What makes a sanction “stick”? The effects of financial and social sanctions on norm compliance

Abstract

The present research shows that, like financial sanctions, social punishment (the mere expression of disapproval with another person's conduct) induces compliance with norms for cooperation in a social dilemma. More importantly, after removing the sanctioning opportunity levels of cooperation decrease more under former financial than under former social sanctioning systems. Hence social sanctions are more effective than financial sanctions at inducing “sticky” norms that guide behavior even in the absence of punishment cues. Public policy implications are discussed.

Keywords:

• Social norms
• Social dilemmas
• Financial sanctions
• Social sanctions
• Cooperation

From the 10 commandments to contemporary constitutions, and from global policies like the Kyoto agreements to domestic measures like the use of energy-saving light bulbs, one can see instances of social norms pervading human societies. Researchers in social psychology (Cialdini, Reno, & Kallgren, 1990), behavioral economics (Fehr & Fischbacher, 2004), and anthropology (Henrich et al., 2006) have argued that social norms are paramount to a profound understanding of human behavior. Indeed, several major behavioral models in psychology include norm-related constructs (e.g., Bandura, 1977; Fishbein & Ajzen, 1975; Triandis, 1977). Still, all legislators face the problem of how to effectively induce norm compliance. The present research is concerned with the question of how to design a sanctioning system that makes people comply with social norms even when their actions cannot be perfectly monitored at all times and sanctions for norm violations therefore cannot be perfectly executed. This appeals to major societal issues, like insurance fraud, tax evasion, and violations of industrial quota for toxic emissions.

A sanctioning system (e.g., fining undesirable conduct) is generally considered the most straightforward tool to induce norm compliance (e.g., tax payment) in the face of opposing incentives (McAdams, 2000; Williams & Hawkins, 1986). Focusing mainly on norms for participation in collective action, numerous studies show that sanctions indeed deter free riding in social dilemmas, resulting in higher levels of cooperation (e.g., Balliet, Mulder, & Van Lange, 2011; Fehr & Gächter, 2002; Yamagishi, 1986). Insofar as sanctions change a decision's payoff, this is somewhat obvious. In principle when selfish behavior becomes more costly, cooperation becomes more likely (but see next paragraph). The more interesting question is whether a sanctioning system generates a norm that people comply with even in the (temporary) absence of punishment opportunities. We will show that this depends on the type of sanctioning system (financial or social) used to punish norm violations.

Voluntary norm compliance is an important issue because sanctioning systems are rarely perfect. Frequently, norm violations will go unnoticed. If people only comply with norms when otherwise facing punishment, a sanctioning system is only as effective as its execution. In spite of their norm-enforcing ability, several studies suggest that sanctions actually undermine voluntary norm compliance. Imposing a sanction may cause people to frame a previously ethical decision (Is it acceptable to do this?) in business terms (Is it cost-efficient to do this?), which may even result in less compliance (Gneezy & Rustichini, 2000; Tenbrunsel & Messick, 1999). Moreover, sanctions may negatively affect trust. Under a sanctioning system people ascribe others’ norm compliance to the prospect of punishment upon violation, rather than to a desire to cooperate. Consequently, when the sanctioning system is removed, levels of cooperation tend to drop (e.g., Chen, Pillutla, & Yao, 2009; Masclet, Noussair, Tucker, & Villeval, 2003; Mulder, van Dijk, De Cremer, & Wilke, 2006).

These “dark sides” of sanctions (i.e., the adoption of an economic decision frame driven by external incentives rather than mutual trust) will impair norm compliance under conditions of imperfect execution; that is, when norm violations are not consistently punished. In the present study, however, we explore the possibility that these drawbacks are an artifact of sanctions usually being modeled as financial punishment. We investigated whether the drop in voluntary norm compliance that is commonly observed after the removal of a (financial) sanctioning system is attenuated in case a social instead of a financial sanctioning system is implemented and subsequently removed. In a social sanctioning system the only form of punishment consists of the mere expression of disapproval with a particular kind of conduct. In the present study the removal of a social sanctioning system is achieved by terminating the possibility for participants to express their disapproval of each other's behavior after each round of contributions in a social dilemma game.

In everyday interactions social rather than financial punishment is the default. Expressing disproval or contempt, gossip, peer pressure, and teasing are informal, non-institutionalized means to punish norm violations. Indeed, social disapproval of non-cooperation (e.g., Masclet et al., 2003; Noussair & Tucker, 2005; Rege & Telle, 2001), but also social approval of cooperation (Chen et al., 2009) boosts cooperation in social dilemmas just like financial punishment does. Ultimately social sanctions hint at the possibility of social exclusion, which probably accounts for their impact (Tyler, 1990). Studies demonstrate that experiencing ostracism has severe consequences (for an overview see Williams, Forgas, & von Hippel, 2005), and shares the same neural substrate as physical pain (Eisenberger, Lieberman, & Williams, 2003). Some even argue that the threat of ostracism is the key mechanism underlying the development of social norms (Ouwerkerk, Kerr, Gallucci, & Van Lange, 2005).

