Philosophical Transactions of the Royal Society B: Biological Sciences
You have accessResearch articles

Are strangers just enemies you have not yet met? Group homogeneity, not intergroup relations, shapes ingroup bias in three natural groups

Gönül Doğan

Gönül Doğan

Faculty of Management, Economics and Social Sciences, Chair of Corporate Development and Business Ethics, University of Cologne, Albertus-Magnus-Platz, 50923 Cologne, Germany

[email protected]

Contribution: Conceptualization, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review & editing

Google Scholar

Find this author on PubMed

,
Luke Glowacki

Luke Glowacki

Department of Anthropology, Boston University, Boston, MA 02215, USA

Department of Human Evolutionary Biology, Harvard University, 11 Divinity Avenue Cambridge, MA 02138, USA

Contribution: Conceptualization, Data curation, Investigation, Project administration, Writing – review & editing

Google Scholar

Find this author on PubMed

and
Hannes Rusch

Hannes Rusch

Department of Microeconomics and Public Economics, School of Business and Economics, Maastricht University, PO Box 616, 6200 MD Maastricht, The Netherlands

Contribution: Conceptualization, Formal analysis, Investigation, Methodology, Writing – review & editing

Google Scholar

Find this author on PubMed

    Abstract

    Humans often favour ingroup members over others, a bias that drives discrimination and intergroup conflicts. Hostile relations between groups and homogeneity within groups may affect such ingroup bias. In an experiment with members of three natural groups in Ethiopia, we vary intergroup relations (neutral versus enmity) and exploit the natural variation in the homogeneity of groups (homogeneous versus heterogeneous) to identify their effect on in- and outgroup concerns. We find that ingroup bias largely manifests as positive concern for ingroup members combined with no concern for outgroup members. Enmity has no effect on ingroup bias, whereas ingroup concern is amplified in homogeneous groups. Group homogeneity, thus, is the primary driver of concerns for others in our study’s context. Our results are relevant to understanding the consequences of exclusionary group identities.

    This article is part of the theme issue ‘Intergroup conflict across taxa’.

    1. Introduction

    Across many societies and settings, people tend to favour their ingroup members over others [14]. However, especially in uncompetitive settings without salient group identities, most people also show positive concern for the welfare of others [57]. Differential cooperation towards in- and outgroups is mostly driven by differential expectations of future reciprocity, supporting the view that within group cooperation shapes individuals’ group identification [8,9]. Outgroup threat increases cooperation within groups in minimal groups [8] which may be due to individuals’ heightened group identification [10]. However, little is known about the level of ingroup bias in members of natural groups experiencing real-life violent conflict [1116], and in particular, a comparison of behaviour when facing natural outgroups with or without enmity—as opposed to unidentified strangers—is lacking. Moreover, in groups with strict identity boundaries, the role of group homogeneity in ingroup bias remains an open question. In this paper, we tackle both dimensions using evidence from a lab-in-the-field experiment in Ethiopia with natural groups whose group identities are a defining feature of daily life, and who have varying intergroup relations (neutral versus enmity) and group homogeneity (ethnic homogeneity versus heterogeneity).

    The nature of the relationship between groups could shape how ingroup bias manifests: with neutral outgroups, we would expect some degree of concern for the wellbeing of others [1723], whereas with rivalry between groups such preferences could be attenuated [3,24]. For example, strong rivalry based on membership of political parties or football clubs increases ingroup favouritism [2527], and members of political groups that also engage in violent conflict show a larger ingroup bias towards the rival outgroup than to strangers [28]. Furthermore, facing a member of a rival group could increase ingroup favouritism by invoking one’s own group identity, that is, by increasing its salience more strongly compared to facing a member of a neutral outgroup [2934]. Moreover, ingroup bias could per se be larger in ethnically and culturally homogeneous groups compared to heterogeneous groups [13,22,23,35,36].

