Experimental studies identify the possibility to punish free-riders as a valuable means to sustain cooperation in social dilemmas (Fehr and Gächter, 2000 and Ostrom et al., 1992; see also the reviews by Gächter and Herrmann (2009) and Chaudhuri, 2011). While exogenously (by the experimenter) installed punishment institutions succeed to increase contributions, they often do not produce significantly higher (overall) payoffs than the voluntary contribution mechanism (VCM) without the punishment option (cf. Egas and Riedl, 2008 and Herrmann et al., 2008).1
A recent strand of studies show that endogenous choice of punishment institutions may also induce high contributions to a public good and increase the institution’s efficiency ( Ertan et al., 2009, Gürerk et al., 2009, Gürerk et al., 2010, Sutter et al., 2010 and Tyran and Feld, 2006).2 Although punishment mechanisms in these studies are the more efficient institutions “in the long run”, their overall efficiency often suffer from two stable behavioral patterns. First, (initially) subjects show a great reluctance to interact with each other in the presence of punishment possibilities. Second, in punishment environments, in the beginning – similar to exogenously installed punishment institutions – there is an efficiency loss due to frequent punishment.
The research question of this paper is this: How can be the initial efficiency loss mitigated in order to improve the overall performance of endogenously chosen punishment institutions? This question is closely linked to the initial poor acceptance of the punishment institution. So, we may reformulate our research question: How the initial acceptance – and linked with that – the initial and hence the overall efficiency of endogenous punishment institutions could be increased?
Before proposing an answer to this question let us speculate on the reasons for the initial reluctance to choose punishment institution. First, subjects may have a “natural aversion” against punishment since they associate negative feelings with it. Social psychologists define negative sanctions as deliberate acts that lead to unpleasant inner states that the punished person wants to avoid. Second, subjects may also fear to be exposed to unjustified punishment. In fact, in experiments, punishment of high contributors is a frequently observed phenomenon (Cinyabuguma et al., 2006 and Herrmann et al., 2008). A third possible explanation is that subjects simply do not anticipate correctly that the punishment institution is the more efficient institution in the long-run. If subjects knew ex-ante that the punishment institution generates low efficiency in the beginning but yields high payoff in the future, they probably focus on the long-term benefit and join it right from the beginning. However, subjects could also focus on the downside caused by the short-term loss and hence shun away from the punishment institution.
In this study, we investigate the last proposed possible explanation, i.e., how ex-ante information about the punishment institution affects subjects’ acceptance for this institution. For this, we conduct a social history treatment in which we provide subjects with the complete history of a endogenously chosen punishment institution as occurred in a previous experiment. This social history treatment (in the following abbreviated by SHT) is the exact replication of the PUN treatment from the study of Gürerk et al. (2010) in which subjects individually choose in each period between institutions with and without punishment possibilities before interacting with others who choose the same institution in a public goods setting. The only difference between the PUN and the SHT treatments is that in SHT, a social history3 reporting the main results of PUN is given to the subjects.
The social history provides subjects with information that may help them to identify cognitively that the punishment institution is the more efficient mechanism in the long run. On the other hand, social history may lead subjects to simply imitate the most common behavior in PUN without realizing intellectually that doing so they will be better off in the end. For example, subjects may imitate by simply choosing the historically most popular institution.
To identify unambiguously whether the behavior we would possibly observe in SHT is due to social learning and not mere imitation we conduct another treatment called SH-Half which provides subjects only with a subset of social history information given in SHT. Specifically, in SH-Half, we provide subjects solely with the history of institutional choice as occurred in PUN but with no other information given in the social history of SHT. If behavior observed in SH-Half is closer to PUN and different from SHT, we may be more confident that what we observe in SHT is indeed social learning. If, however, behavior in SH-Half is more similar to SHT, then the data may be interpreted as an evidence for mere imitation.
Previous experimental studies show that social history may affect subjects’ behavior. In their influential “trust game” study, Berg, Dickhaut, and McCabe (1995) find significant effects of social history on subjects’ choices. With social history, both amounts invested by the sender and the amount sent back by the responder increase. A replication study by Ortmann, Fitzgerald, and Boeing (2000) with some additional treatments confirms the results of Berg et al. (1995). On the other hand, there is also some literature reporting no change in subjects’ behavior when they are provided with information on earlier play of another cohort (Fehr & Rockenbach, 2003). This study is a variant of the trust game in which the sender states how much she wants the responder to transfer back and may activate a punishment option (or not) for the case the responder does not meet the desired amount. The results show if senders choose to activate the punishment option, then the actual back transfers are lower and senders earn less than when senders deliberately refrain from using the fine. In a social history treatment, senders were informed about this result. Knowing that, roughly the same percentage of senders still activated the punishment option, i.e., social history did not change senders’ behavior.
