Reputation Concern versus In-group Identification as
Underlying Mechanisms of Parochial Cooperation in relation with Social Value Orientation
Thanaphat Thongpaibool University of Amsterdam
Hillie Aaldering University of Amsterdam
Abstract
In intergroup settings where cooperation is possible between two or more groups, parochial cooperation—making self-costly contributions to one‘s in-group—is pervasive and generally preferred over universal cooperation which benefits all groups. In two experiments, using the Collective Incentive game, which allows for cooperation on the in-group and/or the collective level, we investigated the potential drivers of parochial cooperation by manipulating in-group identification and reputation concern based on Social Identity Theory (SIT) and the Theory of Bounded Generalized Reciprocity (BGR). We also examined the moderating role of social value orientation on parochial cooperation. Experiment 1 tested whether low in-group identification and/or low reputation concern decreased parochial cooperation, compared to a control condition (N = 204). Experiment 2 compared the differences between high vs. low in-group identification and high vs. low reputation concern on parochial cooperation (N = 374). Results showed
tentative support for SIT only in Experiment 2, and no firm conclusion can be reached regarding the drivers of parochial cooperation. Parochial cooperation consistently prevailed over universal cooperation and was motivated by a preference to favor the in-group. When parochial
cooperation harms the out-group, universal cooperation was preferred (Experiment 1 and 2). A pro-social inclination was positively associated with universal cooperation (Experiment 1 and 2), but with parochial cooperation only in Experiment 2. Pro-sociality, however, did not affect how in-group identification and reputation concern impact parochial cooperation (Experiment 1 and 2). We discuss implications for both theories and another possible determinant driving parochial cooperation, namely outcome dependence.
Keywords: Intergroup Setting, Parochial Cooperation, Universal Cooperation, Social Value Orientation
Reputation Concern versus In-group Identification as
Underlying Mechanisms of Parochial Cooperation in relation with Social Value Orientation
As social beings, cooperation and interdependency among individual humans within and between groups are ubiquitous. The concept of groups is pervasive and entrenched into our complex social lives, ranging from small-sized kin-based families to nationalities. A group can be defined as “a collection of individuals who perceive themselves to be members of the same social category…” (Tajfel & Turner, 1979, p. 40). In this sense, group affiliation may be multidimensional and cross-cutting, based on a variety of objective or subjective criteria
(Brewer, 1999). For example, a single individual may perceive him/herself as member of a group by virtue of nationality, occupation, religious belief, gender, or even an arbitrary artificially-induced group feature. Throughout human history, cooperation between groups is necessary for individuals as well as their groups to survive and prosper (Brewer, 1999). Especially in the present globalized world, individuals not only interact and cooperate with their own group but also with individuals from other groups—e.g. through research collaborations and international business deals (e.g. Arrow, 2007; Choi & Bowles, 2007). An interesting question, then, is: once individuals associate themselves with particular groups, how does this impact their cooperative behaviors within and between groups?
Previous research has well documented that in social decision-making situations
individuals have a tendency to relatively favor the members who belong to the same group (i.e. in-groups) and to discriminate against members of different group(s) (i.e. out-groups). This behavioral tendency usually manifests itself as a positive evaluation towards the in-group members, namely in-group favoritism or in-group love, and/or a negative evaluation of the
out-group members, namely out-out-group derogation or out-out-group hate (for a review see De Dreu, Balliet, & Halevy, 2014). For example, people perceive members of the in-group more positively than members of the out-group (Balliet, Wu, & De Dreu, 2014; Brewer, 1979).
However, situations become more complicated in an intergroup setting, where interdependency among individuals is usually not restricted within a group but expands to cooperation between members of different groups. In general, intergroup situations are considered to have a nested social structure, where individuals not only belong to small subgroups but also to a collective group as a whole (Wit & Kerr, 2002). For instance, in an organization, individuals usually do not only work with their own team (i.e. a subgroup) but they have to collaborate with other teams within the department (i.e. a collective group) as well. Hence, in intergroup situations, the outcome of cooperation for each individual is also
interdependent between both in- and out-groups. In this way, intergroup situations create a social dilemma such that individuals have to decide between self-interest and other available
cooperative options (Aaldering et al., In prep; see Dawes, 1980). A bulk of evidence suggests that when cooperation is possible with two or more parties, individuals, even at the expense of themselves, are generally inclined to make contributions to their in-group which may
consequently negatively impact the out-group, at the expense of themselves—i.e. parochial cooperation (e.g. Aaldering, 2014; Balliet et al., 2014).
In this research, inspired by the findings of Aaldering, Ten Velden, Van Kleef, & De Dreu (in prep), we conducted two experiments using an experimental game with a nested intergroup structure to investigate the psychological and personality determinants that may explain the emergence of parochial cooperation in such interdependent intergroup situations. Despite the high prevalence of parochial cooperation, it is still unclear why individuals would
engage in parochial cooperation. In other words, what exactly is the driving mechanism or motive of parochial cooperation? Here, we report two experiments to precisely test predictions from the Social Identity Theory (Tajfel & Turner, 1979) and the Bounded Generalized
Reciprocity theory (Yamagishi, Jin, & Kiyonari, 1999) whether in-group identification and/or reputation concern underpins parochial cooperation. Specifically, the first experiment examines whether low in-group identification and/or low reputation concern would decrease parochial cooperation. The second experiment investigates whether there would be a difference in parochial cooperation when in-group identification is high vs. low and/or when reputation concern is high vs. low. Furthermore, in addition to such psychological drivers, individuals‘ tendency to cooperate may, to some extent, be determined by individuals‘ dispositional traits or personality associated with pro-social behaviors—namely, social value orientations. Therefore, we also investigate how social value orientations interact with psychological motives (in-group identification or reputation concern) in a manner that may influence individuals‘ parochial cooperation.
The remainder of this introduction is organized as follows: in the next section, we provide a general outline of the theoretical background, primarily define the term ‗parochial cooperation‘, and discuss the influential theories—i.e. The Social Identity Theory (SIT) and the Bounded Generalized Reciprocity (BGR) theory—on the psychological mechanisms that potentially drive parochial cooperation. We also review existing research that has compared the predictive power of each theory. We then review how previous studies investigate cooperation in intergroup situations using experimental intergroup games and discuss the advantages and drawbacks of existing games. Next, we review how individuals‘ dispositional traits, namely social value orientation, may fundamentally influence cooperative behaviors, and we provide
existing empirical examples found thus far in intergroup situations where cooperation is possible between two or more parties. Based on the theories and existing literature, we put forward corresponding hypotheses and test them in two experiments.
Theoretical Accounts of Parochial Cooperation
In the current study, following Aaldering et al. (In prep), we define parochial cooperation exclusively in terms of behavior, rather than motives or preferences, as “in-group bounded cooperation which has direct or indirect adverse consequences for the out-group, in the sense that the out-group profits less from the cooperation than the in-group or is even directly harmed by the in-group cooperation.” By this definition two types of parochial cooperation can be distinguished: one that solely benefits the in-group (henceforth: ―benign parochial cooperation‖) and one that favors the in-group while simultaneously hurts the out-group (henceforth:
―competitive parochial cooperation‖). This clearly disambiguates whether the necessity to harm the out-group is a prerequisite for parochial cooperation. Also, with this definition, parochial cooperation can be distinguished from ―universal cooperation,‖ which, here, is defined as cooperation that is beneficial to the collective whole—i.e. including both the in-groups and the out-groups (Aaldering et al., In prep; Thielmann & Böhm, 2016).
