Explaining a Rare Null Relationship Between Group Identification and Social Protest Through a Relational Form of Coping With Low Group Status

Primary and meta-analytic research strongly suggests that group identification motivates disadvantaged group members for social protest to achieve social change. However, most studies on social protest are conducted in contexts that are already conducive to this positive relationship (i.e., conditions of hope and scope for social change). Two studies of Indonesian ethnic minority group members’ coping with low group status add to this literature by testing (a) whether group identification motivated social protest in a cultural context in which group identity is valued while hope and scope for social change are lacking, and (b) alternatively, whether individuals engage in a relational form of coping that revolves around seeking shelter in social relationships to protect well-being. In both studies (N = 80 and N = 132), we predicted and found a rare null correlation between group identification and social protest, and rather strong support that coping with low group status revolved around seeking shelter in social relationships. We discuss the theoretical and practical implications of our findings for identity management and relationship regulation processes in contexts that lack hope and scope for social protest to achieve social change.

However, most collective action research is conducted in contexts in which collective action is likely or already ongoing (Van Zomeren, 2013). According to social identity theory (Tajfel, 1978;Tajfel & Turner, 1979), such contexts already offer hope and scope for social change and it is exactly under these socio-structural conditions that group identification motivates collective action, reflecting social competition for intergroup status to achieve social change (Ellemers, 1993;Scheepers, Spears, Doosje, & Manstead, 2006). Under conditions of low hope and scope for social change, however, social identity theory predicts that group identification should not serve this mobilization function. Although it is less clear how individuals manage their identity under these conditions (see Cohen-Chen, Halperin, Saguy, & Van Zomeren, 2014;Klein, Spears, & Reicher, 2007;Tausch et al., 2011), it does suggest that increasing group members' group identification should not increase their motivation to undertake social protest to achieve social change.
We therefore purposefully selected a cultural context in which, across the board, group identities are valued while hope and scope for social change are lacking and boundaries between groups are closed. Specifically, we focused on how Indonesian ethnic minority group members cope with low group status. There are two key reasons for this choice. First, the Indonesian context features a large ethnic majority group and many different small ethnic minority groups, which together reflect a clear ethnic hierarchy. Second, based on Hofstede's (2001) classification of Indonesian mainstream culture as revolving around strong collectivism (defined as whether individuals are interdependent on others and loyal to their in-groups) and power distance (defined as whether those low in power legitimize, accept, and respect intergroup power relations), the ethnic hierarchy can be conceptualized as a system in which ethnic group identities are valued while hope and scope for change is lacking and intergroup boundaries are closed.
Within this context, we tested whether collective action was still a viable way to cope with low group status, or, if this was not the case, whether individuals engage in a relational form of coping that revolves around seeking shelter in one's social relationships to protect well-being (e.g., Bowlby, 1969;Mikulincer & Shaver, 2007;Van Zomeren, 2014, 2015, 2016. We conducted two studies (N = 80 and N = 132) in which we thus sought to replicate a rare null correlation between group identification and collective action, and to identify this new relational form of coping with low group status.

Identity Management Strategies to Cope With Low Group Status
Psychological models of collective action developed over the past decades to allocate a pivotal role to the notion of group identification (e.g., Drury & Reicher, 2009;Thomas, Mavor, & McGarty, 2012;Van Zomeren et al., 2008). Because of individuals' identification with the group, they think, feel and act on behalf of their group when they engage in collective action to achieve social change (e.g., Wright, Taylor, & Moghaddam, 1990; see also Van Zomeren & Iyer, 2009). Indeed, group identification strongly and positively predicts participation in collective action or intentions thereof (Klandermans, 1997;Simon et al., 1998; for a meta-analysis see Van Zomeren et al., 2008).
Theoretically, collective action represents an identity management strategy (Tajfel, 1978; see also Ellemers, 1993;Scheepers et al., 2006), which motivates individuals to actively and competitively seek to improve their group's position relative to the relevant outgroup (Van Zomeren, Leach, & Spears, 2012). This focus on seeking social change through collective action is very different from other identity management strategies such as individual mobility (which refers to leaving the group to join the higher status group) and social creativity (which refers to a motivated shift from a status-relevant dimension to a status-irrelevant dimension in order to maintain or increase positive self-evaluation; Ellemers, 1993;Tajfel, 1978). Surprisingly, the collective action literature has paid little attention to the social identity prediction that social competition only represents a viable way to cope with low group status when individuals can envision hope and scope for social change (i.e., status illegitimacy and instability, coupled with impermeable group boundaries that prevent individual mobility; Tajfel & Turner, 1979; see also Bettencourt, Charlton, Dorr, & Hume, 2001;Ellemers, 1993;Tausch et al., 2011). Most research on collective action has indeed focused on contexts in which collective action is likely or already ongoing; that is, in which there is hope and scope for social change (Van Zomeren, 2013). Surprisingly little is known, therefore, about how individuals cope with low group status in contexts in which there is little hope and scope for social change (for discussions, see Klein et al., 2007;Tausch et al., 2011). Exactly because collective action and individual mobility may not be viable coping strategies under such conditions, this raises the question how group members actually cope with low group status i . Below, we propose a new relational form of coping that serves to protect individuals' well-being -a coping strategy that seems well-suited to members of low status groups under conditions of low hope and scope for social change.

