Explaining Antisocial Punishment

1 Punishing group members who parasitize their own group’s 2 resources is an almost universal human behavior, as evidenced by 3 multiple cross-cultural and theoretical studies. Recently, researchers in 4 social and behavioral sciences have identified a puzzling phenomenon 5 called “antisocial punishment”: some people are willing to pay a cost to 6 “punish” those who act in ways that benefit their shared social group. 7 Interestingly, the expression of antisocial punishment behavior is 8 regionally diverse and linked to the socio-psychological dimensions of 9 local cultural values. In this review, we adopt an ecological perspective 10 to examine why antisocial punishment might be an advantageous strategy 11 for individuals in some socio-economic contexts. Drawing from research 12 in behavioral economics, personality, social psychology and 13 anthropology, we discuss the proximate mechanisms of antisocial 14 punishment operating at an individual level, and their consequences at 15 the group and cultural levels. We also consider the evolutionary 16 dynamics of antisocial punishment investigated with computer 17 simulations. We argue that antisocial punishment is an expression of 18 aggression, and is driven by competition for status. Our review elucidates 19 the possible socio-ecological underpinnings of antisocial punishment, 20 which may have widespread repercussions at a cultural level. 21 EXPLAINING ANTISOCIAL PUNISHMENT 3 Homo homini lupus? Explaining antisocial punishment 22 “It is not surprising that there should be a struggle in man 23 between his social instincts, with their derived virtues, and his lower, 24 though at the moment, stronger impulses or desires.” (Darwin, 1871, 25 p.104) 26 Recent reports on antisocial punishment have drawn attention to 27 the duality of human nature. Antisocial punishment can be defined as 28 paying a cost to reduce the resources of a person whose previous 29 cooperative behavior benefited the punisher and their group. In past 30 research, the focus tended to be on altruistic punishment – paying a cost 31 to reduce the resources of a person who previously exploited group 32 resources. Altruistic punishment has become an area of particular interest 33 because it offers a potential resolution of the quest to understand human 34 cooperation. Extensive cooperation in humans, often considered 35 surprising in light of Darwinian natural selection theory, has been 36 investigated in numerous empirical and theoretical studies (e.g. Gintis, 37 Bowles, Boyd, & Fehr, 2005; Henrich et al., 2004). Altruistic 38 punishment, despite its negative proximate motives and, sometimes, 39 detrimental effect on average payoffs, has been proposed as a form of 40 1 Rather than turning the other cheek and continuing to cooperate, motivated by anger humans use punishment towards selfish individuals (Fehr & Gächter, 2002). 2 (Dreber, Rand, Fudenberg, & Nowak, 2008; Wu et al., 2009) EXPLAINING ANTISOCIAL PUNISHMENT 4 pro-social behavior promoting cooperation (Fehr & Gächter, 2002). 41 Moreover, it inspired a new theory of the evolution of human cooperation 42 strong reciprocity (Fehr & Fischbacher, 2003; Gintis, 2000). However, 43 more recent investigations of the full range of available and expressed 44 punishment behavior across cultures have highlighted the existence of 45 antisocial punishment. This has led some to reconsider the “dark side” of 46 human behavior, including a tendency for spite and hyper47 competitiveness (Abbink & Herrmann, 2011; Abbink & Sadrieh, 2009; 48 Herrmann, Thöni, & Gächter, 2008; Jensen, 2010). 49 Our review is motivated by the unexplained cultural variation in 50 antisocial punishment revealed by Herrmann et al. (2008). We propose 51 that the high levels of punishment directed toward cooperators in places 52 like Muscat, Athens and Riyadh reflect different pressures in these socio53 economic or cultural environments. These pressures affect the perception 54 of group identity, which leads to changes in individual behavior. We 55 argue that, despite lowering absolute levels of resources across a society 56 taken in aggregate, antisocial punishment may constitute a successful 57 individual strategy for establishing social status and receiving its 58 benefits. This ecological interpretation of costly punishment allows us to 59 present it devoid of ethical loading and enables a better understanding of 60 EXPLAINING ANTISOCIAL PUNISHMENT 5 its functional causes. In addition to proposing and justifying this 61 theoretical framework, we also emphasize some unresolved questions 62 about costly punishment, and offer testable predictions. 63 The review is organized as follows. We first focus on the various 64 definitions of costly punishment and how they relate to the concept of 65 altruism in different disciplines. Next, we discuss how methodological 66 manipulations of the cost-to-impact ratios of costly punishment affect its 67 use. We observe that the amount of costly punishment meted out to 68 others (in particular, antisocial punishment) is rationally adjusted to 69 exploit its effect of increasing the positive difference between one’s own 70 and others’ payoffs. In the proceeding sections we discuss antisocial 71 punishment at three levels: cultural, group and individual. At each level, 72 we show how antisocial punishment could bring advantages despite its 73 initial cost. Crucially, the benefits from using antisocial punishment may 74 result from punishers acquiring a higher status within their groups. In the 75 last section, we present the evolutionary perspective on antisocial 76 punishment and its ultimate consequences for a population, as well as, for 77 individuals. 78 3 Reproductive timing in human females viewed from an ecological perspective is a notable example of how socially undesirable behaviour, such as teenage pregnancies, can be neutrally explained and considered a biologically sensible strategy. EXPLAINING ANTISOCIAL PUNISHMENT 6 In this review, to fully understand antisocial punishment, we 79 consider both its proximate and ultimate causes (Scott-Phillips, Dickins, 80 & West, 2011). A proximate explanation refers to the mechanism that 81 leads an individual to express a behavior, while an ultimate one describes 82 the evolutionary context that resulted in the appearance of (normally, 83 selection for) a behavior or trait. While many authors have shown that 84 this distinction can be difficult to make (Scott-Phillips et al., 2011; 85 Thierry, 2005), drawing it allows us to investigate two complimentary 86 explanations for why antisocial punishment occurs. First, we focus on the 87 workings of antisocial punishment – the proximate mechanisms that 88 drive it; then, we discuss why it might have evolved – the evolutionary 89 dynamics might have caused it. The answer to the former question is 90 provided primarily by experiments using behavioral economics games 91 while the answer to the latter one comes from computer simulations of 92 evolutionary processes. 