Laboratory experiments have provided evidence that many individuals are willing to punish at a personal cost those favoring their private over the public interest. This type of punishment has been dubbed “altruistic” because it benefits third parties by discouraging free-riding (Fehr & Gächter Reference Fehr and Gächter2002). Guala's article is an excellent critical review of the literature on altruistic punishment.Footnote
1
The careful discussion of the field data on altruistic punishment is particularly useful. It is important to note, however, that this data comes from small-scale societies. While altruistic punishment could, in principle, explain cooperation when individuals interact repeatedly, one should not hurry to infer by the infrequency (or even the absence) of altruistic punishment in these cases that it is not an important force in supporting cooperation in general. The reason is that the expected cost of altruistic punishment can be larger in repeated than in one-shot interactions. For example, free-riders have an incentive to counter-punish in repeated interactions to signal that punishment will not be tolerated in the future.
Counter-punishment raises the cost of altruistic punishment and can also lead to feuds (i.e., cycles of retaliation) that further increase the cost of enforcing cooperation. Altruistic punishment can also destroy social ties, which means that future benefits from interacting with a particular individual are foregone. The demand for altruistic punishment has been shown to decline when the cost of punishment increases, all else equal, in laboratory experiments (Anderson & Putterman Reference Anderson and Putterman2006; Carpenter Reference Carpenter2007; Egas & Riedl Reference Egas and Riedl2008; Nikiforakis & Normann Reference Nikiforakis and Normann2008). Therefore, the rarity of altruistic punishment in small-scale societies does not necessarily imply that altruistic punishment plays no role in one-shot interactions. In addition, recent studies have found that individuals can be quite forward looking if there exists a prospect of future interactions (Cabral et al. Reference Cabral, Ozbay and Schotter2011; Reuben & Suetens Reference Reuben and Suetens2011).
The existence of altruistic punishment can be tested more clearly in one-shot interactions. To my knowledge, the only experimental field evidence on altruistic punishment in one-shot interactions comes from a natural field experiment recently run by Balafoutas and Nikiforakis (Reference Balafoutas and Nikiforakis2011) in Athens, Greece. In this study, the authors exogenously violated two well-established, efficiency-enhancing social norms (littering and standing on the left side of escalators in a central subway station). The goal was to examine whether civilians punish norm violators. The individuals were unaware they were taking part in an experiment. The rate of altruistic punishment is overall low: Altruistic punishment is observed in only 11.7% of violations (35 cases out of 300). Surprisingly, violations of the more well-established of the two norms (littering) are less likely to be punished (4% versus 19.3%). Questionnaire data indicates that the reason for the low occurrence of altruistic punishment is that people are concerned about being counter-punished by the norm violator. Further, they consider individuals who litter more likely to counter-punish than those who stand on the left side of the escalators. Interestingly, the vast majority of people do adhere to the two norms. This raises the question whether the low frequency of altruistic punishment in our experiment can explain by itself the widespread adherence to the norm that is observed. We believe that this is unlikely to be the case. For example, some individuals will be unwilling to litter because they have internalized this norm early in their lives. However, one cannot rule out that altruistic punishment plays a significant disciplining role. It is possible that some individuals may not litter because they are concerned that they will be punished: The probability may be low, but it is not zero.
Is there any evidence that the threat of altruistic punishment can sustain cooperation even when the overall threat of punishment is very low? Nikiforakis and Engelmann (Reference Nikiforakis and Engelmann2011) study a public-good game in which altruistic punishment could lead to counter-punishment and lengthy feuds. The design imposes minimal restrictions on the punishment strategies subjects can adopt and thus allows them to use a range of complex strategies that are often found in the field (e.g., punish non-punishers, intervene to stop a feud). The most surprising finding is that the likelihood that an extreme free-rider gets punished is as low as that of a cooperator being punished. However, despite the low frequency and severity of altruistic punishment, cooperation rates are higher than those typically observed in public-good experiments without any punishment opportunities. This suggests that the mere possibility of triggering a feud may be sufficient to stop some individuals from free-riding. Therefore, the results in Nikiforakis and Engelmann (Reference Nikiforakis and Engelmann2011) suggest that the low frequency of altruistic punishment reported in the field experiment of Balafoutas and Nikiforakis (Reference Balafoutas and Nikiforakis2011) may be sufficient to support (at least to some extent) cooperation.
I am referring to the study by Nikiforakis and Engelmann (Reference Nikiforakis and Engelmann2011) to emphasize that laboratory experiments can help us understand the determinants of cooperation in the field by allowing researchers to study counterfactuals (e.g., cooperation in the absence of altruistic punishment) in ways that is difficult (if not impossible) to do in the field. Field data – both experimental and not – is essential in order to understand the forces that exist outside the laboratory, the strategies employed by individuals, and the different institutions that emerge in different circumstances. However, field data will prove insufficient in some cases to explain the determinants of cooperation by itself. The reason is that key variables such as the perceived threat posed by altruistic punishment and the risk preferences of potential norm violators are difficult to measure in the field. The difficulty of measuring important variables in the field and the limited control of researchers over the experimental environment (e.g., researchers may not be able to remove a certain strategy in a natural field experiment) is what makes laboratory experiments useful in understanding factors that facilitate cooperation such as altruistic punishment.