A New Relational Form of Coping
Humans have a fundamental need for a safe haven they can seek shelter in (Ainsworth, 1989;Baumeister & Leary, 1995;Bowlby, 1969;Mikulincer & Shaver, 2007;Van Zomeren, 2016). Although some approaches conceptualize this as a general need to belong (e.g., Baumeister & Leary, 1995), others focus more specifically on attachment processes such as attachment anxiety and avoidance (e.g., Mikulincer & Shaver, 2007). Consistent with this view, social support is essential for successful coping with adversity to the extent that it protects well-being (Lazarus, 1991). This also fits with the approaches that suggest that successful relationship regulation is essential for human survival, procreation, and well-being (Fiske, 1992;Rai & Fiske, 2011;Van Zomeren, 2016). Indeed, having access to supportive social relationships buffers stress (Lazarus, 1991) and reduces mortality risks (Holt-Lunstad, Smith, & Layton, 2010). This importance of social relationships as a basis for coping has therefore led some to argue that humans are essentially relational beings (Van Zomeren, 2014, 2015, 2016 who have a social baseline (Beckes & Coan, 2011).
Applying this line of thought to coping with low group status under conditions that make individual mobility and collective action unlikely, we propose that individuals can engage in a relational form of coping that leads them to seek shelter in their relationships (e.g., partner, family, friends) in order to protect their well-being. Note that this line of thought does not imply that this makes group identity less relevant or salient. In fact, because the social relationships individuals seek shelter in will likely be ingroup members, those relationships may come to define the group identity (e.g., Easterbrook & Vignoles, 2013;Postmes, Spears, Lee, & Novak, 2005). However, because individuals seek to regulate those relationships to protect well-being (through seeking shelter in them), it seems unlikely that they will engage in identity management strategies such as individual mobility, social creativity, or indeed collective action.

The Current Research
To test our line of thought, we selected a cultural context in which ethnic group identities but also stable status and power differences between them are valued, thus offering little hope and scope for social change through collective action (and also little room for individual mobility, as group boundaries are closed). According to Hofstede (2001), Indonesia's mainstream culture ii is characterized by a strong tendency towards accepting and legitimizing inequality (i.e., a power distance score of 78 out of 100). This suggests low hope and scope for social change with respect to status differences between ethnic groups, which makes collective action less likely. At the same time, Indonesia's mainstream culture is characterized as strongly collectivist (i.e., an individualism score of 14), which implies a strong preference for a stable societal framework that invites conformity to the ideals of society and the in-groups to which they belong. Thus, group boundaries are likely to be impermeable, making individual mobility less likely. We therefore expected that ethnic group identification among members of ethnic minority groups, in this context, should not predict collective action; and that our proposed new relational form of coping with low group status should revolve around seeking shelter in social relationships to protect well-being. We thus conducted two studies to replicate a rare null correlation (between group identification and collective action) and to examine a new relational form of coping with low group status (with seeking shelter in social relationships positively predicting well-being).