93 Costly punishment terminology 94 Economists, psychologists and biologists often use the same 95 phrases to mean different things. When drawing together knowledge 96 from various disciplines, it is important to precisely determine what is 97 understood by terms such as altruistic or antisocial punishment in each, 98 and to define the specific usage in the present discussion. Our use of the 99 word punishment originates within the context of behavioral economics 100 EXPLAINING ANTISOCIAL PUNISHMENT 7 experiments, in which researchers typically employ the Public Goods 101 Game (PGG) with punishment, the Ultimatum Game (UG) and/or the 102 Third Party Punishment game (TPP). PGGs can be played one-shot or for 103 multiple rounds (for the implications which follow from this difference, 104 see Hertwig & Ortmann, 2001). They can also be played with or without 105 punishment opportunities. If a sequence of PGGs is played, the player’s 106 group membership can be maintained or different participants may be 107 grouped together in each round. In the latter case, any consequences of 108 punishment do not affect the punisher. UGs and TPPs are, typically, only 109 played for a single round. 110 PGGs represent a social dilemma because the individual’s 111 interests are in conflict with the group’s interests. In PGGs, a group of 112 individuals can contribute some portion of their allocation to the public 113 pool, which benefits everyone equally. Individuals who do not contribute 114 anything, or contribute less than others, gain a payoff advantage. In 115 PGGs with punishment, after a round of the PGG, individuals can 116 anonymously punish others (usually at a cost-to-impact ratio of 1/3). In 117 UGs, one individual (the proposer) can share an amount of money 118 between themselves and a recipient. After the proposer’s offer, the 119 recipient decides whether they accept it, in which case both parties 120 receive the respective amounts. Alternatively, the recipient can reject the 121 offer, in which case no one receives anything. The act of rejection 122 represents the act of costly punishment because both the recipient and the 123 EXPLAINING ANTISOCIAL PUNISHMENT 8 proposer suffer a cost. TPPs greatly resemble UGs, with the major 124 difference being that the recipient in the TPP is passive and cannot 125 punish. Instead, an extra third person, not benefitting from the split, has 126 an opportunity to spend money on punishing the proposer. 127 In an experimental setting, people mete out costly punishment 128 towards uncooperative individuals, even when there is no opportunity to 129 interact with them again (Fehr & Gächter, 2002). Such punishment has 130 been dubbed “altruistic” because the punisher decides to pay a fee to 131 reduce the payoff of free-riders, and this action is likely to make free132 riders increase their cooperative contributions in future interactions. 133 Hence, in congruence with the biological definition of altruism (West, 134 Griffin, & Gardner, 2007), punishment is costly to the actor and 135 beneficial to the recipient, where the recipients are individuals interacting 136 with the punished person in the future. The biological definition of 137 altruism refers to the lifetime fitness consequences of a behavior, which 138 4 Fehr and Gächter’s definition of altruistic punishment is encapsulated in the following two quotes: “Punishment may well benefit the future group members of a punished subject, if that subject responds to the punishment by raising investments in the following periods. In this sense, punishment is altruistic.” (p.137, Fehr & Gächter, 2002). “Thus, the act of punishment, although costly for the punisher, provides a benefit to other members of the population by inducing potential non-cooperators to increase their investments. For this reason, the act of punishment is an altruistic act.” (p.139, Fehr & Gächter, 2002). EXPLAINING ANTISOCIAL PUNISHMENT 9 are impossible to capture in behavioral economics experiments. For the 139 sake of convenience, we adopt Fehr and Gächter’s term “altruistic 140 punishment” to describe a phenomenon occurring in short-term 141 experimental interactions, although we acknowledge that this definition 142 might be misleading (see Sylwester, Mitchell, & Bryson, submitted). 143 Altruistic punishment requires that (a) punishers suffer a cost for 144 punishing and (b) punished individuals are thereby induced to become 145 more pro-social. Hence, in behavioral economics, the term “altruistic 146 punishment” is defined through the negative economic outcomes to the 147 punisher and positive economic outcomes to the group. When 148 psychological drives are considered, altruistic punishment seems to be 149 motivated not by the altruistic desire to help the group but rather by 150 negative feelings towards cheaters and the willingness to harm them 151 (Fehr & Gächter, 2002). It could be argued that these negative emotions 152 are a consequence of egalitarian preferences and that the underlying 153 psychological motivation is, therefore, altruistic (Cinyabuguma, Page, & 154 Putterman, 2006; Denant-Boemont, Masclet, & Noussair, 2007; 155 Nikiforakis, 2008). However, studies investigating egalitarian 156 preferences typically use games that measure the degree to which people 157 are willing to reduce others’ income, rather than their own income. A 158 reduction of others’ income is as likely a result of competitive 159 preferences as egalitarian ones. Therefore, it is questionable whether 160 EXPLAINING ANTISOCIAL PUNISHMENT 10 punishment behavior should ever be considered “altruistic”, in the folk161 psychological sense. 