In intergroup situations where two or more groups have conflicting interests, individuals often face a dilemma between fulfilling their self-interest, cooperating only with their own in-groups which may (i.e. competitive parochial cooperation) or may not hurt the out-group (i.e. benign parochial cooperation), or contributing to the collective benefit of both the in-group and the out-group (i.e. universal cooperation). Previous research illustrates that individuals are more likely to engage in parochial cooperation over universal cooperation (Polzer, 2004; Wit & Kerr, 2002), despite the fact that universal cooperation benefits both the in-group and the out-group
and usually yields the highest possible outcome for all the involved parties. An example of such complex intergroup situations can be well captured during war. While an act of competitive parochial cooperation involves taking part in the military as a frontline soldier, volunteering to be a military doctor can be considered as a form of benign parochial cooperation. On the other hand, since in the end both opposing sides would suffer from the war, being a peace activist supporting an armistice illustrates universal cooperation. Although parochial cooperation may provide a functional advantage that benefits the welfare of the in-group (e.g. strengthening the bond and integrity among members of the in-group), in the long run it can also cause adverse societal effects to the collective whole (e.g. fueling more conflicts in war time). As parochial cooperation seems to be pervasive in human society, understanding how and why it emerges is vital.
Several theories from evolutionary as well as social psychology have been proposed to explain how parochial cooperation may emerge in intergroup situations. Here, we discuss two prominent theories: the Social Identity Theory (Ellemers, Spears, & Doosje, 2002; Tajfel & Turner, 1979) and the Bounded Generalized Reciprocity (Yamagishi, Jin, & Kiyonari, 1999).
From the field of social psychology, the Social Identity Theory (SIT) rests on three general assumptions (e.g. Tajfel, Flament, Billig, & Bundy, 1971). First, individuals are motivated to maintain and develop their self-esteem. Second, social groups are connoted with positive or negative values, and these group values also reflects a social identity of its
individuals‘ group members. Third, as individuals self-categorize or identify as belonging to a specific group, group evaluation is determined comparatively to other salient groups (i.e. the out-groups). According to the theory, a social identity can be distinguished from a personal identity as „the part of individual‟s self-concept which derives from his knowledge of his membership of a
social group (or groups) together with the emotional significance attached to that membership‟ (Tajfel, 1974, p. 69). In other words, individuals obtain part of their social self-image from the social group with which they identify (Aron, Aron, & Smollan, 1992; Tropp & Wright, 2001). On the basis of these assumptions, the SIT posits that to enhance and maintain self-esteem individuals strive to differentiate their own in-group positively from the out-group in order to develop a positive social identity, as part of a group to which they belong (Tajfel & Turner, 1979). In the context of intergroup cooperation, this can be achieved via parochial cooperation: either only contributing to the in-group without hurting the out-group (benign parochial
cooperation) or favoring the in-group while harming the out-group (competitive parochial
cooperation). In this way, individuals can make favorable comparisons between the in-group and the out-group, which may but not necessarily include aggression towards the out-group. In other words, SIT endorses both forms of parochial cooperation, though hostility towards the out-group can be a by-product and is not necessary. According to SIT, the more strongly individuals
identify with the group, the more likely they would engage in parochial cooperation. Thus, within SIT, a key psychological mechanism underlying parochial cooperation is in-group identification, which is the degree to which individuals identify themselves as part of a specific group. For example, Leonardelli and Brewer (2001) found that when in-group identification was experimentally induced, participants (from both minority and majority groups) who identified more strongly with their in-group allocated more rewards to their in-group, compared to those who identified less strongly.
Another theory that attempts to explain parochial cooperation is the Theory of Bounded Generalized Reciprocity (BGR) (Yamagishi et al., 1999). It is based on an evolutionary
and prosper (Aaldering et al., In prep; Balliet et al., 2014; Brewer, 1999). Because of
interdependent relationships within a group, throughout human evolution, a norm of cooperation and (indirect) reciprocity can be established based on group membership within a group (Mifune, Hashimoto, & Yamagishi, 2010). Within BGR, individuals are motivated to make costly
contributions to their group members (such as, via parochial cooperation), and expect other in-group members to do the same/reciprocate (see also Everett et al., 2015; Kiyonari & Yamagishi, 2004). This network of (indirect) reciprocity creates an outcome-dependent structure between in-group members (see Rabbie, Schot, & Visser, 1989). That is, individuals‘ outcomes are
dependent not only on their own decisions, but also on other in-groups‘ reciprocity. According to BGR, since individuals‘ outcomes are dependent on their in-group, individuals‘ decisions to engage in parochial cooperation are driven by the expectation of (indirect) reciprocity that the other in-group members—though not necessarily the same in-group members they previously cooperated with—will reciprocate (Kiyonari & Yamagishi, 2004; Rabbie et al., 1989; Yamagishi et al., 1999). To achieve such in-group cooperation and indirect reciprocity, developing and maintaining a positive reputation of being cooperative among in-group members is thus
necessary (Aaldering et al., In prep; Everett et al., 2015; Yamagishi et al., 1999). On top of that, because in-group membership also indicates a higher probability of repeated contact (Rusch, 2014), the presence of a salient group membership, consequently, provides a heuristic about whom to trust, depend on, cooperate with, and expect reciprocity from in the future encounters (Everett et al., 2015). Thus, individuals are more likely to try to build a positive reputation within their own group (relative to the out-group) by favoring the in-group members (e.g. through parochial cooperation) because they expect some form of beneficial outcomes from the other in-group members. Importantly, within the BGR, the motive for parochial cooperation is mainly to
favor the in-group (benign parochial cooperation) in order to maintain a positive reputation and to reach valuable group outcomes, thus it does not necessarily involve hostility towards the out-group (competitive parochial cooperation). Therefore, within the BGR theory, because of the in-group outcome dependency and the expectation that their in-in-group members will reciprocate, individuals are concerned about their (positive) reputation, and this reputation concern drives parochial cooperation.
In sum, both SIT and BGR predict that individuals are inclined to engage in in-group bounded parochial cooperation. However, this prediction for parochial cooperation is based on different psychological underlying mechanisms, and it is unclear to what extent parochial cooperation is driven by each mechanism (see Everett et al., 2015). While SIT predicts that parochial cooperation is driven by in-group identification, BGR predicts that parochial cooperation is a result of reputation concern because individuals strategically expect in-group members to reciprocate.