    An important feature of most past research is its focus on WEIRD (Western, educated, industrialized, rich and democratic) societies [37] where group identities and rivalries are not a defining feature of everyday life. Therefore, we take our test to a region where group relations and identities play a crucial role in daily life. We recruited male participants belonging to one of three natural groups inhabiting the South Omo Valley in Ethiopia. Two of these groups, Daasanach and Nyangatom, are long-standing enemies with an extensive history of lethal raids between each other [38,39]. Conflicts are usually motived by a desire to capture livestock or obtain revenge for earlier acts of violence. Nyangatom and Daasanach are distinct ethnolinguistic groups with strong customary institutions that structure daily life. Marriage is primarily between members of the same group requiring the transfer of bridewealth in the form of livestock [40]. Social life is structured by internal institutions specific to each group including generation sets and territorial sections [41,42]. Further, they have strong group identities including essentialist aspects and rigid boundaries [43]. The Ethiopian government has recognized Nyangatom and Daasanach as members of distinct ethnolinguistic groups giving them the right to self-governance at the local (Woreda) level [44]. The third group, Highlanders, has neutral relations with the other two groups, primarily working as shop keepers and traders within the areas of Nyangatom and Daasanach. They are usually recent immigrants to the area from other regions of Ethiopia. Many identify as Wolayta, a large ethnic group numbering over two million, but some are members of other groups including Oromo, Konso or Tigray. Thus, Highlanders in the study area are ethnically heterogeneous without a cohesive cultural identity as found in Nyangatom and Daasanach.

    Behaviour in neutral group pairings forms our benchmark and comes closest to the setups of existing studies on ingroup bias using WEIRD samples [1,13,37]. By comparing behaviour in neutral and enemy group pairings we can measure the directional and level effects of intergroup relations on ingroup bias. The natural variation in ingroup homogeneity in our sample allows us to test whether the two ethnically and culturally homogeneous groups exhibit a larger ingroup bias than the heterogeneous one.

    We employ a simple group decision-making task, a variant of the multiple dictator game [45], in which two decision-makers simultaneously decide whether to appropriate the earnings of a third player. Using a within-subject manipulation, we vary the group memberships of the decision-makers and the affected party resulting in four group constellations. Our design offers multiple advantages over the commonly used two-person dictator game: it captures the essence of team-work that is needed to extract resources from third parties, and as such, is especially relevant in the context of our sample. Since expropriation may be more psychologically costly than giving when facing an outgroup but not when facing an ingroup [29,46], our design provides a stringent test of the difference between in- and outgroup concerns. The need to coordinate on an unfair action diffuses responsibility [45,47], therefore, behaviour is not confounded by image concerns to the extent it is in the standard dictator game [48]. While it is simple to administer, it allows us to cleanly categorize participants with respect to their level of ingroup bias, ranging from purely selfish concerns, over some level of ingroup love, to fully egalitarian. Finally, the group constellation that emerges from the full-factorial design, in which the decision is whether to coordinate with an outgroup member to take from an ingroup member, tests ‘treasonous’ behaviour. To the best of our knowledge, this has not been investigated before. Such behaviour may emerge when the difference between ingroup and outgroup love is rather small and the selfish concern is rather large; however, it would be starkly at odds with group norms when facing an enemy. We devise a simple utility model to organize our hypotheses and to classify participants according to their level of concern for the welfare of members of their ingroup and the outgroup. Ingroup bias in our framework is thus the difference between the levels of concern for ingroup and outgroup members. We also evaluate our results in light of both the bounded generalized reciprocity theory and social identity theory [1].

    We find that participants show a universal lack of concern for outgroups’ welfare irrespective of intergroup relations. Thus, ingroup bias manifests as concern for ingroup members’ payoffs with no such concern for outgroups. Enmity does not affect ingroup concerns whereas group homogeneity matters. The two homogeneous groups in our sample show larger ingroup concerns than the heterogeneous one. In our pooled sample, about a quarter are classified as selfish and more than 40% show large ingroup concerns. In sum, (i) ingroup bias manifests as positive welfare concerns for the ingroup for the majority of participants, (ii) is unaffected by intergroup relations and (iii) is stronger in homogenous groups.