162 Researchers working on costly punishment noticed that in 163 behavioral economics experiments some punishment is directed not to 164 free-riders but to cooperators instead (the earliest record of this 165 phenomenon is provided by Ostrom, Walker, & Gardner, 1992). This 166 punishment type has been dubbed, antisocial (Herrmann et al., 2008), 167 spiteful (Falk, Fehr, & Fischbacher, 2005) or perverse (Cinyabuguma et 168 al., 2006). Antisocial punishment, the “sanctioning of people who behave 169 prosocially” (p.1362, Herrmann et al., 2008), is defined in a broader 170 manner than altruistic punishment (see Table 1). Both altruistic and 171 antisocial punishment are costly to the punisher and even more so to the 172 punished, but the definition of antisocial punishment makes no reference 173 to the consequence of such punishment to group cooperation and welfare. 174 Rather, antisocial punishment focuses on the punishment’s target: it is the 175 punishment of those who give more than the punisher. 176 Hermann et al. (2008) found a statistically significant negative 177 correlation between antisocial punishment and cooperative contributions 178 measured across all subject pools. However, as shown in Table 1, 179 antisocial punishment can sometimes be functionally neutral or even 180 altruistic, in the sense that punishing an individual with a higher 181 cooperative contribution can prevent this person from reducing the level 182 EXPLAINING ANTISOCIAL PUNISHMENT 11 of their contributions or even encourage them to contribute more. Such 183 an effect can be enhanced by the fact that, in PGG, punished individuals 184 typically do not know who punished them. As a result, they may suspect 185 that the punishment came from a cooperator and hence is deserved. This 186 thread of reasoning finds support in Herrmann et al.’s (2008) data. In 12 187 out of 16 participant pools, receiving antisocial punishment did not 188 correlate negatively with contributions in the following rounds. 189 In this review, we will stick to the terms “altruistic” and 190 “antisocial” punishment because, although imprecise and ethically 191 loaded, they are well established in the literature. In our opinion, the 192 evidence suggesting the psychologicallyor biologically-altruistic 193 character of punishment is weak. In the experimental setting, the 194 altruistic nature of punishment can be identified only when repeated 195 PGGs are played with different participants in each round, or in one-shot 196 TTPs, but even then it is possible to find selfish explanations for 197 punishment, for example spite. Moreover, punishment of free-riders, 198 instead of positively affecting future contributions, may actually decrease 199 5 Such an effect has been noticed by Herrmann et al. (p.1366, 2008): “Some antisocial punishment can be efficiency-enhancing in intent to induce the punished individual to increase his or her contributions.” 6 See Table S7B in Herrmann et al.’s (2008) supplementary material. Cities where participants decreased cooperation after being a victim of antisocial punishment: Bonn, Minsk, Samara and Istanbul. EXPLAINING ANTISOCIAL PUNISHMENT 12 them (Sylwester, Mitchell & Bryson, submitted). Therefore, in this 200 review we will use altruistic to denote any punishment meted out by 201 cooperators to free-riders. Depending on the study, cooperators are either 202 defined with respect to individual cooperativeness (those who contribute 203 more than, or equally to, another individual are cooperators, while those 204 who contribute less are free-riders) or to average group contributions 205 (those who contribute more than, or equal to, the group mean are 206 cooperators, those who contribute less are free-riders). Antisocial will be 207 used as it was defined by Herrmann et al. (2008). Therefore, any 208 punishment imposed by free-riders on cooperators, or individuals of 209 equal contributions, will be referred to as antisocial. 210 1. The price of punishment 211 Researchers investigating costly punishment typically assume that 212 punishment is more costly to the punisher than to the punished. Due to 213 convention rather than any particular rationale, the most commonly used 214 cost-to-impact ratio is 1:3; it costs the punisher one point to reduce the 215 payoff of the punished individual by three points. Although costly 216 punishment can be considered irrational from the perspective of 217 maximizing the absolute payoff, it does follow a rational rule when 218 relative payoff is prioritized. 219 Expenditure on punishment is strongly affected by the cost-to220 impact ratio. The general finding is that the use of punishment decreases 221 EXPLAINING ANTISOCIAL PUNISHMENT 13 as the punishment price increases (e.g. Anderson & Putterman, 2006). 222 Despite this, some costly punishment (mostly directed at uncooperative 223 individuals) is observed even when the cost to the punisher is larger than 224 the cost to the punished individual. Antisocial punishment does occur, 225 though rarely, even with a high relative cost of punishment (Anderson & 226 Putterman, 2006). 227 There is variation in the results reported concerning sensitivity to 228 the relative cost of punishment. Using data from U.S. participants, 229 Carpenter (2007) analyzed the behavior of free-riders who punished 230 cooperators, cooperators who punished free-riders and free-riders who 231 punished other free-riders. Out of the three groups, free-riders punishing 232 other free-riders were most sensitive to the price of punishment. Free233 riders who punished cooperators did not condition their punishment 234 decisions on price. Carpenter’s results contrast with those obtained by 235 Falk, Fehr & Fischbacher who used a sample of Swiss participants 236 (2005). These researchers found that when the cost of punishment is the 237 same to the punisher as to the punished, antisocial punishment 238 7 In Anderson and Putterman’s (2006) study there were three price-to-impact conditions with ratios in condition 1: 0/100, 30/100, 60/100, 80/100, 120/100, condition 2: 0/100, 5/100, 10/100, 20/100, 30/100 and condition 3: 30/100, 40/100, 50/100, 60/100, 70/100. 8 Free-riding was defined as a negative deviation from the group average. Punishment price-to-impact ratios were as follows:1/4, 1/2, 1/1, 2/1, 4/1. EXPLAINING ANTISOCIAL PUNISHMENT 14 disappears. When punishment resulted in lowering the payoff of the 239 punished person to a greater extent than reducing the cost to the punisher, 240 sanctioning of cooperators by defectors and defectors by other defectors 241 occurred frequently. 