Laboratory experiments have provided evidence that many individuals are willing to punish at a personal cost those favoring their private over the public interest. This type of punishment has been dubbed “altruistic” because it benefits third parties by discouraging free-riding (Fehr & Gächter Reference Fehr and Gächter2002). Guala's article is an excellent critical review of the literature on altruistic punishment.Footnote 1 The careful discussion of the field data on altruistic punishment is particularly useful. It is important to note, however, that this data comes from small-scale societies. While altruistic punishment could, in principle, explain cooperation when individuals interact repeatedly, one should not hurry to infer by the infrequency (or even the absence) of altruistic punishment in these cases that it is not an important force in supporting cooperation in general. The reason is that the expected cost of altruistic punishment can be larger in repeated than in one-shot interactions. For example, free-riders have an incentive to counter-punish in repeated interactions to signal that punishment will not be tolerated in the future.
Counter-punishment raises the cost of altruistic punishment and can also lead to feuds (i.e., cycles of retaliation) that further increase the cost of enforcing cooperation. Altruistic punishment can also destroy social ties, which means that future benefits from interacting with a particular individual are foregone. The demand for altruistic punishment has been shown to decline when the cost of punishment increases, all else equal, in laboratory experiments (Anderson & Putterman Reference Anderson and Putterman2006; Carpenter Reference Carpenter2007; Egas & Riedl Reference Egas and Riedl2008; Nikiforakis & Normann Reference Nikiforakis and Normann2008). Therefore, the rarity of altruistic punishment in small-scale societies does not necessarily imply that altruistic punishment plays no role in one-shot interactions. In addition, recent studies have found that individuals can be quite forward looking if there exists a prospect of future interactions (Cabral et al. Reference Cabral, Ozbay and Schotter2011; Reuben & Suetens Reference Reuben and Suetens2011).
The existence of altruistic punishment can be tested more clearly in one-shot interactions. To my knowledge, the only experimental field evidence on altruistic punishment in one-shot interactions comes from a natural field experiment recently run by Balafoutas and Nikiforakis (Reference Balafoutas and Nikiforakis2011) in Athens, Greece. In this study, the authors exogenously violated two well-established, efficiency-enhancing social norms (littering and standing on the left side of escalators in a central subway station). The goal was to examine whether civilians punish norm violators. The individuals were unaware they were taking part in an experiment. The rate of altruistic punishment is overall low: Altruistic punishment is observed in only 11.7% of violations (35 cases out of 300). Surprisingly, violations of the more well-established of the two norms (littering) are less likely to be punished (4% versus 19.3%). Questionnaire data indicates that the reason for the low occurrence of altruistic punishment is that people are concerned about being counter-punished by the norm violator. Further, they consider individuals who litter more likely to counter-punish than those who stand on the left side of the escalators. Interestingly, the vast majority of people do adhere to the two norms. This raises the question whether the low frequency of altruistic punishment in our experiment can explain by itself the widespread adherence to the norm that is observed. We believe that this is unlikely to be the case. For example, some individuals will be unwilling to litter because they have internalized this norm early in their lives. However, one cannot rule out that altruistic punishment plays a significant disciplining role. It is possible that some individuals may not litter because they are concerned that they will be punished: The probability may be low, but it is not zero.
Is there any evidence that the threat of altruistic punishment can sustain cooperation even when the overall threat of punishment is very low? Nikiforakis and Engelmann (Reference Nikiforakis and Engelmann2011) study a public-good game in which altruistic punishment could lead to counter-punishment and lengthy feuds. The design imposes minimal restrictions on the punishment strategies subjects can adopt and thus allows them to use a range of complex strategies that are often found in the field (e.g., punish non-punishers, intervene to stop a feud). The most surprising finding is that the likelihood that an extreme free-rider gets punished is as low as that of a cooperator being punished. However, despite the low frequency and severity of altruistic punishment, cooperation rates are higher than those typically observed in public-good experiments without any punishment opportunities. This suggests that the mere possibility of triggering a feud may be sufficient to stop some individuals from free-riding. Therefore, the results in Nikiforakis and Engelmann (Reference Nikiforakis and Engelmann2011) suggest that the low frequency of altruistic punishment reported in the field experiment of Balafoutas and Nikiforakis (Reference Balafoutas and Nikiforakis2011) may be sufficient to support (at least to some extent) cooperation.
I am referring to the study by Nikiforakis and Engelmann (Reference Nikiforakis and Engelmann2011) to emphasize that laboratory experiments can help us understand the determinants of cooperation in the field by allowing researchers to study counterfactuals (e.g., cooperation in the absence of altruistic punishment) in ways that is difficult (if not impossible) to do in the field. Field data – both experimental and not – is essential in order to understand the forces that exist outside the laboratory, the strategies employed by individuals, and the different institutions that emerge in different circumstances. However, field data will prove insufficient in some cases to explain the determinants of cooperation by itself. The reason is that key variables such as the perceived threat posed by altruistic punishment and the risk preferences of potential norm violators are difficult to measure in the field. The difficulty of measuring important variables in the field and the limited control of researchers over the experimental environment (e.g., researchers may not be able to remove a certain strategy in a natural field experiment) is what makes laboratory experiments useful in understanding factors that facilitate cooperation such as altruistic punishment.