Like social history, advice giving4 also unfolds its impact through social learning. Schotter (2003) reviews a series of studies on advice giving, most of them “intergenerational games”. In these studies, successors who take advice from their predecessors play the same game differently than their advice givers did. The advice takers tend to follow the advices of the advice givers. Chaudhuri, Graziano, and Pushkar (2006) report that advice – given as free-form text messages by individuals – increases contributions to a public good if it becomes common knowledge.
With one exception, the above studies show that people apparently react to information provided in social history and advices. In some settings, this information helps increase trust and positive reciprocity (Berg et al., 1995 and Ortmann et al., 2000), in others it serves as a coordination device, e.g., in a public goods setting moving the contributions to rather high levels even in the absence of any sanctions (Chaudhuri et al., 2006). Thus, altogether, we believe that the findings from the existing studies on social history and advice giving presented above support the hypothesis that social history in our experiment could have an enhancing effect on cooperation and an accelerating effect on the speed of convergence to the socially efficient state of full cooperation.
Our main results support the conjectures posed above. With social history, the initial acceptance of the punishment institution increases significantly. Contributions towards the public good are significantly higher than without social history right from the beginning and they converge earlier to socially efficient level. If we consider the society as a whole including both punishment and non-punishment institutions, the payoffs are the highest with social history. Our results underline the relevance of transferring experience based information to increase efficiency. We contribute to the existing literature by confirming the cooperation enhancing effect of social history in a more complicated experimental setting than previous studies. Furthermore, we show that the driving force behind the success of social history is social learning rather than mere imitation.
The following section discusses the related literature on endogenous institution choice. Section 3 describes the experimental design and procedure. Results are presented in Section 4. Section 5 concludes.
In this study, we explore whether informed subjects are less reluctant than uninformed subjects to join an institution with a peer punishment mechanism in a social dilemma situation and whether the society consisting of informed subjects obtain a greater efficiency. We observe a clear effect of social history on the institution choice. With social history, initially, significantly more subjects join the punishment community than in the baseline treatment. Moreover, with social history, subjects joining the punishment community start to cooperate on a much higher level than the subjects who join the punishment community in the baseline treatment. With social history, the punishment expenses are also lower. In the beginning, in both treatments, the payoff differential between the punishment and the non-punishment community is negative for the punishment institution. However, with social history, the payoffs in PuC catch up with the payoffs in NPC more quickly than they do in PUN. As a consequence, the punishment institution of SHT attracts subjects more quickly than the PuC of the baseline treatment. The socially efficient “ideal” state of “full participation” with (almost) full contributions is reached significantly earlier in PuC of SHT than in PuC of PUN. With social history, overall efficiency gains for the society as a whole are higher than without it.
Hence, with respect to our research question, the data clearly show that ex-ante knowledge about the punishment institution does not shun away the subjects from joining it. This information, provided to the subjects with the social history, does rather increase the initial acceptance of the punishment institution significantly. Moreover, social history works as a catalyst in reaching full contributions more quickly and in obtaining a higher overall efficiency.
To identify whether the driving force behind the success of social history is social learning or imitation we conducted the SH-Half treatment in which subjects were provided with less information from the baseline treatment than in SHT. For all important variables, the data we observe in SH-Half is closer to PUN than to SHT. This regularity in behavior supports the conjecture that what is going on in SHT is indeed social learning and not mere imitation. This finding is a novel contribution of our paper to the literature compared to previous experimental studies on social learning.
Although our findings unambiguously show that social history has clearly increased the initial acceptance and the overall efficiency of the punishment institution still there is potential for “improvement”. Roughly half of the subjects (45.8%) do not join the punishment community initially. Why is this? A possible reason could be the fear of anti-social punishment. Future research could clear this question by conducting an experiment that could restrict unjustified punishment in the punishment institution. Probably, more subjects choose then the punishment institution initially. Moreover, since unjustified punishment would be ruled out, the efficiency of the punishment institution may also increase.