242 Egas and Riedl (2008) varied the cost and the impact of 243 punishment and investigated how such a manipulation affected 244 cooperation and punishment decisions in repeated PGGs played by Dutch 245 speakers from around the world. As in Falk, Fehr and Fischbacher’s 246 study, cooperative individuals were willing to punish when the cost to the 247 punisher was equal to, and even when it exceeded the cost to the 248 punished, though in such cases cooperation was not maintained. Unlike 249 in Falk et al.’s study, Egas and Riedl observed antisocial punishment of 250 more cooperative individuals in all cost-to-impact conditions. In 251 agreement with Falk et al.’s results, antisocial punishment was highest 252 when its cost was relatively low in comparison with the impact on the 253 punished (28% of all punishment acts). However, it remained at the level 254 of 22.3% and 18.5% in the two conditions where the cost to the punisher 255 9 In their study there were two price conditions: a low-sanction condition with a price-to-impact ratio of 1/1 and a high-sanction condition in which the price-to-impact ratio of punishing cooperators was 1/3.33 while punishing defectors was 1/2.5. 10 The price-to-impact ratios used by Egas and Riedl were: 1/3, 3/1, 1/1 and 3/3 11 The researchers call this counter-intuitive punishment. EXPLAINING ANTISOCIAL PUNISHMENT 15 was equal to the impact on the punished. Surprisingly, even when the 256 punishing cost exceeded its impact by three times, antisocial punishment 257 was still present (13% of all punishment acts). 258 What happens when punishers themselves can decide about the 259 cost-to-impact ratio of their punishment? Theories of inequality aversion 260 (e.g. Fehr & Schmidt, 1999) suggest that the punisher should use a ratio 261 that would result in minimizing the payoff difference between themselves 262 and the punished. However, if punishment is motivated by the desire for 263 revenge, competition or the pursuit of social status, punishers should 264 adjust the ratio in a way to create an inequality favorable to them. A 265 critical test of these predictions was conducted using the Dictator game 266 with punishment, in which recipients were allowed to decide how much 267 money they wished to deduct from the dictator’s account and where the 268 cost of punishment to the punisher was always $1. Two-thirds of the 269 resultant punishments were inequality-seeking. That is, the punisher 270 decided to deduct from the Dictator more money than was necessary to 271 maintain equality. One-third did deduct only the amount of money 272 necessary to reach equality or less (Houser & Xiao, 2010). 273 Researchers have tended to focus on costly punishment where 274 both the punisher and the punished suffer a cost. It is possible to imagine 275 that non-monetary punishment, in the form of a reprimand that does not 276 affect either the punisher’s or the punished’s payoff, has some effect on 277 cooperation. Indeed, both costly and non-monetary punishment were 278 EXPLAINING ANTISOCIAL PUNISHMENT 16 found to increase cooperation, but the effect of non-monetary sanctions 279 weakened over time (Masclet, Noussair, Tucker, & Villeval, 2003). As in 280 other studies on costly punishment, monetary sanctioning was predicted 281 by both negative and positive deviation from the punisher’s cooperation 282 level, indicating the presence of altruistic and antisocial punishment. 283 However, in the condition where non-monetary sanctions were used, 284 while the effect of altruistic punishment persisted, antisocial punishment 285 was absent. Masclet et al.’s (2003) study is important in that it gives 286 insight into the motivations behind antisocial punishment. The fact that 287 non-monetary reprimands were not used to punish antisocially indicates 288 that the reason for using antisocial punishment is not to change other 289 individuals’ future economic behavior but to negatively affect their 290 payoffs. 291 The presented evidence does not allow for an unequivocal 292 conclusion about how the cost-to-impact ratio of punishment affects 293 antisocial punishment. While some studies show that changing the cost294 to-impact ratio affects antisocial punishment to a greater extent than 295 altruistic punishment and that antisocial punishment is more likely to be 296 reduced when the ratio is unfavorable to the punisher, others do not 297 report such an effect. Despite the mixed findings reported in the studies, 298 it appears that antisocial, rather than altruistic, punishment is more 299 sensitive to the manipulations of the cost-to-impact ratio. In line with this 300 conclusion is the fact that sanctioning cooperators does not occur when 301 EXPLAINING ANTISOCIAL PUNISHMENT 17 their payoffs cannot be altered. Moreover, free-riders who are potential 302 antisocial punishers are less willing to buy costly information about 303 other’s contributions than more cooperative individuals who become 304 altruistic punishers (Page, Putterman, & Garcia, 2008). This suggests that 305 some instances of costly punishment, in particular antisocial punishment, 306 may function as aggressive acts, and are not contingent on the previous 307 cooperative behavior of the punished individuals. In sum, in apparently 308 irrational costly antisocial behavior, the decisions to punish are, at least 309 in some studies, logically tied to the effectiveness of such punishment 310 and to the ability to increase the positive difference between others’ 311 payoffs and one’s own. 312 Cross-cultural variation in punishment 313 A human sense of fairness is omnipresent but takes on different 314 forms around the world (Henrich et al., 2005). A cross-cultural analysis 315 of punishment in UGs and of TTP games revealed a consistent trend; as 316 the offered amount approached an equal split, recipients in the UG and 317 observers in TTP were less willing to punish (Henrich et al., 2006). 318 Interestingly, in some societies a small fraction of recipients sanctioned 319 those whose offers were hyper-fair i.e. those who donated more than an 320 equal split would predict. The suggested reason for such behavior, 321 observed mostly in gift-giving cultures, was the reluctance of recipients 322 to feel indebted to the proposers and the subordinate position resulting 323 EXPLAINING ANTISOCIAL PUNISHMENT 18 from such a debt. In consequence, cooperators were punished 324 antisocially, but, interestingly, in this situation the cost to the punisher 325 was even higher than the cost to the punished. 