Thus far, only a few prior studies have compared and provided some clues on the
predictive power of each theory, by manipulating various aspects of each theory such as outcome dependence, group goals, and in-group identification (Balliet et al., 2014; Everett et al., 2015; Gaertner & Insko, 2000; Rabbie et al., 1989; Romano, Balliet, & Wu, 2017; Scheepers, Spears, Doosje, & Manstead, 2002; Stroebe, Lodewijkx, & Spears, 2005), but the findings are
inconclusive. For example, Rabbie et al. (1989) found that participants allocated more rewards to in-group members as they had higher expectation for reciprocity from the in-group members. However, their allocations were not particularly motivated by group identification. The results thus support BGR in contrast to SIT. Gaertner and Insko (2002) also found more support for the BGR as they unconfounded outcome dependency and group identification. Their findings
showed that even though participants identified with their in-groups, parochialism occurred only when the outcome was in-group dependent. The authors concluded that the in-group
identification, although powerful, is not sufficient to promote parochial cooperation. Along similar lines, Stroebe and colleagues (2005) orthogonally manipulated outcome dependence contrasting between the in-group and the out-group, as well as the strength of in-group
identification. They found that when comparing strong versus weak in-group identification, there was no effect of in-group identification on in-group-rewarding allocations. However, participants made more in-group-rewarding allocations in the in-group outcome dependence condition
compared to the out-group outcome dependence condition, which supports BGR. Gagnon and Bourhis (1996) also examined the influence of outcome dependence and identification, but on the contrary they found more support for SIT that participants who identified more strongly with their in-group allocated more rewards to the in-group. The outcome dependence and the
expectation of in-group reciprocity did not have an effect on in-group favoring reward
allocations, thus providing no support for BGR. However, their results were criticized by Stroebe et al. (2005) who claimed that identification was not varied independently but was assessed after the completion of the allocation task, which undermines causal conclusions (this was also acknowledged by the authors themselves). Most recently, Romano and colleagues (2017) found more support for BGR compared to SIT by manipulating reputation concern, group
identification, and partner‘s group membership across five studies. They found that participants cooperated more with the in-group, as well as the out-group, when their reputation was at stake. In-group identification, however, did not have an influence on cooperation.
While the findings discussed above seem to point in the direction of the BGR, research that precisely tests the influence of reputation concern within the BGR (with an exception of
Ramono et al., 2017) is still lacking, and thus is needed to understand how reputation concern may play a role in driving parochial cooperation. Also, importantly, the research paradigm used in these studies did not have a clear intergroup structure, which is characterized by outcome interdependence between both in-group and out-group. For example, some studies (e.g. Gagnon & Bourhis, 1996; Rabbie et al., 1989) compared the predictive power of each theory using the reward allocation task which only encapsulates how much individuals are willing to favor the in-group, relative to the out-group. Thus, a better research paradigm that can capture different cooperative decisions in the intergroup setting—where cooperation is possible and
interdependent between two or more groups—needs to be conducted in order to gain a more thorough understanding of the nature of parochial cooperation.
Intergroup Situations as Experimental Team Games
Recall that in intergroup situations individuals often encounter a tension between two or more of the following: self-interest, benign parochial cooperation, competitive parochial
cooperation and universal cooperation. Such complex social dilemmas can be modeled by a variety of experimental team games. Generally, in these games, individuals are assigned to one of the two groups, which are (either implicitly or explicitly) nested in the collective group as a whole. Individuals are then given a certain amount of endowments and have to decide between different cooperative decisions. Among existing literature, different interpretations of parochial cooperation and their assumed underlying motivations vary depending on the game structure. For example, in the standard Nested Social Dilemma (NSD) (Aaldering et al., In prep; Halevy et al., 2012; Polzer et al., 1999; Wit & Kerr, 2002), individuals face decision options: 1) to keep the endowments to themselves, to contribute 2) solely to the in-group (benign parochial cooperation) and/or 3) equally to the collective (universal cooperation). As such, in NSD the definition of
parochial cooperation does not require hostility towards the out-group (or out-group hate) since the situation allows individuals to show preference towards the in-group while not hurting the out-group (i.e. benign parochial cooperation). Nonetheless, some researchers have argued that inaction/avoidance to help the out-group when having an opportunity to do so can be viewed as a form of out-group hate, since it indirectly withholds benefits from the out-group (Weisel, 2015; Weisel & Böhm, 2015). Thus, in this way, the NSD could not entirely unveil the motivation behind parochial cooperation.
Next to the NSD is the Intergroup Prisoner‘s Dilemma (IPD) game (e.g. Bornstein, 2003). In this game, individuals are also divided into two groups, though a nested structure within a collective whole is not explicit. Individuals also received endowments and have to decide between two options: 1) to pursue self-interest and/or 2) to contribute to the in-group while directly hurting the out-group. In this case, parochial cooperation may only be viewed as an inclination to favor the in-group as well as to derogate the out-group (i.e. competitive parochial cooperation). However, in this game, the motivation to help the in-group is indistinguishable from the desire to hurt the out-group. Also, the preference for universal cooperation might be disguised in selfish behavior, since individuals may simply not want to favor the in-group at the expense of the out-group and, in consequence, choose to pursue self-interest instead (Halevy, Bornstein, & Sagiv, 2008). Therefore, the underlying motivation of parochial cooperation in IPD is ambiguous.
In comparison, in the Intergroup Prisoner‘s Dilemma—Maximizing Difference (IPD-MD) game (Halevy et al., 2008) individuals are confronted with three options 1) to pursue self-interest, 2) to contribute to the in-group while directly hurting the out-group (competitive parochial cooperation), and 3) to benefit the in-group while not hurting the out-group (benign
parochial cooperation). Thus, in this version, the two contrasting scopes of parochial cooperation (i.e. benign and competitive) can be clearly distinguished. However, IPD-MD cannot distinguish between individuals‘ preference for parochial cooperation and universal cooperation, as one does not have an option to benefit both the in-group and the out-group. Therefore, all the three games previously discussed cannot completely encapsulate all possible cooperative options and their underlying motives in intergroup situations.
Recently, Aaldering et al. (In prep) designed the Collective Incentive (CI) game to comprehensively model cooperative behaviors in intergroup situations. In this game, participants can choose between keeping an endowment to themselves and investing it in three pools that correspond to different cooperation options—including benign parochial cooperation,
competitive parochial cooperation, and universal cooperation. That is, individuals can choose between: 1) pursuing their self-interest, 2) benefitting the in-group while not directly hurting the out-group (benign parochial cooperation), 3) contributing to the in-group while directly hurting the out-group (competitive parochial cooperation), and 4) investing in the collective that benefits both the in-group and out-group (universal cooperation) (See Table 1). In this game, individuals personally always benefit the most when keeping endowments to themselves but are better off when all in-group members contribute to the in-group. However, the game is designed to make the possible maximum outcomes achievable at the universal level if everyone from both the in-group and the out-in-group contributes to the collective more than to the in-in-group. In comparison to other games previously mentioned, CI provides an explicit option for universal cooperation and clearly separates the pools for benign parochialism and competitive parochialism. In other words, the game design can disentangle subtle differences between benign parochial cooperation,
examination of individuals‘ cooperative decisions and allows for a more comprehensive interpretation of parochial cooperation.
For example, Aaldering et al. (In prep) conducted a series of experiments investigating situational and personality (specifically, social value orientation) factors that drive individuals to different cooperative actions, especially parochial cooperation. In one experiment using CI the authors found that individuals engage more in benign parochial cooperation in comparison to competitive parochial cooperation and universal cooperation. Hence, parochial cooperation seems to be driven mainly by the motivation to contribute to the in-group, relative to the desire to hurt the out-group (for the same argument, see also Balliet et al., 2014; De Dreu, 2010;
Thielmann & Böhm, 2016). This cooperative pattern appeared particularly strong in pro-social individuals (Aaldering et al., In prep; Thielmann & Böhm, 2016), which suggests that individual differences in pro-social inclination also play a role in cooperative behaviors.