    We contribute to the existing literature as follows: first, we add to the literature on ingroup bias by showing how real-life enmity affects it. With respect to the intensity of enmity and violent conflict in the populations studied, the closest to ours are those that investigated ingroup bias during or after periods of violent Jewish-Arab conflict in Israel and Palestine [49,50], during Protestant-Catholic conflict in Northern Ireland [51] and around the time of violent clashes of political rivals in Thailand [28]. Our study extends this literature to a previously unstudied non-WEIRD sample using resource allocation games and neutral outgroups as comparison benchmarks to pin down the role of enmity. Thus, we investigate the role of enmity vis-à-vis a peaceful relationship when facing outgroups for both types of relationship, rather than, as previously has been done, inferring the extent of ingroup bias in a peaceful relationship based on encountering strangers without a salient group identity or whose group identity is not known. Second, our design allows us to separate ingroup and outgroup concerns, and isolate the effects of enmity and group homogeneity on both types of concerns. Third, we document the determinants of ingroup bias in a population in which group membership is a defining feature of daily life. Previous experimental work on ingroup bias mostly relies on either artificially created, minimal groups or on geographically, ethnically, socially, or otherwise demarcated subgroups within the WEIRD sphere (see the reviews [1315]). One notable exception is Romano et al. [4] who found ingroup bias in several non-WEIRD countries. Our results strikingly diverge from previous findings as we find a universal lack of concern for outgroups, and most of our participants are ‘groupy’ [52,53]. Thus, we provide new insights to the ongoing quest for a systematic understanding of how group identities shape social behaviour.

    2. Methods

    (a) Experimental design and participants

    This study was conducted in Nyangatom and Daasanach Woredas (governmental units) of southwest Ethiopia. As described in the introduction, Nyangatom and Daasanach are small ethnic groups with strong identities, living primarily as nomadic pastoralists and inhabiting largely ethnically homogenous regions. They have a long history of violent conflicts with each other that primarily take the form of tit-for-tat raiding [39,54]. Though they also have trading relationships occasionally, they recognize each other as enemies; in fact, the Nyangatom refer to Daasanach as ‘emoit’, literally meaning ‘enemy’. Ethiopian Highlanders are ethnically heterogeneous and do not engage in intergroup conflict with either Daasanach or Nyangatom.

    We recruited only adult male participants to best simulate real-world between-group decision-making processes, where only adult men are involved in intergroup conflict. In total, 192 males participated in the experiment who belong to one of three natural groups: Daasanach (N = 48), Nyangatom (N = 96) and Highlanders (N = 48). Our sample size lets us detect an effect size of w = 0.23 with 80% power and 5% significance, calculated using G * Power 3.1.9.4 [55]. Such an effect size amounts to a 12% change in the outcome variable of our main treatment comparison (expropriation rate in enmity versus neutral relations). Nyangatom and Highlander participants were recruited from the town of Kangaten while Daasanach participants were recruited from the town of Omorate. Participants from the same cultural group were assembled in groups of 12 in a private area and participated in the study individually. Verbal informed consent was obtained from all participants and the Committee on the use of Human Subjects at Harvard University approved this study.

    The experiment was conducted using verbal instructions with neutral framing; all possible choices and their payoff consequences were common knowledge. Participants remained anonymous and did not learn the decisions of anyone else throughout the experiment. A local research assistant who was familiar with the purposes and procedures of the study translated study instructions and participant responses on the spot. Participants were informed of the study procedures and told that they would be paid on the basis of one randomly chosen allocation decision they made. For each of the four choices they made, they were told the cultural identities (Nyangatom, Daasanach or Highlander) of all other players involved. After completing all four choices which were presented in a counterbalanced order, participants were debriefed and asked not to talk about the experiment. Participants then received a show-up fee of 20 birr (approx. 1 USD at the time). Payments were based on random selection of one condition and were made after the conclusion of data collection for all participants. Verbal instructions and payment procedures can be found in the electronic supplementary material.