326 Punishing generous individuals appeared as a leitmotiv in 327 Herrmann et al.’s (2008) cross-cultural study on costly punishment, 328 conducted in 16 comparable subject pools. Participants from different 329 cities across the world played multi-round PGGs, with each round 330 followed by a punishment opportunity. Herrmann et al. (2008) found that 331 the level of antisocial punishment, measured as punishment towards 332 individuals whose PGG contributions were equal to or exceeded the 333 punisher’s contributions, varied dramatically across societies. Notably, 334 high levels of antisocial punishment were observed in Greece, Turkey, 335 the former Soviet Union and the Middle East while lower levels were 336 found in the U.S, Australia, the Far East and Northwestern Europe. 337 Previous experiments, conducted in places with low levels of antisocial 338 punishment, showed that the opportunity to punish positively affected 339 group cooperation. However, not surprisingly, in subject pools where 340 12 In splits where the proposer offers more than a fair share to the recipient (e.g. 30 for the proposer and 70 for the recipient), a recipient who rejects the offer suffers a higher cost (70) than the “punished” proposer (30). 13 Scandinavia, the UK, Germany & Switzerland. Southwestern Europe, e.g. France, Spain & Italy were not tested. EXPLAINING ANTISOCIAL PUNISHMENT 19 cooperators were punished heavily, cooperation levels did not increase 341 with punishment. 342 In an attempt to explain the observed cross-cultural variation, 343 Herrmann et al. investigated possible relationships between antisocial 344 punishment and a number of socio-demographic factors. Democracy 345 ranking and a measure of the prosperity of a country (GDP per capita) 346 were negatively correlated with antisocial punishment, suggesting that 347 high socio-economic development coincides with the cooperation348 enhancing function of punishment. Antisocial punishment was also 349 related to various cultural dimensions of the investigated countries (see 350 Hofstede, 2001) e.g. it occurred more often in places where the inequality 351 in society was high (high Power Distance), where ties between 352 individuals and their in-group are strong (low Individualism), where 353 gender differences tend to fade away (low Masculinity) and where 354 uncertainty avoidance is high. 355 In their analysis, Herrmann at al. (2008) emphasized two factors 356 as possible explanations for the cross-cultural variation in antisocial 357 punishment: the norms of civic cooperation and the rule of law. The 358 norms of civic cooperation is a measure based on questions used in the 359 World Values Survey describing the strength of abiding cooperative 360 norms in a society and the level of disapproval for breaking them. The 361 rule of law is an indicator developed by the World Bank to describe the 362 extent to which people perceive their government, police, courts and 363 EXPLAINING ANTISOCIAL PUNISHMENT 20 authorities as fair, trustworthy and effective at law enforcement. Both 364 measures were negatively correlated with antisocial punishment. 365 Additionally, the researchers investigated a link between Inglehart’s cultural 366 dimensions “traditional vs. secular-rational values” and “survival vs. self367 expression values” and antisocial punishment. They found less antisocial 368 punishment in cities where self-expression values i.e. social liberties and 369 personal freedom mattered more than survival values, which represent 370 economic and physical security. 371 With so many interdependent predictors of antisocial punishment, 372 it is difficult to determine their relative importance and assess their 373 explanatory power. While Herrmann et al. focused on predictors 374 involving ethical evaluation of certain behaviors by the society (norms of 375 civic cooperation); and the quality, efficiency and fairness of a 376 centralized sanctioning system within a society (rule of law), it is 377 possible to imagine that differences in antisocial punishment are driven 378 by other societal characteristics. For example, if antisocial punishment is 379 proximately motivated by dominance and the desire for social control, it 380 would be reasonable to focus on its relationship with power distance and 381 survival/self-expression values. High levels of antisocial punishment 382 would be expected in places where social hierarchy and demonstration of 383 14 This correlation is unsurprising given that Inglehart’s “survival vs. selfexpression values” are related to Hofstede’s power distance and IndividualismCollectivism dimensions (Inglehart & Welzel, 2005). EXPLAINING ANTISOCIAL PUNISHMENT 21 power play an important role, and in harsher environments where 384 individuals need to focus on local competition with their neighbors in 385 order to succeed. 386 The variation in cooperation observed in Herrmann et al.’s (2008) 387 data was affected by individual heterogeneity and group-level differences 388 and most importantly by the membership in a “world culture” (Gächter, 389 Herrmann, & Thöni, 2010). Apart from the cultural differences in the 390 average cooperation level when punishment was possible, there were also 391 some interesting differences in the patterns of reacting to punishment. In 392 subject pools with high levels of antisocial punishment, the level of 393 cooperation remained low but relatively stable. In contrast, in places 394 where punishment of free-riders dominated and antisocial punishment 395 was scant, some participants, when the opportunity to punish was 396 introduced, almost immediately increased their pro-social contributions 397 (e.g. Boston, Nottingham, Copenhagen, Bonn, Zurich and St Gallen). In 398 other subject pools the increase in cooperation occurred gradually over 399 the course of rounds (e.g. Seoul, Chengdu and Melbourne). In general, 400 clustering the subject pools according to the Inglehart and Baker (2000) 401 schema did approximate the patterns of the reactions to punishment but 402 15 World cultures have been defined following Inglehart and Baker (2000) and Hofstede (2001) as a way to capture their historical and cultural backgrounds. EXPLAINING ANTISOCIAL PUNISHMENT 22 there were exceptions. Melbourne, categorized as an English speaking 403 culture, together with Nottingham, displayed a pattern similar to those 404 observed in the cities of the Confucian culture-type. Boston, on the other 405 hand, resembled the pattern observed in protestant non-English speaking 406 Europe. 