Social Value Orientation and Parochial Cooperation
While the theoretical accounts previously discussed (i.e. SIT and BGR) provide a possible explanation for the general underlying mechanism of parochial cooperation, none actually considers internal personality factors that may fundamentally influence how individuals behave differently in social contexts. In the domain of interdependent decision behaviors,
especially cooperation in social dilemmas, one fundamental psychological construct that has long been studied is social value orientation (SVO), which determines an individual‘s disposition to value not only personal welfare but also cooperation and welfare of others in a situation of interdependence (Balliet, Parks, & Joireman, 2009; Bogaert, Boone, & Declerck, 2008; Murphy et al., 2011; Van Lange, 1999). In other words, SVO reflects motivational differences in how individuals weigh between outcomes for themselves and for others, which consequently
influences their cooperative behaviors (see Balliet et al., 2009; Bogaert et al., 2008; De Dreu, 2010).
Most studies on the role of SVO on pro-social behaviors have been done by categorizing people into two main groups: pro-social and pro-self (Balliet et al., 2009; Van Lange, 1999)— e.g., using the pioneering decomposed game which involves consecutive decisions for
distributing hypothetical points between oneself and others (see Messick & McClintock, 1968). Previous studies have consistently shown that pro-socials tend to cooperate with and care for others, whereas pro-selves only value their own outcomes and tend to be non-cooperative in order to gain personal benefits (Balliet et al., 2009; Bogaert et al., 2008; Messick & McClintock, 1968; Van Lange, 1999). However, some researchers have argued that since SVO indicates the quantity of how much individuals are willing to contribute or sacrifice for the sake of others, it is actually a continuous construct (e.g. Murphy et al., 2011). In other words, as SVO quantifies individual differences in interdependent utilities, it is best represented in a continuous scale. A categorical approach, hence, leads to a loss of full explanatory power of SVO (see also Cohen, 1983), as a considerable amount of information about individuals‘ SVO, e.g. the extent to which individuals are pro-social or pro-self, is ignored and discarded. To resolve and improve upon this problem, recently, Murphy and colleagues (2011) have developed a novel measurement (called ‗the slider measure‘) that computes SVO as a continuous, rather than categorical, spectrum of pro-sociality. Since the slider measure provides continuous information on individuals‘ SVO, there is no need to categorize participants as pro-social or pro-self, preventing information loss due to categorization. Consequently, no individual is excluded as unclassifiable, and a wider range of social preferences can be taken into account. Thus, using the slider measure SVO can be
assessed as a level of pro-sociality, providing a more sensitive and higher resolution evaluation of SVO.
Research on the relationship between SVO and cooperation in social dilemmas has shown that in general pro-socials demonstrate more willingness to cooperate and greater concern for the larger collective than pro-selves (see Bogaert et al., 2008). For example, Simson (2004) found that pro-socials readily showed a high level of cooperation even in anonymous situations in a one-shot game, while pro-selves did not. Pro-sociality was also found to be associated with higher reciprocity and tendencies to value joint outcomes and equality in the outcomes (Van Lange, 1999). Additionally, pro-socials contributed more to the collective when asked about their willingness to contribute to the establishment of a public good, as compared to pro-selves (De Cremer & Van Vugt, 1999; De Cremer & Van Dijk, 2002).
In an intergroup setting, however, prior findings on the relationship between SVO and cooperation are less conclusive. Specifically, previous research provides mixed results whether cooperation in pro-socials is parochial or universal in nature. On the one hand, since, by
definition, pro-socials in general value both their own and others‘ welfare (Balliet, Parks, & Joireman, 2009; Bogaert, Boone, & Declerck, 2008; Murphy et al., 2011; Van Lange, 1999), pro-socials‘ cooperative tendency should be universal—i.e. irrespective of group membership. This argument was supported by Thielmann and Böhm (2016). The authors compared cooperative behaviors in three different versions of the IPD game in relation to SVO using the slider
measure. They found that individuals high in pro-sociality displayed more universal cooperation relative to parochial cooperation as their cooperation was not restricted to their in-group but extended to the out-group as well. Also, especially highly pro-social individuals preferred universal cooperation over competitive parochial cooperation, demonstrating a reluctance to
impose out-group harm among highly pro-social individuals. This implies that cooperative tendencies of individuals with high pro-social inclination can extend towards the out-group and may be universal in nature.
On the other hand, it could be argued from the evolutionary perspective that, because humans are social animals, parochialism and pro-sociality provide an adaptive function, increasing individual fitness within a group and a probability of group survival in time of intergroup conflicts (Aaldering, 2014; Rusch, 2014). Because of this, pro-social and parochial tendencies might have concurrently evolved (Rusch, 2014). Therefore, pro-socials‘ cooperative tendency should be parochial. That is, pro-social behavior is in-group bounded in nature
(Aaldering, 2014; De Dreu, 2010; De Dreu et al., 2014; Weisel & Böhm, 2015). For example, using the standard NSD, Polzer (2004) found pro-socials, compared to pro-selves, to engage more in parochial cooperation, relative to universal cooperation. Aaldering et al. (In prep) also found that pro-socials are (benign rather than competitive) parochial cooperators. However, when parochial cooperation does harm the out-group, pro-social individuals turn to universal cooperation. The results seem to suggest that the cooperativeness of pro-socials is benign, rather than competitive, parochial in nature. That is, pro-socials‘ cooperation is in-group bounded and is motivated by a tendency to favor their in-group, rather than hurting the out-group (e.g. Aaldering et al., In prep; Balliet et al., 2014; Brewer, 1999; De Dreu, 2010; De Dreu et al., 2014).
Nonetheless, when an option for universal cooperation is not available, previous studies have consistently shown that SVO may play a role in moderating individuals‘ tendency towards parochial cooperation (Abbink, Brandts, Herrmann, & Orzen, 2012; De Dreu, 2010; De Dreu et al., 2015). For example, De Dreu (2010) observed more benign (relative to competitive)
parochial cooperation among pro-socials compared to pro-selves in the IPD-MD game. Moreover, De Dreu et al. (2015) also used the IPD-MD game and replicated the previous
findings that pro-socials engaged in parochial cooperation more than pro-selves, and furthermore after lowering impulse control pro-socials increased their contribution towards parochial
cooperation. In a more competitive setting of an intergroup conflict game, pro-socials displayed competitive parochial cooperation, at the expense of the out-group, more than pro-selves in order for their group to win an intergroup rivalry (Abbink et al., 2012). Combined, these findings seem to suggest that, depending on the game and the setting, (benign as well as competitive) parochial cooperation is especially prevalent in pro-socials. Though, importantly, in a setting where competition is not explicit, pro-socials still prefer benign relative to competitive parochial cooperation.