    We employed a simple group decision-making task, a variant of the multiple dictator game [45]. In our task, two players decide whether to expropriate the endowment of a passive third player, the ‘victim’ for short. In the experimental instructions, this choice was denoted as ‘taking’ the tokens of the third player in a group of three. Initially, each player has 10 tokens. The victim’s endowment is equally redistributed to the two decision-makers only if both choose to expropriate. Thus, with redistribution, decision-makers each earn 15 tokens and the victim earns zero. Conditional on group identities of all players, we focus on three concerns that may play a role in players’ decisions: a selfish concern, a concern for the welfare of one’s ingroup members and a (lack of) concern for the welfare of the outgroup members. We assume that these concerns directly affect behaviour and are therefore reflected in the resulting taking behaviour in this game. Thus, we infer participants’ concern types based on their expropriation behaviour.

    Using a within-subject manipulation, we varied the group memberships of the decision-makers and the victim. Each participant made four decisions, one for each of the following cases: Out-In, In-In, In-Out and Out-Out. Herein, the first word denotes the co-decision-maker’s group membership and the second denotes the victim’s group membership, both relative to the participant. At the end of the experiment, one group constellation was randomly chosen for payment (see electronic supplementary material for detailed procedures). Observing players’ choices in these four group constellations enables us to classify participants according to their concern for the welfare of members of their ingroup and the outgroup. Using a between-subject manipulation, we varied the group relations (neutral versus enmity). Participants were informed of the identity of the outgroup (Nyangatom, Highlander, Daasanach) prior to decision-making.

    (b) Theoretical framework

    We use a relatively simple linearly additive utility model to organize our hypotheses and results. Individuals derive utility from their own payoff, xi, the total payoff of their ingroup excluding themselves, XI, and the total payoff of the outgroup, XO, based on quasiconcave functions f, w, and h with f1, w1, h1 > 0, f11, w11, h11 ≤ 0, and h(0) = w(0) = 0. Parameters μi and βi model the level of concern for the welfare of ingroup and outgroup players, respectively. In the absence of competition or conflict, most individuals show a concern for others’ welfare regardless of group membership [5,56,57]. We conceptualize ingroup bias as a positive difference between the level of concern for the welfare of ingroup members and outgroup members. Thus, we assume that individuals may have a preference for the welfare of the outgroup (outgroup love), but at a lower level than their preference for the welfare of their own group (ingroup love), which is at a lower level than their preference for their own payoff, leading to 1 > μi > βi. Further, we assume that participants exhibit ingroup love, that is, μi ≥ 0. We also assume βi to be non-negative in neutral group pairings based on the findings of previous research [20,58]. Next, we assume that players always prefer to increase their group members’ payoffs over outgroup members’ payoffs, therefore, for all positive x, w1(x) > h1(x). Together:

    Ui=f(xi)+μiw(XI)+βih(XO).

    Based on this model and some general assumptions about distributional and efficiency preferences (see model predictions in the electronic supplementary material), the behavioural predictions for neutral group pairings are straightforward and as follows: given the same co-decision-maker, taking money from an outgroup victim is weakly preferred over taking money from an ingroup victim. Likewise, given an ingroup victim, taking together with an ingroup co-decision-maker is weakly preferred over taking together with an outgroup co-decision-maker. Notice that the higher the concerns for ingroup members in the group, the lower the expropriation rate of ingroup members. The fact that people care more about their ingroup members’ welfare than their outgroup members’ welfare then gives the following proposition.

    Prediction 2.1.

    Group constellations ordered according to predicted expropriation rates are:

    Out-In<In-In<min{Out-Out,In-Out}.

    Members of homogeneous groups may care less about the outgroup, i.e. 0 ≤ βi < μi. Moreover, outgroup hate, as possible in the case of enmity, would lead to a negative β [11,24,59,60]. We would then expect homogeneous groups to expropriate more often in constellations with an outgroup victim (In-Out and Out-Out), and less in the constellation with an ingroup victim and outgroup collaborator (Out-In) compared to the heterogeneous groups.