Of particular importance to the present study, social sanctions seem less likely to suffer the previously listed limitations accompanying financial sanctions. First, as no money is involved, social sanctions are unlikely to evoke economic framing of the decision at hand. Second, in the absence of financial incentives for cooperation, individual contributions in a social dilemma are more likely to be attributed to stable, individual preferences for cooperation (e.g., Chen et al., 2009; Mulder et al., 2006), especially if people have previously been able to express their disapproval with non-cooperators as was the case in the present study. There is no reason to assume that these preferences will change if people no longer have an opportunity to explicitly communicate them to each other. Consequently we expected social sanctions to be more resilient than financial sanctions to imperfect execution because they make people more willing to comply with norms even in the absence of direct punishment cues.

Suggestive support for the greater resilience to removal of social compared to financial incentives was provided by Chen and colleagues (2009, Study 2). Specifically their study demonstrated that the removal of financial rewards (for cooperation) and sanctions (for non-cooperation) resulted in a collapse of individual contributions in a social dilemma setting. Moreover they found that moral appeals to individual cooperation increased contributions just like financial incentives did, but that individual contribution levels did not collapse after “removing” the moral appeal.

Even though the results by Chen and colleagues (2009) clearly established that removing financial sanctions reduced subsequent cooperation, they remain inconclusive about the consequences of removing social sanctions. A moral appeal in the Chen et al. study consisted of an explicit statement, issued by the project supervisor of a company, concerning the desirability and benefits of individual efforts in a team project. In our terminology, then, a moral appeal could be reframed as a social reward (rather than a sanction) for cooperation, as it involved the positive evaluation of individual contributions in a social dilemma. Admittedly this distinction would amount to nothing more than conceptual quibble, if we did not believe that removing the social reward as operationalized in the Chen et al. study was practically impossible. Because Chen and colleagues framed their moral appeal as a statement about the general desirability of individual contributions to the group project, their social reward could not actually be withdrawn without implying that individual efforts to the team project were from that moment on no longer considered desirable.1 Because, in our study, social incentives are operationalized in terms of expressions of disapproval between individual players in a social dilemma, rather than a general, moral appeal for cooperation that is issued by an external authority, we are able to examine the consequences of the removal of this social incentive simply by not allowing further expressions of disapproval from a certain point onward.

In order to investigate the effectiveness of financial and social sanctioning systems on voluntary norm compliance, we compared (to a control group) the effect of imposing either a financial or a social sanctioning system on (1) the level of cooperation in a public good game and, more importantly, (2) on the endurance of that effect after removing the sanctioning opportunity. We predicted that after removal of the sanctioning system, levels of cooperation would remain higher in the social than in the financial sanctioning condition.

To be clear, we did not expect that levels of cooperation would remain high indefinitely after the removal of a social sanctioning system. Sure enough, when people do not disapprove each other's non-cooperation any more, the norm will ultimately evaporate. Still, such extended non-punishment seems unlikely, as sooner or later non-compliance will be noted again and met with disapproval. Removing the sanctioning opportunity was our experimental proxy for the inherent lapses in the execution of a sanctioning system that follow from the inability to always monitor other people's social behavior. Our prediction strictly concerns the fact that compared to a financial sanctioning system, a social sanctioning system is less vulnerable to (short) lapses in its execution and will more readily uphold levels of voluntary norm compliance.

Method

Participants and design

A total of 84 undergraduate students (43 male, 41 female, M _age = 20.57 years) participated in this study. Participants were randomly assigned to the three conditions (i.e., no sanction, financial sanction, and social sanction).

Procedure

Participants were seated in individual cubicles and received instructions on the computer screen. Consistent with the actual experimental procedure2 they learned that they were to participate in a four-player interaction, that they were randomly paired with three other participants, and that all group members would remain anonymous throughout and after the study. In the interaction they could earn money dependent on their own and the other players’ decisions. We told participants that the interaction would consist of several rounds and that they would not know the number of rounds beforehand. At the beginning of each round every player was endowed with 4 Euro (about 5.60 US dollars), of which they could invest some amount in a group fund at their own discretion. After all group-members had decided on their contributions, the total amount in the group fund was multiplied by 1.5 and distributed evenly. Subsequently individual contributions were displayed on screen in such a way that they could not be assigned consistently to any individual player. Hereto we randomly alternated the order in which players’ contributions were listed between rounds. This excluded retribution effects on contribution and punishment decisions (e.g., Houser, Xiao, McCabe, & Smith, 2008).