Despite their initial inclination to cooperate, pro-socials‘ cooperation might be susceptible to external, contextual influences, e.g. reciprocity cues and induced group identification (see Bogaert et al., 2008). Previous studies have shown that pro-socials have a stronger tendency and expectation towards reciprocity than pro-selves such that pro-socials expect others to behave cooperatively, whereas pro-selves expect others to similarly act based on their self-interest (Iedema & Poppe, 1994a, 1994b; Kelley & Stahelski, 1970). Also, pro-socials are likely to reciprocally cooperate with their cooperative counterparts, but will react non-cooperatively towards non-cooperative counterparts (e.g. Kanagaretnam, Mestelman, Nainar, & Shehata, 2009; Kelley & Stahelski, 1970; Parks & Rumble, 2001; Van Lange, 1992). These findings seem to suggest that individuals with high, compared to low, pro-sociality would be more influenced by reputation concern as a result of higher expectation for indirect reciprocity. Put differently, pro-socials might be more concerned about their reputation as they tend to
cooperate and highly expect others to reciprocate. In addition to their high expectation of reciprocity, when group identification is induced and made salient, pro-socials also have a tendency to identify more strongly with their in-group, compared to pro-selves (De Cremer & Van Vugt, 1999; De Cremer & Van Dijk, 2002). The fact that individuals high in pro-sociality are sensitive to these external influences, i.e. in-group identification and reputation concern, might in turn impact their parochial cooperation according to the predictions of SIT and BGR, respectively. That is, especially pro-socials might engage more/less in parochial cooperation depending on the strength of in-group identification (SIT) and/or the level of reputation concern (BGR).
To sum up, existing research on the relationship between cooperation and SVO suggest that pro-socials generally tend to cooperate more than pro-selves. In the intergroup setting, highly pro-social individuals are both more parochial and universal in their cooperation than low pro-social individuals. Furthermore, highly pro-social individuals prefer benign parochial
cooperation to competitive parochial cooperation, and would turn to universal cooperation when their parochial cooperation directly hurts other parties, i.e. the out-group. Highly pro-social individuals are also more likely to be affected by external influence than low pro-social individuals, which may in consequence lead to changes in their parochial cooperation. In two experiments, we investigate how social value orientations together with psychological motives based on SIT and BGR (i.e. in-group identification or reputation concern) may impact
individuals‘ cooperation.
Experiment 1
The first experiment aims to reaffirm established findings and investigate the unresolved questions outlined earlier. First, in intergroup situations, is cooperation generally benign
parochial, competitive parochial or universal? Second, what is the psychological mechanism that drives parochial cooperation? Specifically, is parochial cooperation better predicted by
reputation concern (following BGR) or by in-group identification (following SIT)? Third, what is the relationship between pro-sociality and cooperation? Specifically, are pro-socials especially benign parochial, competitive parochial or universal cooperators? Finally, since individuals with high pro-sociality are more influenced by external influences, are they especially affected by these motives (i.e. reputation concern or in-group identification)? That is, what is the interaction between individuals‘ SVO and the underlying psychological motives on parochial cooperation? In this experiment, we investigated these questions by using the Collective Incentive game, to capture different cooperative decisions, and the slider measure, to measure individuals‘ SVO, and by manipulating in-group identification and reputation concern.
Accordingly, we put forward and test the following hypotheses. First, we expect a difference between cooperative tendencies—i.e. benign parochialism, competitive parochialism, and universal cooperation. Specifically, in accordance with previous studies (e.g. Aaldering et al., In prep), we expect a preference for parochial cooperation relative to universal cooperation. In other words, individuals in general will engage more in parochial cooperation compared to universal cooperation (Hypothesis 1A). That is, investments in parochial cooperation (combining both benign and competitive parochialism) will be higher than investments in universal
cooperation. Furthermore, since a competition in the Collective Incentive game is not explicit, we expect parochial cooperation to be benign rather than competitive (Balliet et al., 2014; De Dreu, 2010; De Dreu et al., 2015). That is, individuals would prefer benign parochial
cooperation over competitive parochial cooperation (Hypothesis 1B). In other words, investments in the benign parochial pool will be higher than the competitive parochial pool.
Additionally, previous research demonstrates that when parochialism is hostile, individuals turn to universal cooperation (e.g. Aaldering et al., In prep; Thielmann & Böhm, 2016). Thus, we also expect more investments in the universal pool than in the competitive parochial pool (Hypothesis 1C).
Second, based on the corresponding predictions from the SIT and BGR, there are two competing hypotheses for which mechanism—in-group identification or reputation concern— drives parochial cooperation. We test this by decreasing the strength of in-group identification and the degree of reputation concern. Because in previous studies (e.g. Aaldering, 2014; Aaldering et al., In prep), parochial cooperation is present without any manipulation, which means parochial cooperation already emerges even in a default or control situation. This suggests that by decreasing either the strength of in-group identification or the degree of reputation
concern, parochial cooperation should decrease, depending on which mechanism actually drives parochial cooperation. Also, because of the uncompetitive setting of the Collective Incentive game, we expect the influence of these driving mechanisms especially on benign parochial cooperation. According to SIT, in-group identification is the key driving mechanism for (benign) parochial cooperation (Tajfel & Turner, 1979). Therefore, in line with SIT, we expect in-group identification to drive (benign) parochial cooperation (Hypothesis 2A). Specifically, lowering the degree to which individuals identify with their in-group should decrease parochial cooperation, compared to a control condition. That is, investments in the benign parochial pools in the low in-group identification condition will be lower than in the control condition. Alternatively,
according to BGR, individuals behave parochially to maintain their in-group reputation as they expect in-group members to reciprocally cooperate (Yamagishi et al., 1999). In line with BGR, we expect reputation concern to drive (benign) parochial cooperation (Hypothesis 2B).
Specifically, decreased reputation concern would decrease parochial cooperation, compared to the control condition. That is, investments in the benign parochial pool in the low reputation concern condition would be lower than in the control condition.
Third, we expect a relation between cooperative behaviors and pro-sociality, especially among individuals high in pro-sociality. On the one hand, since pro-sociality means concern for the welfare of others (e.g. Balliet et al., 2009; Bogaert et al., 2008; Murphy et al., 2011; Van Lange, 1999), which could extend beyond in-group members (Thielmann & Böhm, 2016), a cooperative tendency of individuals high in pro-sociality should be universal (Hypothesis 3A). That is, the higher the pro-sociality, the more investments in universal cooperation. On the other hand, previous studies in a less competitive setting (in our case, the Collective Incentive game) found benign, rather than competitive, parochial cooperation to be more pronounced in highly pro-social individuals (e.g. Aaldering et al., In prep), which is in line with the evolutionary perspective that parochialism and pro-sociality might have concurrently evolved (Rusch, 2014). Hence, a cooperative tendency of individuals high in pro-sociality might be benign parochial (Hypothesis 3B). That is, the higher the pro-sociality, the more investments in benign parochial cooperation.
Finally, since highly pro-social individuals are more affected by external influences than low pro-social individuals (e.g. Bogaert et al., 2008), we expect an interaction between the manipulated conditions (low in-group identification and/or low reputation concern) and SVO on benign parochial cooperation. Specifically, the difference between conditions would be
pronounced especially in individuals with high pro-sociality. Such a difference in parochial cooperation would not be observed among individuals low in pro-sociality.
On the one hand, within SIT, parochial cooperation is driven by in-group identification, in order to create favorable comparisons between the in-group and the out-group (Tajfel & Turner, 1979). As individuals high in pro-sociality tend to identify strongly with the in-group (De Cremer & Van Vugt, 1999; De Cremer & Van Dijk, 2002), this would predict that especially among individuals high in pro-sociality parochial cooperation would be more influenced by in-group identification (Hypothesis 4A). That is, especially individuals higher in pro-sociality would show less benign parochial cooperation in the low in-group identification compared to the low reputation concern and control conditions. There would be no difference among individuals low in pro-sociality.