    Prediction 2.2.

    Consider a homogeneous group whose members have smaller β than the members of a heterogeneous group. Then, expropriation rates are higher in Out-Out and In-Out, unchanged in In-In, and smaller in the Out-In group constellation in the homogeneous group compared to the heterogeneous one.

    Consider now outgroup hate as a special case. With outgroup hate, β is negative. In that case, there is full expropriation of an outgroup victim, i.e. in Out-Out and In-Out. Further, in Out-In expropriation rates are smaller with smaller β because of the disutility from the payoff the outgroup member receives from expropriation.

    Enmity or group homogeneity could heighten ingroup concerns. A group could have a larger ingroup concern when facing an enemy than when facing a neutral outgroup [30,61], i.e. μi > μi. Likewise, members of a homogeneous group could have larger ingroup concerns than members of a heterogeneous group. In such a case, the rate of expropriation of ingroup members will be lower, i.e. in In-In and Out-In, and expropriation in In-Out will be higher with positive β.

    Prediction 2.3.

    Assume that facing an enemy increases ingroup concerns for all i such that μi > μi. Then, expropriation rates in In-In and Out-In are lower, in Out-Out unchanged, and in In-Out larger with μi compared to μi.

    To sum up, compared to neutral outgroups, if enmity decreases concerns for the outgroup’s welfare or leads to outgroup hate, then members of enemy outgroups would be targeted more often, with full expropriation in case of outgroup hate, i.e. in Out-Out and In-Out. Further, if being reminded of the enemy makes one’s own group identity more salient, thereby increasing concerns for ingroup welfare, ingroup members would be expropriated less often, i.e. in Out-In and In-In. The same comparative statics would also be obtained for more homogeneous groups. Notice that expropriation in Out-In is akin to treason in real-life groups with enmity relations, where any such collaboration with the members of enemy groups would be prohibited and highly sanctioned in real-world contexts. Therefore, we would not expect any expropriation in Out-In in the enmity condition. Any expropriation in Out-In can thus only be explained by fully selfish concerns, that is, a complete lack of ingroup concerns.

    3. Results

    Figure 1a shows the percentage of participants who expropriate per group constellation and per group relationship where group constellations with an ingroup victim are depicted on the left. Figure 1b shows the percentage of participants expropriating from an ingroup member per group type (homogeneous versus heterogeneous).

    Figure 1.

    Figure 1. Percentage of expropriators (a) per group relationship and group constellation; (b) per group type and group constellation with ingroup victims. Daasanach and Nyangatom are enemies, and Highlanders have neutral relationships with both Daasanach and Nyangatom. Highlanders are a heterogeneous group, and Daasanach and Nyangatom are strongly homogeneous. Error bars indicate 95% Wilson’s confidence intervals for single proportions.

    Two results stand out: first, remarkably, almost everyone (96.6%) chose to expropriate when the victim is an outgroup member regardless of intergroup relations. Thus, participants do not care enough about the welfare of even the neutral outgroup members in order to establish the equitable outcome. In terms of our model, our participants’ βis are too low to compensate for foregone earnings from expropriation. Second, for a substantial proportion of participants, behaviour is consistent with positive welfare concerns for one’s ingroup. As opposed to virtually full exploitation of outgroup victims, only 40.1% of participants chose to expropriate ingroup victims. The difference is highly significant and robust to the type of intergroup relation and group homogeneity (all McNemar exact tests ps < 0.0002; odds ratios > 3.048). Further, participants were indeed reluctant to expropriate ingroup victims when this required coordinating with outgroup members, that is, in Out-In. Nonetheless, such ‘treason’ is still remarkably prevalent, even in the enmity condition.