407 Running identical experiments with the same experimenter and 408 instructions allows us to unravel cross-cultural variation in antisocial409 punishment behavior. By employing a slightly different design, and 410 comparing the behavior in subject pools from two countries, we may 411 illuminate other cross-cultural patterns, not visible using the earlier 412 experimental method. While costly punishment increases cooperation in 413 Boston (Dreber, Rand, Fudenberg, & Nowak, 2008), it does not do so in 414 Beijing (Wu et al., 2009). In contrast, Herrmann et al. (2008) found that 415 the opportunity to use punishment positively affected contributions in 416 both subject pools, and that both Chinese participants from Chengdu and 417 US participants from Boston exhibited similar levels of costly 418 punishment, with only marginally higher level of antisocial punishment 419 in China. Unlike in Hermann et al.’s paradigm with a PGG, in Dreber et 420 al.’s and Wu et al.’s experiments participants had an opportunity to 421 cooperate, defect or punish within a dyad, in each round. Wu et al. (2009) 422 discovered high levels of indiscriminate punishing in China in 423 comparison to the US. The researchers explained the differences between 424 theirs and Herrmann et al.’s study by the differences between protocols 425 EXPLAINING ANTISOCIAL PUNISHMENT 23 used. In the repeated PGG, Chinese participants might have recognized 426 the concept of reputation, so important in their culture, whereas in the 427 dyadic encounters this concept was not applicable. 428 Another cross-cultural study, conducted by Gächter and 429 Herrmann (2009), supported their 2008 results. In an experiment 430 comparing antisocial punishment in Swiss and Russian participants, it 431 was confirmed that the punishment directed at cooperators in one-shot 432 games meted out by Russian participants was higher than antisocial 433 punishment in Switzerland. What merits attention is that participants in 434 both investigated regions could accurately predict the levels of antisocial 435 punishment, which suggests that common cultural origins predispose 436 people to correctly assess the cooperative and uncooperative intentions of 437 the members of their cultural group. In Russia, participants exhibited 438 more exploitative behavior in the sense that, even if they expected high 439 levels of cooperation from others, their own cooperative contribution was 440 lower than Swiss participants’ contributions. Introducing punishment had 441 a positive effect on cooperation in Switzerland but a detrimental effect on 442 cooperation in Russia. In the latter case, this effect was mostly driven by 443 the change in the behavior of top contributors, who, presumably 444 expecting antisocial punishment, became less cooperative. 445 16 The reference level was the group average. EXPLAINING ANTISOCIAL PUNISHMENT 24 Evidence that an opportunity to punish produces different types of 446 behavior in different cultures is growing. In a recent study, American and 447 Romanian students showed a similar level of cooperative behavior when 448 it was measured by games without punishment (Ellingsen et al., 2012). 449 However, in repeated PGGs with punishment, American students tended 450 to use cooperation-enhancing altruistic punishment, while Romanian 451 students frequently meted out antisocial punishment. Interestingly, 452 Romanian students often used indiscriminate punishment targeting both 453 cooperators and non-cooperators. This finding is in line with our re454 analysis of Herrmann et al.’s dataset (Sylwester, Mitchell & Bryson, in 455 preparation), showing a non-exclusive use of antisocial and altruistic 456 punishment. 457 It is plausible to expect that, within a given culture, socio458 demographic factors will modulate the occurrence of antisocial 459 punishment, as they do with cooperation and third-party punishment 460 (Marlowe et al., 2011). In a study conducted in rural and urban Russia, 461 socio-demographic variables were found to affect cooperative but not 462 punishing behavior (Gächter & Herrmann, 2011). High levels of 463 antisocial punishment were unrelated to the age group and region of the 464 sample but, surprisingly, participants with a university degree and those 465 who were members of a voluntary organization exhibited higher levels of 466 EXPLAINING ANTISOCIAL PUNISHMENT 25 antisocial punishment. It is important to note, however, that one-shot 467 games were used in that experiment and different patterns might be 468 revealed if participants are allowed to interact in the same group for a 469 longer period of time, as in Herrmann et al. (2008). 470 So far, the evidence gathered by Herrmann et al. (2008) provides 471 the most complete picture of antisocial punishment in different cultures. 472 The patchwork of other studies that differ in methodology do not 473 facilitate a coherent theory of the driving forces behind the variation in 474 antisocial punishment. The direction of the correlations between 475 antisocial punishment and different socio-economic factors suggests that 476 certain conditions can contribute to its occurrence. More specifically, it 477 appears that antisocial punishment frequently takes place in cultures 478 where the potential cost of it is low in relation to its benefits, for 479 example, in places where norms are frequently infringed, free-riding is 480 commonly approved of and legal sanctioning institutions are not 481 perceived as being fair or efficient. In such places, the potential cost of 482 being caught red-handed when punishing cooperators is low in 483 comparison to places where unethical behavior is strongly penalized and 484 disapproved of by both members of the society and legal institutions. On 485 the other hand, we observe antisocial punishment in places where there is 486 17 Though voluntary organisations in the former Soviet Union might have a different character from voluntary organisations in established market economies. EXPLAINING ANTISOCIAL PUNISHMENT 26 a lot to be gained from acquiring a higher rank in the group (even at a 487 cost of the absolute payoff) and where status and power may have a 488 dramatic impact on the quality of life and survival. In cultures with high 489 power distance the benefits coming from having a dominant status are 490 much higher than where power distance is low. In places abundant in 491 resources and with low inequality, gaining power might bring smaller 492 ecological benefits than in places where resources are low and 493 competition is fierce. 