On the other hand, according to BGR parochial cooperation is driven by reputation concern as a result of an expectation of (indirect) reciprocity (Yamagishi et al., 1999). As
individuals high in pro-sociality tend to have a strong regard towards reciprocity, and hence high reputation concern (Kanagaretnam et al., 2009; Kelley & Stahelski, 1970; Parks & Rumble, 2001; Van Lange, 1992), their parochial cooperation would be influenced more by reputation concern (Hypothesis 4B). That is, individuals higher in pro-sociality would show less benign parochial cooperation in the low reputation concern compared to the low in-group identification and control conditions. Such difference would not be expected among individuals low in pro-sociality.
Methods
Participants and design. 214 participants participated in the experiment at University of Amsterdam. Sample size was determined based on previous experiments with a similar design (Aaldering et al., In prep). We excluded ten participants from the data analysis due to a technical error; participants‘ SVO scores were not recorded. The total sample was N = 204, 72.5% female,
mean age = 22.14, SD = 3.43. Participants were randomly assigned to low reputation concern, low in-group identification, or control condition. Social value orientation was measured as a between-subjects continuous factor (see below). The order of the tasks (i.e. whether participants started with the Collective Incentive game or the SVO slider measure) was counterbalanced across all participants, and was treated as a covariate. Main dependent variables were the
investments in different pools (i.e. the benign parochial pool, the competitive parochial pool, and the collective pool) of the Collective Incentive game. All subjects gave informed consent prior to the experiment, and after completing the experiment all received the participation fee of €10 or course credit. The experiment was approved by the Ethics Committee of Psychology Research Institute of the University of Amsterdam (2016-WOP-7520).
Procedure and manipulation. The experiment was conducted in a laboratory with eight separate cubicles at University of Amsterdam. Participants arrived in the experiment room, were assigned to one of the cubicles, and provided written informed consent. The experiment was computerized and self-paced. During the experiment, participants played the Collective Incentive game and completed the slider measure. The order of each task was counter-balanced for all participants. That is, at the beginning of the experiment half of participants completed the slider measure first; the other half first played the Collective Incentive game first. Each participant was randomly assigned to one of the experimental conditions (i.e. low reputation concern, low in-group identification, or control) and received corresponding task instructions on the computer screen.
Low in-group identification was manipulated by informing participants that they would be categorized in a group based on their answers on (bogus) questionnaires—e.g. ―I generally have a strong opinion‖ and ―Poor people who are convicted for theft need to get a sentence
discount‖ (for similar manipulations based on bogus group assignment see e.g. Gaertner & Insko, 2000; Stroebe et al., 2005). After participants answered the questionnaire, they were told that no group members had a similar profile (i.e. having similar answers) and that they were assigned to an already-existing group, whose other members had similar profiles, to fill in a gap. This would indicate that participants had nothing in common with their in-group members, and thus should lower their in-group identification.
In the low reputation concern condition, the manipulation was performed by informing participants that other in-group members would not see their decisions in the game. In other words, both group membership and decisions of each participant will be kept anonymous (unknown to other participants) while participants themselves are still aware of a group which they belong to (e.g. Kiyonari, 2002; Romano et al., 2017; Yamagishi et al., 1999). Hence, this should not affect in-group identification and only lower reputation concern since participants make contributions anonymously and therefore cannot (to a great extent) expect reciprocation from the in-group members. In the control condition, similar to the study of Aaldering et al. (In prep), the instruction does not mention any of the manipulations—i.e. the anonymity of choices and the bogus assignment of groups.
In all conditions, participants received hypothetical endowments of 10 Euros and learned that there are two groups of four members in this game (Team Triangle and Team Circle) and that he/she and three others belong to the ‗Team Triangle‘. Participants also learned that their final outcomes are interdependent on the decisions made by the members of both their and other teams. Participants then played the Collective Incentive game for five rounds to ensure
reliability. After the game, participants completed questions/items assessing their understanding of task structure and manipulation instruction checks to assess the in-group identification and
reputation concern conditions (see below). They also completed measurements for the degree of in-group identification and the level of reputation concern, see under ‗measures‘.1 Finally, after the experiment, participants received either €10 or course credits.
Materials
Social value orientation. To measure participants‘ SVO, we used the more recently developed slider measure for a sensitive and high resolution measure of individuals‘ social preference (see Murphy et al., 2011). This measure gives the resulting SVO as a continuous angle of pro-sociality, which does not allow for participants to be unclassifiable. The SVO measure consists of six decisions about how to allocate hypothetical money between oneself and others. Each decision presents a slider with nine possibilities of a variety of outcomes (e.g. for self and other respectively: 50-100; 54-89; 59-79; 63-68; 68-58; 72-47; 76-36; 81-26; 85-15). By calculating the mean allocations to oneself (As) and others (Ao), an individuals‘ SVO score is
computed as the arctangent of a ratio of Ao – 50 and As – 50, and is represented as an angle of
pro-sociality. A zero degree angle indicates perfect self-interest; a more positive angle reflects higher pro-sociality. For example, categorically, individuals with a SVO score higher than 22.45⁰ are considered pro-socials, while individuals with a SVO score less than 22.45⁰ are considered pro-selves. However, here, SVO was treated as a continuous variable.
Collective incentive game. To play the Collective Incentive game, participants received 10 Euros and decided how much money to keep to themselves or to invest in three different pools—i.e. the benign parochial pool (Pool A), the competitive parochial pool (Pool B), and the universal pool (Pool C) (see Table 1). Specifically, investing one euro in the benign parochial pool would yield a 0.5 euro payoff to all in-group members. One euro invested in the competitive parochial pool would yield 0.5 euro to all in-group members but also subtract 0.5 euro from all
out-group members. One euro invested in the universal pool would yield 0.40 euro for all
members of both groups. Lastly, the final option was not to invest in any of the pools. Hence, the allocation decisions of all partaking players interdependently determine the final outcome. When all members of each group contribute 1 euro, the maximum total outcome at the individual level is 2 vs. 3.2 for the benign parochial pool and the universal pool, respectively. In contrast, if all members of each group contribute 1 euro to the competitive parochial pool, the final outcome for each player received would be 0 (they would cancel each other out). All participants made 5 rounds of decisions. Investments in benign parochial cooperation, competitive parochial
cooperation, and universal cooperation were computed as means across the 5 rounds. For Pool A (benign parochial cooperation) Cronbach‘s alpha was .91. For Pool B (competitive parochial cooperation) Cronbach‘s alpha was .94. For Pool C (universal cooperation) Cronbach‘s alpha was .96. Finally, for the last option of keeping endowments to themselves Cronbach‘s alpha was .94.
Understanding of task and manipulation instructions. To check participants‘
understanding of the Collective Incentive game, we asked 6 questions about the consequence of investing in different pools: 1) What is a reason for your group to contribute to pool A (Benign parochial cooperation)? 2) What is a reason for your group to contribute to pool B (Competitive parochial cooperation)? 3) What is a reason for your group to contribute to pool C (Universal cooperation)? 4) What is a reason for members of the other group to contribute to pool A (Benign parochial cooperation)? 5) What is a reason for members of the other group to
contribute to pool B (Competitive parochial cooperation)? 6) What is a reason for members of the other group to contribute to pool C (Universal cooperation)? The order of questions was randomized. For all questions, the possible responses are: 1) to generate profit for my group, 2)
to decrease gain from the other group, 3) to generate profit for both groups, and 4) both 1 and 2. The correct answers for each set of three questions (questions 1-3 and 4-6) are 1, 4, and 3,
respectively. If participants correctly understand the game, participants should answer ‗4‘ to item 2 more often than item 1, and also answer ‗3‘ should be indicated more often when comparing item 3 to item 2.
