The perils of government enforcement

It has been shown that conventions, customs or social norms do emerge spontaneously from the repetition of interactions between individuals (Sugden 1986; Josselin and Marciano 1995, 1997). Critics of the spontaneous emergence of norms argue that there is no reason to assume that individuals will engage in interactions with others if the environment is not peaceful. In other words, norms are a public good and self-interested individuals have no incentives to contribute to their creation, which implies that norms must be created by the state [see, e.g., Sened (1997) and Aviram (2004)].

Private enforcement mechanisms may be decentralized (such as shaming) or centralized (such as private judicial institutions). What these enforcement mechanisms have in common is that they are private—i.e., non-governmental. One should not necessarily equate private with decentralized (also referred to as informal) enforcement. The literature on self-governance refers to both private centralized/formal and private decentralized/informal enforcement mechanisms [(see Leeson 2009, 2013, 2014c]. Other examples of private enforcement institutions in the literature on self-governance have less clear categories [see, e.g., Leeson 2007b, 2014a]. In this paper, a private enforcement mechanism is characterized when the individuals affected by an offender’s behavior take it upon themselves to enforce a rule of conduct through the expression of disapproval. Government enforcement is characterized when punishment is top-down/centralized. However, one should not necessarily equate centralized or top-down punishment with governmental mechanisms. We thank one referee and the associate editor of the journal for having raised this important distinction.

Recent works along these lines are Leeson (2014b), Stringham (2005, 2015) and Stringham and Powel (2009).

Even more remarkable is the case of outlaw societies, such as pirates or prison gangs. Indeed, rule compliance in these groups can only be ensured privately, precisely because these groups are outside of the law. As a related point, Fleury and Marciano (2014) provide an analysis of private-governance among the undead.

Indeed, we do not envisage the replacement of a private enforcement mechanism with a government enforcement mechanism, but rather analyze the effects of adding government enforcement to complement the private enforcement mechanism.

Stringham (2015), pp. 195–200) also emphasizes the risk that government might crowd-out private governance. In our experiment, the crowding-out is the consequence of removing government enforcement mechanisms.

Romaniuc (2016) found that this way of presenting a punishment expresses cooperation norms.

It is important to keep in mind that game theory, in general, is mute about the content of agents’ utility. The experimental literature testing game theoretic predictions chose to use self-regarding preferences in formulating the standard Nash equilibrium (see Cox (2004)). It is, however, possible to include a preference for others’ income in agents’ utility functions and then apply Nash or subgame perfection to the resulting game.

Dunbar (2004) estimates that gossip constitutes 65 % of all spoken communication. On gossip, see also Giardini (2012) and Giardini and Conte (2011).

Another major difference with our study is that Nelissen and Mulder (2013) use peer-monetary punishment whereas in our experiment the punishment is centralized.

As we know from the existing literature in experimental economics, average contributions decline significantly over time [see Keser and van Winden (2000)]. Consequently, we might reasonably expect lower contribution levels in our second segment compared to the first, ceteris paribus. However, if with this design, we can show that peer disapproval creates and sustains cooperative behavior, then this is the strongest kind of support one could obtain for the existence of norm-driven behavior backed by peer control. In doing so, we identify the lower bound of the PE treatment.

The reason we can pool the data from these four sessions for segment 1 and 2 is that we always have the Baseline in the first 10 periods and the PE in periods 11–20 and the Mann-Whitney ranksum test indicates that we cannot reject the null hypothesis that the means are drawn from the same distribution.

The null hypothesis is that the means are drawn from the same distribution. We treat average contributions of each four-person group as a single independent observation.

In effect, we constantly varied the order according to which each subject’s decision was presented to the other group members, thus rendering even non-costly retaliation impossible (see the experimental instructions in the online supplementary materials—B).

Although some types of interventions can result in higher average contributions, many generally fail to maintain those high contribution rates over time (Fehr and Gaechter 2000).

In most repeated public good experiments, it is common to detect a clear difference between the first and the last periods of the game [see Keser and van Winden (2000)].

Someone who contributes less than the average contribution in his or her group reasonably may be considered as breaking a norm since the norm is generally described as the average behavior in one’s group, community or society (for a thorough analysis of various accounts of norms, see Bicchieri (2006).

Another effect from the exposure and then removal of government enforcement concerns the threshold above which subjects raise their contributions from one period to the next. As in Masclet et al. (2003), we find that without previous exposure to government enforcement, subjects raise their contributions when they receive an amount greater or equal to about 15 disapproval points. Recall that the maximum number of disapproval points that one can send is ten. Consequently, receiving ten disapproval points clearly does not count for social disapproval—it may indeed indicate that only one, out of three, peers are highly dissatisfied with the contribution of the one who receives points of disapproval. However, receiving 50  of the maximum possible points of disapproval (the maximum being 30) may suggest general agreement that one’s contribution was not appropriate. When a government-like enforcement had been in place previously, we find that subjects do not increase the amount they contribute in \(t+1\) even when they receive amounts of disapprobation up to and slightly exceeding 20 points—at is, an amount of disapproval close the maximum possible. This is shown in Fig. 4(a, b) in the online supplementary materials—A.

However, as we explained in the previous section, this increase is not significant, perhaps owing to the small sample size, although it is obvious by eyeballing the evolution in contributions in Figure 1 that contributions are higher under the combination of government and private enforcement than when private enforcement is used in isolation in periods 21–30.

Khadjavi (2015) develops a similar argument and shows experimentally that government-like punishment may deactivate pro-social emotions.