    Surprisingly, enmity plays no role in ingroup bias. If the victim is an ingroup member, expropriation rates in enmity are not significantly different from those in neutral (50.0% versus 43.8% in In-In, 36.5% versus 30.2% in Out-In in enmity versus neutral conditions, respectively, and tests of proportions ps > 0.358). Since outgroup members are universally exploited, expropriation rates in enmity and neutral are virtually the same in Out-Out and In-Out.

    Finally, group homogeneity lowers ingroup expropriation rate, and this behaviour is consistent with larger ingroup concerns in homogeneous groups: the heterogeneous group, Highlanders, had the highest rate of expropriation when targeting an ingroup member. Comparing Highlanders and the pooled data of Daasanach and Nyangatom, expropriation rates in the Out-In and In-In constellations are significantly higher for Highlanders (test of proportions p = 0.0133 and p = 0.0015, respectively). Further, about half of the Highlanders collaborate with an outgroup member to expropriate another Highlander, pointing to rather weak concerns for the welfare of their ingroup members.

    Next, we classify our subjects into types based on their decisions in the four group constellations that are consistent with the in- and outgroup concerns we modelled. Table 1 shows that 172 out of 192 participants (89.58%) can be classified as one of four types. Expropriating in all cases but Out-In is consistent with a positive but low level of ingroup concerns. Expropriating outgroup members but not ingroup members is consistent with a high level of ingroup concerns. Overall, about 60% of our participants exhibit ingroup bias manifesting as some level of ingroup love and no detectable outgroup love, and in most, behaviour is consistent with high levels of ingroup love. More than a quarter of participants show no ingroup bias in our setup as they always chose the selfish option (see e.g. [52,53] for results in a standard subject pool). Consistent with the aggregate results, ingroup bias is larger in homogeneous groups: whereas the modal type is selfish in the heterogeneous group, in the two homogeneous groups, almost half of the participants exhibit a high level of concern for their fellow ingroup members. Thus, ingroup bias is larger in homogeneous groups and manifests as a higher level of positive concern for the welfare of one’s ingroup members.

    Table 1. Percentage of types in the population and per group type (homogeneous versus heterogeneous). Number of observations in overall are 51, 36, 83, 2 and 20, per type, respectively. The corresponding number of observations in the homogeneous group are 21, 10, 13, 1 and 3, and in heterogeneous groups are 30, 26, 70, 1 and 17, respectively.

    choice pattern type overall heterogeneous homogeneous
    always expropriate selfish 26.56% 43.75% 20.83%
    expropriate except Out-In low ingroup concerns 18.75% 20.83% 18.05%
    expropriate outgroup high ingroup concerns 43.23% 27.08% 48.61%
    never expropriate fair 1.04% 2.08% 0.69%
    other other 10.42% 6.25% 11.81%

    4. Discussion and conclusion

    Overall, our results support theories that conceive of ingroup bias as a concern for the wellbeing of ingroup members with no or little concern for the outgroup, even in neutral group pairings. The wide-spread exploitation of outgroup members prevents us from testing whether enmity decreases concerns for the outgroup relative to neutral group relations. Group homogeneity heightens ingroup bias in our setup: the two homogeneous groups we study exhibit larger ingroup concerns compared to the heterogeneous one. However, the marginal effect of group homogeneity on ingroup bias is much smaller than that of facing of an outgroup victim.

    Our study has the following limitations. In our design, we balanced observations across the enmity and neutral relation conditions, as a result of which, the test for the effect of group homogeneity relies on a smaller sample of observations for the heterogeneous group compared to the homogeneous groups. Although the observed effect sizes are small to medium when comparing ingroup expropriation rates between homogeneous and heterogeneous groups (ws > 0.179, see electronic supplementary material for details), future work can generate a more precise estimate of this effect while also showing whether group homogeneity impacts expropriation behaviour when coordinating with an outgroup member. Second, our study relies on three natural groups. Whether our results are robust in other natural groups with enmity, both in this region and elsewhere, and whether ethnically or culturally demarcated homogeneous groups with peaceful relations to one another show a similar type of ingroup bias remain open questions. Third, the dichotomous nature of decisions in our setup, although making it particularly suitable for our sample due to its ease of administration, provides only crude information on in- and outgroup concerns. Extending the choice set by allowing participants to choose how much to expropriate where possible would generate a more refined picture of ingroup bias.