The manipulation check of in-group identification instruction was assessed with three items (1 = ―completely disagree‖ and 7 = ―completely agree‖): 1) ―The group allocation was random‖ (reverse coded), 2) ―I would be allocated to a group with people that are similar to me‖, and 3) ―My answering profile would determine to which group I would be allocated‖.
Cronbach‘s alpha was 0.72.
The manipulation check of reputation concern instruction was also assessed with three items: (1 = ―completely disagree‖ and 7 = ―completely agree‖): 1) ―My group members of team Circle will learn how much I contributed‖, 2) ―I expected that the members of my group, Team Circle, would know how much I contributed‖, and 3) ―I don‘t think that the members of my group, Team Circle, would get to know how much I contributed‖ (reverse coded). Cronbach‘s alpha was .56. Deleting any of the items did not improve the reliability.
Measurement of driving mechanisms. In addition to the manipulation check items which only examine whether participants understood the task instructions, we also measured the underlying mechanisms (i.e. the level of reputation concern and the strength of in-group
identification), to explore these as possible mediating variables. The level of reputation concern was assessed by measuring four items (1 = ―completely disagree‖ and 7 = ―completely agree‖) adapted from Wu, Balliet, and Van Lange (2015): 1) ―During the decision making task, I thought about how my other group members would think about me‖, 2) ―It‘s important to me
that others have a positive evaluation about me‖, 3) ―It‘s important that others will accept me, and 4) ―The fact that my other group members knew nothing about my decisions played an important role when I made my investments in the decision making task‖. Surprisingly, Cronbach‘s alpha was .55. To improve reliability we consequently removed the last item from this scale and the final Cronbach‘s alpha was .72.
The degree of in-group identification was measured using 11 items from Leach et al. (2008). Specifically, the items measured group-level self-investment (Solidarity and Satisfaction) and group-level self-definition (Individual Self-Stereotyping and In-group Homogeneity),
excluding Centrality items. For example, ―I feel a bond with group]‖, ―I am glad to be [In-group]‖, and ―I have a lot in common with the average [In-group] person‖) (1 = ―completely disagree‖ and 7 = ―completely agree‖). Cronbach‘s alpha was .94.
Results
Descriptive statistics. Correlations between main measurements are shown in Table 2. Investments in each of the pools were weakly negatively correlated. Pro-sociality was positively associated with investments in universal cooperation, but was negatively related with the selfish choice, which was to keep endowments to oneself. Surprisingly, the degree of reputation concern (3 items from Wu et al., 2015) did not correlate with investments in any pool, including benign and competitive parochial cooperation, providing no preliminary support for the Bounded Generalized Reciprocity theory. However, in-group identification (11 items from Leach et al., 2008) was positively related with investments in both benign and competitive parochial cooperation, but negatively correlated with keeping endowments to oneself. This suggests that the degree to which individuals identify with their group influences their decisions to parochially
cooperate, in line with Social Identity Theory. In-group identification was also positively correlated with reputation concern.
Task understanding and manipulation instruction checks. To ensure participants‘ understanding of the game, the result from a chi-square test suggests that when comparing answers to item 2 (i.e. What is a reason for your group to contribute to Pool B (Competitive parochial cooperation)?) to item 1 (i.e. What is a reason for your group to contribute to Pool A (Benign parochial cooperation)?), participants indicated the answer 4 (i.e. both to generate profit for my group and to decrease gain from the other group) more for item 2 than item 1, χ2 (9, N = 204) = 23.90, p = .004. Also, when comparing item 3 (i.e. What is a reason for your group to contribute to Pool C (Universal cooperation)?) and item 1, the answer 3 (i.e. to generate profits for both groups) was chosen more for item 3 than item 1, χ2 (9, N = 204) = 19.68, p = .02. Thus, participants indeed understood the game correctly.
To check whether the manipulations of in-group identification and reputation concern had worked, i.e. whether participants correctly understood the instructions, a MANCOVA with conditions (i.e. low in-group identification, low reputation concern, and control) as between-subject variables and task order of SVO measures (assessed before or after the decision making task) as covariate was conducted on the mean of the three identification manipulation check items and the mean of the three reputation concern manipulation check items as dependent variables. If the manipulation of group identification had worked, the scores in the low in-group identification manipulation should be higher than the other two conditions (i.e. low reputation concern and control conditions). Also, if the reputation concern manipulation had worked, the scores in the low reputation concern should be lower than the other two conditions. The results revealed a multivariate main effect of conditions, Pillai‘s Trace = .38, F (4, 400) =
23.45, p < .001, ηp2 = .19. There was a main effect of conditions on the manipulation check of
in-group identification, F (2, 200) = 56.07, p < .001, ηp2 = .36. Indeed, the scores of the in-group
identification manipulation check items in the low in-group identification condition were higher (M = 4.14, SD = 1.41) than in the low reputation concern condition (M = 2.19, SD = 1.09, Contrast Estimate = 1.95, SE = .21, p < .001, 95% CI [1.44, 2.45], Bonferroni corrected) and the control condition (M = 2.23, SD = 1.52, Contrast Estimate = 1.90, SE = .21, p < .001, 95% CI [1.39, 2.41], Bonferroni corrected). The in-group identification scores in the low reputation concern and control conditions did not differ (Contrast Estimate = -.05, SE = .21, p = 1.00, 95% CI [-.56, .46], Bonferroni corrected). This illustrates that participants correctly understood the manipulation of in-group identification.
However, to our surprise, there was no main effect of conditions on the scores of the reputation concern manipulation check items, F (2, 200) = .18, p = .84, ηp2 = .002. Neither was
there a difference between the low reputation concern condition (M = 2.52, SD = 2.20) and the low in-group (M = 2.51, SD = 1.16, Contrast Estimate = .01, SE = .22, p = 1.00, 95% CI [-.51, .53], Bonferroni corrected), nor the control condition (M = 2.63, SD = 1.41, Contrast Estimate = -.11, SE = .22, p = 1.00, 95% CI [-.63, .42], Bonferroni corrected). This indicates that participants did not understand the reputation concern manipulation.
Hypothesis testing. To test Hypothesis 1A that individuals engage more in parochial cooperation than universal cooperation, we combined investments from benign and competitive parochial cooperation pools into a new variable ―parochial cooperation‖. A paired samples t-test revealed more parochial cooperation (M = 3.94, SD = 2.70) than universal cooperation (M = 2.26, SD = 2.57), t (203) = 5.40, p < 0.001. Thus, Hypothesis 1A was supported. Similarly, two paired samples t-tests were conducted to test Hypothesis 1B that parochial cooperation is benign
rather than competitive and Hypothesis 1C that when cooperation hurts the out-group universal cooperation is preferred over competitive parochial cooperation. The results showed that
investments in benign parochial cooperation (M = 2.67, SD = 2.16) were higher than investments in competitive parochial cooperation (M = 1.26, SD = 1.96), t (203) = 6.46, p < 0.001, supporting Hypothesis 1B. Investments in universal cooperation (M = 2.26, SD = 2.57) were also higher compared to competitive parochial cooperation, t (203) = 3.90, p < 0.001, which supports
Hypothesis 1C. The Bonferroni correction was applied to the resulting p-values for all three tests. These results suggest that, as expected, individuals in general have a preference for parochial cooperation relative to universal cooperation, and the underlying motive for parochial
cooperation is benign rather than competitive. Furthermore, individuals in general are more likely to engage in benign parochial cooperation and universal cooperation over competitive parochial cooperation.
