Large-scale cross-societal examination of real- and minimal-group biases

Research Question 1 (RQ1) asks whether the MGE holds across cultures and so addresses the possibility that a basic tendency towards ingroup preference is a human universal (e.g., Brown, 2004). Correspondingly, Hypothesis 1 (H1) predicts that the MGE is culturally ubiquitous, i.e., that the effect of ingroup bias will be statistically significant and positive in all societies tested (or, at a minimum, that the pattern of observed results is consistent with a universal presence of the MGE coupled with sampling error). However, we anticipate that6even against the backdrop of a culturally ubiquitous MGE, we will nonetheless observe cultural variation.Research Question 2 (RQ2) asks whether the magnitude of the MGE varies as a function of several moderators. Because we conceptualize group biases as culturally situated, where possible, we examine societal-level and individual-level operationalizations of each moderator.We focus on several of the most critical potential moderators identified via past work.First, societies vary in the ease with which people conceive of themselves as able to form and dissolve their memberships in real-world groups. This phenomenon has been referred to as group permeability in some areas of psychology (e.g., Ellemers et al., 1990) and is closely related to the construct of relational mobility (Thomson et al., 2018); we will refer to it as permeability going forward. In societies such as Morocco or Egypt in which the extended family or a small network of close relationships are the central reference groups, permeability and relational mobility tend to be lower because the most critical social groups are enduring and difficult to enter and exit. By contrast, in societies such as the USA or France in which the extended family plays a smaller role and where groups are formed and dissolved more frequently, permeability and relational mobility are high. This factor has been highlighted as an important potential contributor to the MGE going as far back as the early minimal groups work (Tajfel, 1975), but has not been explored extensively in this context (cf., Fischer & Derham, 2016). That said, conceptually related constructs including Hofstede¿s (2001) individualism (which positively correlates with permeability) and the strength of family ties (which negatively correlates with permeability) have been central predictors in many studies of other forms of cross-societal variation (Alesina & Giuliano, 2010; Gorodnichenko & Roland, 2011; Schulz et al., 2019; Thomson et al., 2018; Van de Vliert, 2020).7Given that minimal groups seem to be better analogs for rapidly formed and dissolved groups than stable elements of social life such as the family, Hypothesis 2.1 (H2.1) predicts a positive relationship between permeability and the MGE. In other words, the MGE will be stronger in societies and individuals who show more flexibility moving between new and various social groups. Conversely, in contexts in which long-standing social groups that are difficult to depart from are paramount (e.g., intensive kinship networks), individuals will be less given to affiliate with newly formed transient social identities such as minimal groups. Thus, we predict a positive relationship between permeability/relational mobility and the MGE. Note that this prediction holds only for minimal group preferences; we do not predict this pattern for real-world groups. Indeed, some past theorizing suggests the opposite direction of relationship for some social groups, grounded in the notion that one must be more committed to groups from which one is less able to depart (e.g., Bettencourt et al., 2001; Ellemers, 1993; Hruschka et al., 2014; Tajfel & Turner, 1979).Newly formed social identifications (as appear to be generated by the MGP) entail a degree of risk due to the possibility of forming a link with an untrustworthy partner (Yamagishi et al., 1999) and/or not having recourse available if exploited. The extent to which one is open to such new (and potentially risky) links is plausibly related to societal-level trust and the strength of institutions that enforce social contracts and mediate between individuals engaged in disputes (Gächter et al., 2010). This motivates Hypothesis 2.2 (H2.2): the MGE will be stronger in societies and individuals with a higher baseline level of trustworthiness towards unfamiliar others and that exhibit greater trust in the institutional structures that regulate cooperative behavior. Further support for this possibility comes from evidence that the MGE depends, at least in part, on the expectation that individuals will behave prosocially towards other ingroup8members (Guala et al., 2013; Yamagishi et al., 1999), an important form of ingroup favoritism (Balliet et al., 2014). Importantly, this prediction again holds for the MGE but not necessarily for affiliation with salient real-world groups. Indeed, past work (e.g., Hruschka et al., 2014; Hruschka & Henrich, 2013) provides evidence that in societies with less secure institutions and reduced trust in the rule of law, some forms of ingroup favoritism, including that directed towards the family and the nation, tend to be stronger. This may occur because, in the absence of strong institutions, one is incentivized to favor closer social ties with highly familiar others. On the other hand, it is possible to motivate a competing hypothesis, namely that in societies and institutions high in generalized trust, trust towards outgroup members will be higher and thus the MGE will be lower. Our data will allow us to adjudicate between these possibilities.Our final moderator is self-esteem. There is a long history of debate about the relationship between self-esteem and ingroup favoritism, both within and beyond the context of the MGE. Early work in the social identity tradition conceptualized ingroup favoritism as a response to low self-esteem, but this relationship was not consistently found (for a classic review, see Rubin & Hewstone, 1998). In the context of the MGE, past work suggests a positive association between the MGE and self-esteem (e.g., Cadinu & Rothbart, 1996; Dunham, 2013; Gaertner et al., 2006; Gramzow & Gaertner, 2005), a finding generally interpreted as suggesting that self-related positivity is automatically extended to the ingroup. The most direct cross- cultural inquiry also provides support for this link, finding that Japanese participants show a reduced MGE as compared to North American participants, and that this difference is mediated by self-esteem (Falk et al., 2014). Building on this finding, Hypothesis 2.3 (H2.3) predicts that the MGE will be stronger in individuals with higher self-esteem. While there are many conceptualizations of self-esteem, some, in particular collective self-esteem (Luhtanen &9Crocker, 1992), overlap considerably with notions of group identity. In the present work, we focus on a much simpler conceptualization grounded in the basic question of how positively one evaluates oneself in a relatively neutral context (i.e., personal self-esteem) and the possibility that this evaluation is extended to the minimal ingroup.Research Question 3 (RQ3) asks whether the magnitude of the MGE is systematically related to real-world intergroup biases. Understanding an individual¿s orientation towards more familiar and locally relevant real-world groups will always be more complex than that towards an artificial group because it is affected by many factors that do not contribute to the MGE (such as social status, cultural stereotypes, etc.). This is why the MGP has frequently been used as a simplified proxy reflecting the consequences of ¿mere membership¿ (Dunham, 2018), but it is unclear whether it can play that role successfully. In other words, does the MGE relate to, or might it even be, the initial foundation of an individual¿s tendency to prefer culturally salient real-world groups? If so, the first-order prediction is that real-world group biases and the MGE should be positively correlated (i.e., the extent of an individual¿s MGE should positively correlate with their bias towards real-world groups); Furthermore, this perspective implies the prediction that at least some individual difference or cultural level moderators should similarly relate to both real and minimal group bias.Alternatively, real-world groups and minimal groups might have a trade-off or compensatory relationship, such that individuals who have stronger biases in one have weaker biases in the other. This relationship could occur if cultural emphasis on cooperation with transient groups serves to fill a gap that arises as more traditional social ties break down, a hypothesis derived from the literature on the role of strong institutions and kinship intensity (e.g., Hruschka et al., 2014; Schulz et al., 2019). A hint of this emerges above when we suggest that10some moderators might relate to real and minimal group bias in different ways; for example, permeability might positively relate to minimal groups while negatively relating to salient real- world groups.In other words, in low-trust and low-permeability societies, people might be so strongly tied to existing real groups (such as kinship networks) that they are less likely to favor groups that are based on minimal cues. If this contrastive pattern is ubiquitous, ingroup bias towards those real groups might be negatively related to the MGE. We will explore these possibilities with respect to the three sets of moderators described above in RQ2.1¿RQ2.3, as well as with a few additional moderators that have been actively discussed in the context of real-world groups, as described in our methods section below (e.g., beliefs about the similarity of group members and the status of the groups involved; Grigoryan, Jones, et al., 2022). Thus, by assessing attitudes and behaviors towards real groups as well as towards a minimal group, we can explore a range of questions concerning how r