    Evaluating our results in light of the two major perspectives in social psychology employed to study ingroup bias, social identity theory and bounded generalized reciprocity [1,62,63], we find some support for the latter. Since intergroup relations and group homogeneity are not originally incorporated in these perspectives, we derive predictions based on their main principles. According to bounded generalized reciprocity, people in groups cooperate with each other to keep a positive reputation as a cooperator, and cooperative behaviour within groups evolved as it increases the group’s survival chance. In a heterogeneous group with weak social ties, since reputation is harder to build and maintain, cooperation with ingroup members may be lower and ingroup expropriation may be higher. Extending this reasoning to intergroup relations, it may be expected that the presence of an enemy moderates the relationship between cooperation and group survival, and therefore, cooperation within the group increases in the presence of an enemy. Social identity theory posits that ingroup bias emerges as a result of the positive social identity that ingroup cooperation gives. While one’s social identity from cooperation would not be expected to be affected by the degree of ingroup homogeneity, the presence of an enemy outgroup may heighten the social identity derived from ingroup cooperation. As such, ingroup bias would increase when facing an enemy compared to a neutral outgroup. Our finding that ingroup expropriation is lower in homogeneous groups is thus in line with the prediction derived from the approach of bounded generalized reciprocity whereas the absence of an enmity effect on ingroup bias goes against the predictions of both the social identity theory and bounded generalized reciprocity approaches. All in all, our results provide partial support for the bounded generalized reciprocity approach.

    The preferences our participants revealed via their behaviour differ in important ways from those of more commonly investigated Western samples. In our study, almost all participants chose to expropriate the earnings of another from a neutral outgroup and about half showed concerns for ingroup members’ welfare. The latter is akin to the rate of fair choices in Western samples with comparable payoff structures and without a salient group identity [45,6466]. Supporting previous findings on the investigation of the role of culture on ingroup bias [63] as well as the conclusions based on the latest developments of the social identity perspective [10], the reason why a large proportion of participants in WEIRD countries acts fair towards others and avoids harming them [67,68] could be common group identities that are not strongly ethnically or culturally demarcated, such as being residents of the same country. Our findings suggest, however, that such fair behaviour may vanish if homogeneous group identities take hold and exclusionary group identities become salient.

    Ethics

    Verbal informed consent was obtained from all participants and the Committee on the use of Human Subjects at Harvard University approved this study.

    Data accessibility

    The data and analysis are included in the electronic supplementary material [69].

    Authors' contributions

    G.D. and H.R.: conceptualization, formal analysis, investigation, methodology, writing—original draft, writing—review and editing; L.G.: conceptualization, data curation, investigation, project administration, writing—review and editing.

    All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

    Competing interests

    The authors declare no competing interests.

    Funding

    G.D. received financial support by the Dr Jürgen Meyer Foundation. L.G. and H.R. benefitted from participation in the ‘Evolution and Warfare’ Investigative Workshop at NIMBIOS, sponsored by the National Science Foundation through NSF award no. DBI-1300426, with additional support from the University of Tennessee, Knoxville. L.G. declares additional financial support by the Eric M. Mindich Research Fund for the Foundations of Human Behaviour, the Mind Brain and Behavior Interfaculty Initiative at Harvard University. None of the funders had any role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

    Acknowledgements

    We thank Robert Böhm, Bernd Irlenbusch, Rudy Ligtvoet, Florian Morath and participants at the Thurgau Experimental Economics Meeting and the 2019 meeting of the German Association for Experimental Economic Research (GfeW e.V.) for helpful comments and suggestions.

    Footnotes

    One contribution of 19 to a theme issue ‘Intergroup conflict across taxa’.

    Electronic supplementary material is available online at https://doi.org/10.6084/m9.figshare.c.5879945.

    Published by the Royal Society. All rights reserved.