To test the predictions from SIT and BGR on the underlying mechanism of parochial cooperation and its relation with social value orientation (H2, H3 and H4), a MANCOVA with three conditions (i.e. low in-group identification, low reputation concern, and control) as between-subject factors, social value orientation as continuous between-subjects variable and task order as covariate was conducted on investments in the three pools (i.e. benign parochial cooperation, competitive parochial cooperation, and universal cooperation) to examine the main and interaction effects of conditions and social value orientation. Examining Hypothesis 2 on the underlying mechanisms, surprisingly, there was no main effect of conditions on benign parochial cooperation, F (2, 197) = 1.95, p = .15, ηp2 = .02. Pairwise comparisons under Bonferroni
adjustment showed no significant difference in benign parochial cooperation between the low in-group identification condition (M = 2.33, SD = 1.86) and the control condition (M = 3.04, SD =
2.44), Contrast Estimate = -.73, SE = .37, p = .15, 95% CI [-1.62, .16]. Investments in benign parochial cooperation were also not different when comparing the low reputation concern
condition (M = 2.65, SD = 2.10) with the control condition, Contrast Estimate = -.42, SE = .37, p = .77, 95% CI [-1.31, .47], as well as the low in-group identification condition, Contrast Estimate = .31, SE = .37, p = 1.00, 95% CI [-.58, 1.20]. In other words, we could not conclude which underlying mechanism (in-group identification or reputation concern) actually drives benign parochial cooperation. Thus, we did not find any support for Hypothesis 2. Additionally, no main effect of conditions was found on either competitive parochial cooperation, F (2, 197) = .03, p = .97, ηp2 = .00, or on universal cooperation, F (2, 197) = .70, p = .50, ηp2 = .00.
Testing Hypothesis 3 whether highly pro-social individuals are benign parochial or universal cooperators in nature, there was a multivariate main effect of social value orientation on a set of three pool investments, Pillai‘s Trace = .07, F (3, 195) = 4.99, p = .002, ηp2 = .07. In
line with Hypothesis 3A, pro-social value orientation was associated with investments in universal cooperation, F (1, 197) = 11.02, p = .001, ηp2 = .05. Parameter estimates suggest that
investments in universal cooperation predicted individuals‘ social value orientation, B = .71, SE = .33, t (203) = 2.18, p = .03, ηp2 = .02, supporting Hypothesis 3A that the cooperative tendency
of individuals with high pro-sociality is universal in nature. The effect, however, was not found for either investments in benign parochial cooperation, F (1, 197) = .68, p = .41, ηp2 = .00, or investments in competitive parochial cooperation, F (1, 197) = .27, p = .61, ηp2 = .00. Thus,
Hypothesis 3B that that the cooperative tendency of individuals with high pro-sociality is parochial was not supported.
Finally, testing Hypothesis 4—that individuals high in pro-sociality are influenced more by psychological mechanisms—showed no interaction effect between conditions and social
value orientation on benign parochial cooperation, F (2, 197) = .25, p = .78, ηp2 = .00, providing
no support for Hypothesis 4. The interaction effect was also not found on competitive parochial cooperation, F (2, 197) = 1.08, p = .34, ηp2 = .01, and universal cooperation, F (2, 197) = .39, p =
.68, ηp2 = .00. To reaffirm that there was indeed no interaction effect between the specific
conditions either, we computed the hypothesized contrasts for low reputation concern (1) vs. low in-group identification (-1) conditions, low reputation concern (1) vs. control (-1) conditions, and low in-group identification (1) vs. control (-1) conditions. The interactions were subsequently formed by multiplying the standardized SVO angle with each of these contrasts. Hierarchical regression analyses were then performed on investments in benign parochial cooperation with the SVO angle and the contrast variable in model 1, and the interaction in model 2 as
independent variables. The results showed that the comparison between the low reputation concern and low in-group identification conditions did not yield a significant interaction, B = .15, SE = .19, β = .06, t (203) = .80, p = .43, Adj. R2 = -.004. There was also no significant interaction when comparing the low reputation concern condition with the control condition, B = .02, SE = .18, β = -.01, t (203) = -.08, p = .93, Adj. R2 = -.007. Furthermore, no interaction was found when comparing the low in-group identification condition with the control condition, B = .15, SE = .21, β = -.05, t (203) = -.70, p = .49, Adj. R2 = -.001. Thus, overall, we did not find any support for Hypothesis 4.
Unexpectedly, we did obtain a multivariate main effect of the covariate, i.e. the order of the SVO measure, Pillai‘s Trace = .04, F (3, 195) = 2.84, p = .04, ηp2 = .04. Investments in
benign parochial cooperation was higher when the slider measure was completed first (M = 2.99, SD = .19) than when the slider measure was completed after the decision task (M = 2.28, SD = .23), F (1, 197) = 5.36, p = .02, ηp2 = .03. No other main or interaction effects were significant.
Auxiliary analyses.
Driving mechanisms. Given the absence of the main effect of conditions, we also investigated whether participants actually experienced lower in-group identification and/or reputation concern in different conditions. A MANCOVA with conditions (i.e. low in-group identification, low reputation concern, and control) as between-subject variables and task order of SVO measures (assessed before or after the decision making task) as a covariate was also conducted on the mean scores of 3 items from Wu et al. (2015) reflecting reputation concern and the mean scores of 11 in-group identification items from Leach et al. (2008), reflecting in-group identification, as dependent variables. The results yielded no significant multivariate effect of conditions, Pillai‘s Trace = .03, F (4, 400) = 1.73, p = .14, ηp2 = .02. There were no main
univariate effects of conditions on each measure (for in-group identification: F [2, 200] = 2.11, p = .12, ηp2 = .02; and for reputation concern: F [2, 200] = 1.04, p = .35, ηp2 = .01). A multiple
comparisons test under the Bonferroni correction also showed that there was no difference in the degree of in-group identification between the low in-group identification condition (M = 2.92, SD = 1.16) compared to the control condition (M = 3.34, SD = 1.25, Contrast Estimate = -.41, SE = .21, p = .14, 95% CI [-.91, .09], Bonferroni corrected). The low in-group identification
condition also did not differ from the low reputation concern condition (M = 3.21, SD = 1.21, Contrast Estimate = -.30, SE = .21, p = .45, 95% CI [-.79, .20], Bonferroni corrected). This seems to suggest that, despite participants‘ understanding of the in-group identification
manipulation, it did not effectively influence how participants identified with their groups. That is, participants did not experience lower in-group identification from the manipulation. Similarly, a multiple comparisons test also showed that there was no difference in the level of reputation concern between the low reputation concern condition (M = 3.67, SD = 1.34) compared to the