In the sanction conditions participants were then asked whether they wanted to assign a 1 Euro fine (in the financial sanction condition) or to express disapproval (in the social sanction condition) to each of the three other group members. Participants answered this question for each individual player by clicking on a “yes” or “no” button that appeared at their computer screens. Imposing a fine did not reduce the punishing players’ payoff in order to exclude cost effects on the willingness to punish. After all players made their punishment decisions players receiving sanctions were informed of either the number of players that had imposed a fine or that had expressed disapproval with their contribution. The fine(s) (max. €3) were deducted from the target player's payoff for that particular round.

In both sanctioning conditions the seventh round was an extra round in which participants received an error message that told them the program had crashed and they were to call the experimenter. The experimenter jump-started the program, but informed the participants that apparently, due to a malfunction, players could no longer execute and/or receive sanctions from then on. Instead, they would merely learn about individual contributions at the end of each round, after which the next round of play would start immediately. Consequently the final three rounds were identical in all conditions.

Measures

The game started with a practice round to familiarize participants with the procedures. Then the real game was played. Participants were told that they would receive their payoff from one randomly selected round. Each round we measured player's contributions to the group fund. In the sanctioning conditions we also measured the frequency with which group members were punished in each round.

After the final round, as a manipulation check for the type of sanctioning system, we asked to what extent participants felt “they could be punished financially by others” and “their behavior could be disapproved by others” on a scale ranging from 1 (not at all) to 7 (very much).

Results

Manipulation checks

Establishing that our sanctioning manipulations were successful, we found significant differences in the extent people felt they could be punished financially F(2, 81) = 52.38, p < .001, η²= .564. Post-hoc comparisons (LSD) revealed significantly higher ratings in the financial (M = 6.07, SD = 0.94) than in the social (M = 2.50, SD = 1.55) or in the no sanction condition (M = 2.75, SD = 1.76), ps < .001. There was no difference between the social and the no sanction condition, p = .523. Similarly, conditions differed on scores of potential for disapproval by others F(2, 81) = 13.29, p < .001, η²= .247. Here post-hoc comparisons (LSD) indicated higher ratings in the social (M = 4.54, SD = 1.88) than in the financial (M = 2.86, SD = 1.86) and the no sanction condition (M = 2.21, SD = 1.45), ps ≤ .001. There was no difference between the financial and the no sanction condition (p = .171).

Dependent measures

We found that lower contributions to the group fund were sanctioned more frequently, r(56) = −.85, p < .001. So sanctions were indeed applied as a response to norm violations. Moreover, sanctions also induced norm compliance. A one-way ANOVA, comparing differences in average contributions during the first six rounds (i.e., before sanctions were removed, Phase 1), showed that sanctions boosted cooperation F(2, 81) = 44.12, p < .001, η²= .521. Post-hoc comparisons (LSD) revealed that contributions were higher under financial (M = 3.18, SD = .74) and social (M = 2.55, SD = 1.17) sanctioning than in the control condition (M = 0.87, SD = .90), ps < .001. Contributions were also higher in the financial than in the social sanctioning condition, p = .016 (see Figure 1).

Figure 1. Average contributions per round in control (no sanction), social, and financial sanction conditions. Sanctioning opportunities are removed in Round 7.

We subsequently computed average contributions during the final three rounds (i.e., after the sanctions were removed, Phase 2), and conducted a Condition × Phase ANOVA with Phase as a within-participant factor of average contributions before and after removing the sanctioning opportunity. As anticipated, this yielded a main effect of Phase, F(1, 81) = 44.09, p < .001, η²= .352, showing higher average contributions in Phase 1 (M = 2.20, SD = 1.36) than in Phase 2 (M = 1.29, SD = 1.23). This confirms the general effect of sanctioning on contribution levels. There was a main effect for condition, F(2, 81) = 36.19, p < .001, η²= .472. More important, and in line with our predictions, a significant Condition × Phase interaction, F(2, 81) = 9.09, p < .001, η²= .183, showed that the difference between average contributions in Phase 1 and 2 was mainly driven by a reduction of contributions after removal of the financial (but not the social) sanctioning opportunity (see Figure 1).

An analysis of simple effects showed that the effect for phase was larger in the financial sanction, F(1, 81) = 52.53, p < .001, η²= .393, than in the social sanction condition, F(1, 81) = 7.37, p = .008, η²= .083. There was no significant difference between contributions in Phase 1 and 2 in the no sanction control condition, F(1, 81) = 2.36, p = .128, η²= .028. This indicates that the drop in cooperation was significantly larger for the financial than for social sanction condition.