494 Antisocial punishment at the group level 495 Variation in antisocial punishment occurs at various levels. 496 Starting from the top, we can consider cultures (e.g. as defined by 497 Inglehart & Baker, 2000), populations within a culture, groups within a 498 population and individuals within a group. Micro-level behavior 499 modulates macro-level, so examining individual drives and social 500 influences within different environments may help explain variation in 501 the cultural make-up. In this section, we discuss betweenand within502 group competition that may be affecting the observed variation in 503 antisocial punishment. Punishment can be imposed within one’s own 504 close social group or it may be inflicted on individuals from another 505 group. Since altruistic punishment enhances group welfare in the long 506 run (Gächter, Renner, & Sefton, 2008) while antisocial punishment can 507 be expected to decrease it, the use of these two types of punishment 508 EXPLAINING ANTISOCIAL PUNISHMENT 27 towards in-group and out-group members should be contingent on the 509 severity of interand intra-group competition. 510 Inter-group competition 511 The parochial preferences widely documented in humans 512 manifest themselves in people favoring individuals from their own social 513 group (Tajfel, 1970). In-group favoritism can occur in any situation 514 where an individual has an option to positively or negatively affect 515 another individual’s well-being. Hence, we should be able to observe 516 selective use of altruistic and antisocial punishment towards in-group 517 versus out-group members. Costly altruistic punishment might be a 518 useful tool for enhancing a group’s cohesion and cooperation, 519 particularly when it is done within one’s own social group and not 520 inflicted on out-group members. In contrast, antisocial punishment, 521 which is likely to result in reducing group cooperation and coordination, 522 could be an effective way to gain competitive advantage over another 523 group when inflicted on members of an out-group. This in-group out524 group reasoning might be underlying the observed variation in antisocial 525 punishment. Excessive generosity displayed by some individuals can 526 possibly be interpreted as a signal of dominance rather than cooperation. 527 High status of these cooperative individuals distinguishes them form the 528 rest of the group. In consequence, cooperators are not perceived as in529 EXPLAINING ANTISOCIAL PUNISHMENT 28 group members and fall victim of antisocial punishment. 530 When costly punishment is meted out within one’s own group, 531 effective altruistic punishment and inhibited antisocial punishment will 532 positively affect the collective payoffs of individuals as a group This, in 533 turn, can increase the odds of one group gaining advantage over another 534 in between-group competition. Where between-group competition has 535 significant consequences, being a relatively weak individual in a 536 dominant group may be better than being a dominant individual in a 537 subordinate group (Queller, 1994; Wilson, 2004). 538 The same logic can be applied to a situation when individuals 539 have an opportunity to punish members of an out-group. It is reasonable 540 to expect that with a higher degree of between-group competition the use 541 of antisocial punishment towards out-group members will increase. 542 Directing antisocial punishment to out-group members may undermine 543 the out-group’s cooperation or make the mechanism of norm 544 enforcement through altruistic punishment less effective. Either could 545 increase the competitive status of the punisher’s own group. 546 Indeed, experiments conducted in Papua New Guinea with two 547 18 In a recent study, U.S participants voted to expel from the group not only the most selfish members, but also the ones who excessively contributed to the public good and used little of it (Parks & Stone, 2010). Social comparison mechanisms, combined with the unwillingness to adhere to high norms established by the overgenerous individuals, were proposed as explanations for this phenomenon. EXPLAINING ANTISOCIAL PUNISHMENT 29 distinct social groups revealed that altruistic punishment was highest 548 when the person in charge of the split, the recipient and the punisher 549 came from the same social group, and also when only the recipient and 550 the punisher came from the same group (Bernhard, Fischbacher, & Fehr, 551 2006). Most antisocial punishment was observed in the latter case, 552 confirming that punishers were more likely to punish in a way that 553 negatively affected payoffs of an out-group member. In another study 554 with artificially created groups of Japanese participants, a similar pattern 555 was observed (Shinada, Yamagishi, & Ohmura, 2004). Punishing of 556 free-riders by cooperators happened more frequently when done within 557 one’s own group (this result was also obtained by McLeish & Oxoby, 558 2007), but, interestingly, free-riders meted out harsher punishment on 559 other free-riders from an out-group rather than in-group. In Shinada et 560 al.’s (2004) study, antisocial punishment was minimal and no in561 group/out-group effects were reported. Perhaps this is unsurprising, 562 given Japan’s high GDP and the strong rule of law in that country. 