theoretically ambiguous but could have notable empirical significance. To examine this effect, we will introduce variations in group size during our experiments. Specifically, we will have two group sizes: 4 and 8 participants, respectively.

The participants in our experiment will be randomly assigned to one of the following four treatment groups:
1. The basic control treatment, in which the subjects will be asked to make their free choices of donating to the public good.
2. The individual punishment treatment: Participants in this group will have the option to impose monetary punishments on other group members. They can also seek revenge or counter-punishment after being punished.
3. The collective punishment treatment (without communications): In this group, participants will vote on several aspects, including the desired level of initial donation, the specific members to be punished based on their actual donations, and the monetary punishment to be imposed on free riders. Participants who are punished will have the opportunity for revenge.
4. The collective punishment treatment (with communications): In addition to the features in Group 3, participants in this group will be allowed to communicate with each other.

Three critical points should be highlighted. Firstly, in all experimental groups, there is a two minutes’ deliberation on the proper donation amount to the public good. This discussion aims to ensure a consistent donation norm across all groups, with differences primarily arising from the methods of punishment rather than donation norms themselves. At the end of the discussion, there will be a voting on this amount, and the average of the voting outcome will be informed to all group members. Kessler and Leider (2012) and Krupka, Leider, Jiang (2017) demonstrate the importance of informal agreements in enforcing social norms. However, we hypothesize that the credibility of group members donating based on the average varies among different groups, with the group engaging in collective punishment more likely to conform to the average.
Secondly, for the sake of experiment simplification, in our benchmark experiments described above we solely implement monetary penalties for free-riders in donations Nevertheless, we will integrate both non-monetary and monetary punishments in the extension of experiments later on.
Thirdly, the introduction of Treatment 4 underscores the significance of communication in collaborative settings, drawing inspiration from case studies in Ostrom (1990) and various experimental research endeavors (Levy, Padgitt, Peart, Houser, and Xiao 2011, Feltovich and Grossman 2015, Bartling, Valero, Weber, and Yao, forthcoming). As an illustration, Bartling, Valero, Weber, and Yao (forthcoming) conducted an experiment that featured an 8-minute interval allowing participants to communicate via an electronic chat window. They guided discussions on public discourse by encouraging participants to deliberate on the social acceptability or unacceptability of trading a product that negatively impacts third parties. Furthermore, participants were prompted to examine how trading this product aligns with or deviates from widely held norms regarding appropriate, ethical, and moral conduct. Participants were encouraged to freely express their opinions while refraining from personally identifying, obscene, or disparaging comments. The study revealed that such public discourse notably fosters socially responsible market behaviors. In our study, Treatments 3 and 4 diverge in terms of communication: Treatment 4 permits communication, while Treatment 3 does not. Similar to Bartling, Valero, Weber, and Yao (forthcoming), in Treatment 4 we will allow group members to communicate via an electronic chat window for 5 minutes before they take the possible actions against free riders. We hypothesize that Treatment 4 represents a more advanced approach to collective punishment compared to Treatment 3. Communication in Treatment 4 will occur prior to the implementation of punishments.

Not available

randomization done in office by a computer

individual

Yes

600 university students

600 university students

2000