To formally test whether the drop in the financial sanctioning condition was larger than in the social sanctioning condition we compared difference scores (Phase 1 – Phase 2) of average contribution levels between conditions. A main effect of condition replicated the Condition × Phase interaction effect noted above, F(2, 81) = 9.09, p < .001, η²= .183.33 Post-hoc comparisons (LSD) of mean differences revealed higher difference scores for people in the financial sanctioning condition (M = 1.73, SD = 1.55) than for people in the social (M = 0.65, SD = 1.26, p = .002) and no sanction condition (M = 0.37, SD = .87, p < .001), whereas the latter conditions did not differ, p = .407. This supports our prediction that norms are more resilient to imperfect execution of social than of financial sanctions

Further corroborating the greater resilience to removal of social compared to financial sanctions, a one-way ANOVA on final contributions in Round 10 showed a significant overall effect, F(2, 81) = 4.16, p = .019, η²= .093. Post-hoc comparisons (LSD) revealed that contributions in the social sanctioning condition (M = 1.56, SD = 1.61) were ultimately even higher than in the financial (M = 0.79, SD = 1.50, p = .040) and no sanction (M = 0.54, SD = .96, p = .007) conditions, whereas the latter did not differ any more, p = .502.

Notably, the average frequency of punishment in Phase 1 did not differ between financial (M = 6.36, SD = 3.75) and social sanctioning (M = 7.61, SD = 5.19) conditions, t(54) = –1.03, p = .306. Therefore the observed difference in resilience to removal of financial and social sanctioning opportunities is not the result of the amount but of the nature of previous sanctioning.

Discussion

Our results show that expressing disapproval is an efficient means to induce norm compliance in collective action. Just like financial sanctions, social punishment boosted contributions in a public good game. More importantly, average contribution levels did not decline as profoundly if a social sanctioning opportunity was removed than when a financial sanctioning opportunity was removed. Thus social sanctions were more likely to induce voluntary norm compliance in the (temporary) absence of punishment than financial sanctions.

This is the first study to show that a sanctioning system may not only induce compliance with a norm in spite of opposing incentives, but also that this effect is less dependent on the consistent punishment of norm violations. Although previous studies established a norm-inducing effect of social sanctions (e.g., Masclet et al., 2003) and rewards (Chen et al., 2009), they did not demonstrate the resilience of that effect as we did in the present study. As explained in the introduction, in the Chen et al. study removing the social rewards was practically impossible due to their operationalization as general moral appeals. In the Masclet et al. study contributions completely collapsed after lifting social sanctioning opportunities. This discrepancy can be attributed to two important features of our study. First, we attributed the further absence of sanctioning opportunities to an unintended program failure. Just like imposing a sanction signals social undesirability of behavior, lifting it without reason may signal renewed tolerance. Due to the cover story, participants in our study had no reason to assume that lower contributions were acceptable simply because they could no longer be punished. Our set-up more closely reflects actual instances of punishment failure that do not imply the legitimacy of norm violations either. Second, in our study the baseline norm was set with a sanctioning regime already in effect. If sanctioning opportunities are implemented post-hoc the baseline is set in the absence of punishment, which likely entails a lower contribution level. Collapse to this level after lifting the sanctioning opportunity may then be inevitable.

We conclude that, compared to financial sanctions, social sanctions are more likely to elicit voluntary norm compliance even if people's behavior cannot be consistently monitored and their norm violations therefore may go unpunished from time to time as is often the case in real life. In other words, social sanctioning systems are more lenient than financial sanctioning systems to inevitable flaws in their execution. As already stated, we do not claim that people will voluntarily comply with norms indefinitely if violations remain unpunished. When people do not disapprove each other's non-cooperation any more, the norm will ultimately vanish. Although exact statistics on the proportion of an individual's norm violations that go unnoticed are lacking, extended non-punishment seems unlikely. The observed resilience of a social sanctioning system should thus be sufficient to buffer incidental slips of vigilance. Clearly this has important implications for public policy. Our results suggest that successful norm induction requires public communication of social (dis)approval, not only because it increases the salience and thus the effectiveness of norms in guiding behavior (cf. Cialdini et al., 1990; Xiao & Houser, 2006), but also because it makes them stick even if people are not consistently punished for their violations.

Notes

Notes

1. It should be noted that the removal of the financial incentives in the Chen et al., (2009) studies was introduced by an explicit statement that bonuses and fines would no longer be extended. It was not detailed, however, if and how participants were informed that the moral appeal to cooperative behavior was withdrawn. It seems that the moral appeal simply was not reinstated. This is not the same as the actual removal of the social incentive.

2. Apart from the justification of the removal of the sanctioning opportunities, no other forms of “deceit” were employed in this study. Participants were correctly and completely informed of the design and payoff structure of the experiment and were not subjected to pre-programmed “decisions” by fictitious others.

3. Note that the value of this F-test is identical to that of the Condition × Phase interaction for differences in average contributions reported before.