563 One-shot TPP experiments have also been conducted in India to 564 investigate the impact of the different caste memberships on punishing 565 behavior. While high-caste participants punished norm violators more 566 severely than low-caste participants (Hoff, Kshetramade, & Fehr, 2009), 567 19 The group distinction was created by telling participants that the other members either belonged to their own or a different academic unit. EXPLAINING ANTISOCIAL PUNISHMENT 30 the caste differences in the punishment of cooperators were not 568 significant (Fehr, Hoff, & Kshetramade, 2008). Investigating spiteful 569 behavior of low and high castes, using a series of binary choice Dictator 570 games (in which one person decided about the split of a given amount of 571 money), provided mixed results. When presented with a choice between 572 70/90 (other/self) distribution and 90/90 distribution, 42% of high caste 573 participants and only 21% of low cast participants chose the first 574 (spiteful) option. In contrast, when deciding between 150/150 and 575 100/160 distributions, 83% of high caste and only 53% of low caste 576 participants chose the first (equal split) option (Fehr et al., 2008). In the 577 seven possible choices, high caste participants preferred the spiteful 578 distribution more than low caste participants in only one case (in which 579 the p value was marginally significant). However, the researchers 580 concluded that “high-caste subjects (compared to low-caste subjects) are 581 considerably more likely to reduce others’ payoffs if behind, or to take 582 other spiteful actions” (p.499, Fehr et al., 2008). 583 Mere in-group/out-group categorization may not invoke hostility 584 and antisocial sanctions. As argued above, what triggers inter-group 585 conflict and aggression is the social level at which the most significant 586 competition takes place. In a sample from Swiss army platoons, group587 membership per se did not affect the occurrence of antisocial punishment 588 but resulted in more altruistic punishment when the victim of defection 589 was in-group and the defector was out-group (Goette, Huffman, Meier, & 590 EXPLAINING ANTISOCIAL PUNISHMENT 31 Sutter, 2010). However, when between-group competition was 591 introduced, costly punishment was mostly imposed on cooperators and 592 free-riders from the out-group. At the same time, in-group cooperation 593 increased. This points to an important role inter-group competition plays 594 in inducing antisocial punishment (and Schadenfreude, see Leach, 595 Spears, Branscombe, & Doosje, 2003). Competition with the out-group 596 can also induce excessive and wasteful punishment of in-group members. 597 Contests between groups resulted in above-rational expenditures on 598 competition but also in high expenditures on within-group punishment of 599 individuals whose financial engagement in the conflict was lower than 600 the group’s average (Abbink, Brandts, Herrmann, & Orzen, 2010). High 601 expenditures on costly punishment in the presence of competition have 602 also been found by Sääksvuori et al. (2011). 603 The levels of antisocial punishment observed in conventional 604 PGG experiments appear to be low when contrasted with the levels 605 towards the out-group members induced by conflict. A possible 606 interpretation of this would be that punishment in ordinary PGG is only a 607 side-effect of mechanisms evolved for conflict situations. The act of 608 costly punishment, when taken out of the PGG context, can be perceived 609 as mere aggression. Engaging in aggressive interactions with out-group 610 members in the presence of conflict may be advantageous, in that it may 611 help preserve a group’s resources such as territory. Herrmann et al. 612 (2008) found a negative correlation between antisocial punishment and 613 EXPLAINING ANTISOCIAL PUNISHMENT 32 scores on the individualism/collectivism dimension. Antisocial 614 punishment occurred more often in places where group identity plays a 615 great role and where, in general, ethnocentrism and xenophobia are more 616 pronounced. A possible explanation for this pattern might be that 617 participants perceived other anonymous players as members of an out618 group rather than in-group. 619 When extracted from the context of PGGs, costly punishment 620 might be an effective weapon used in inter-group conflicts because the 621 cost of aggression is smaller than its consequences to the opponent. 622 Using altruistic punishment in conflict, although still effective at the 623 individual level, might not work in the long term, because it may result in 624 the out-group becoming more cooperative and coherent. Instead, 625 antisocial punishment of out-group cooperators undermines the stability 626 of the other group’s social norms. 627 Intra-group competition 628 In ecological contexts where intra-group competition is fierce, 629 individuals will use aggression towards members of their own group. 630 Costly punishment typically decreases average payoffs (Dreber et al., 631 2008; Wu et al., 2009), however, it might be useful for displaying 632 aggression and gaining relative advantage over the punished individual. 633 Previous research has shown that people do care about their relative 634 payoff within a group. For example, Saijo and Nakamura (1995) made 635 participants face a non-dilemma in which the payoff maximizing choice 636 EXPLAINING ANTISOCIAL PUNISHMENT 33 was to contribute the whole allocation to the public pool. Although the 637 contributions to PGG in the non-dilemma condition were higher than in 638 the standard dilemma, a considerable number of participants still 639 refrained from contributing and failed to maximize their payoff. The 640 average amounts saved in the non-dilemma situation were higher than the 641 average investments to the public pool in the traditional dilemma version 642 of the PGG. This indicates that in the no-dilemma situation more 643 participants chose the non-optimal (non-payoff-maximizing) outcome 644 than in the traditional dilemma, indicating that the non-dilemma may 645 have been taken as a spiteful dilemma. 646 Saijo and Nakamura (1995) concluded that there exists a 647 population of spiteful individuals who value their ranking within the 648 group more than their absolute payoff. In a similar but more recent study, 649 even when the payoff maximizing decision was to contribute everything 650 to the public pool, a considerable number of participants did not do that 651 (Kümmerli, Burton-Chellew, Ross-Gillespie, & West, 2010). The 652 researchers described this phenomenon as “resistance to extreme 653 strategies” or “imperfections” and discovered that a considerable number 654 of participants perceived their group members as competitors rather than 655 20 Saijo and Nakamura (1995) used two variants of the PGG marginal per capita return from each point invested: low return (standard PGG) where each invested point yields 0.7, and high return (anti-dilemma) where each invested point yields 1.43