Study 1 Method Participants
Participants in this study were 80 ethnic minority group members (65% male, mean age = 32.19, SD = 7.09, range 23-52 years) who were employees of a state-owned company in Indonesia. The second author collected the data via questionnaires in, and in collaboration with, the company. The company reflected country characteristics to the extent that the workforce was dominated by members of Javanese and Sundanese ethnic groups, which together totalled around 70% of the company's population. Throughout the whole study, anonymity was guaranteed, participation was on a voluntary basis, and participants indicated informed consent.
Participants came from twenty-six ethnic minority groups in Indonesia. Seven participants came from the Bugis ethnic minority group, while the remainder of the sample derived from 25 ethnic group minority groups (with 5 or less participants from each group) such as Papua, Nias, Madura, Banjar, Flores, Minahasa, Ambon, Bali, to name a few of them. Most participants were undergraduates (75%) and at the lower staff level (60%).
The questionnaire consisted of two sections in a booklet in Indonesian and English and started off with the sentence: "The purpose of this survey is to examine individuals' sense of ethnic identity and its role in their life". The first part included demographic information such as gender, age and ethnicity. The second section included a set of questionnaire items tapping the key variables under study. Responses were scored on 7-point Likert-type scales ranging from strongly disagree (1) to strongly agree (7). The variables we used in both studies and for which we do not list all items in the text below can be found in Appendix 1.

Procedure and Statistical Power
The second author sent a request for data collection and the questionnaire booklet to the company's human resources department. After receiving approval for the study, the questionnaires were uploaded by the human resources staff via the intranet of the company. It was explained that the participants had to be employees belonging to ethnic minority groups. Questionnaires were returned through email or sent as hard copies to the human resources office. The total number of questionnaires received was 92. However, 12 were not analyzed because 10 were from Javanese participants (belonging to the ethnic majority group), and two participants did not fill in any demographic data (including their ethnicity). Thus, data included in the study, entered in SPSS by the second author, is based on 80 questionnaires in total. We conducted a power analysis with GPower (Faul, Erdfelder, Buchner, & Lang, 2009) to determine which sample size was needed to be able to identify a positive medium-sized effect (i.e., r = .30) of group identification on collective action (as identified in the meta-analysis by Van Zomeren et al., 2008). With a desired power of .90 and an alpha of .05, a sample size of 88 is needed (for a one-tailed test, which is appropriate given that we predicted the effect to be smaller, not greater). Although the current sample size is close to this number, ideally, it should have been somewhat larger.

Measures
We included multiple indicators of group identification to prevent reliance on only one index of the construct (for a review of different measures, see Leach et al., 2008). Specifically, we measured ethnic group identification in four different ways. First, we used Phinney's (1992)  Inspection of the pattern and structure matrices suggested that the third factor relied on only one item that also loaded highly on the first (self-investment) factor. Because the first and second factors were in line with the selfinvestment and self-definition dimensions, we decided to calculate these two scales. The reliability of the self-investment and self-definition scales was very good (Cronbach's alphas of .95 and .91, respectively).
Third, we used a single-item indicator of group identification that scholars recently suggested to be sufficient to tap the construct (Postmes, Haslam, & Jans, 2013). This item was simply: "I identify with my ethnic group".
Finally, we used Valk and Karu's (2001) ethnic pride measure, which consisted of 10 items such as "I am proud of my ethnic group membership" or "Being conscious of my ethnic background increases my feeling of confidence".
Principal axis factoring with oblique rotation suggested one factor with an eigenvalue > 1 underlying these ten items, explaining 57.45% of the variance. The reliability of the scale was very good (Cronbach's alpha = .93).
Thus, we measured ethnic group identification in four different ways that derive from different sources. As can be seen in Appendix 2, and as one could expect, correlations between these different measures were all positive and significant. Moreover, means were around the middle of the scale, suggesting that ethnic group identity was, across the board, psychologically meaningful and relevant.
Collective action tendencies were measured with two items focused on social protest that correlated strongly and positively (r = .86, p < .001) and were therefore combined into one scale. The first item was "I would like to engage iii . In line with expectations, the mean level of collective action was fairly low (see Appendix 2).
We then measured seeking shelter in social relationships with the 12-item Multidimensional Scale of Perceived Social Support (Zimet, Dahlem, Zimet, & Farley, 1988). Examples of items are "There is a special person who is around when I am in need", "I get the emotional help and support I need from my family" and "My friends really try to help me". Principal axis factoring with oblique rotation suggested two factors with an eigenvalue > 1 underlying these twelve items, explaining 55.35% of the variance. However, the second factor relied weakly on only two items and thus we decided to collapse across the two factors. The reliability of the scale was very good (Cronbach's alpha = .93) and the mean level of seeking shelter in social relationships was fairly high (see Appendix 2).
Personal well-being was measured using the 5-item Satisfaction with Life Scale (Pavot & Diener, 1993) such as "I am satisfied with my life", "In most ways my life is close to my ideal" and "The condition of my life is excellent". Principal axis factoring with oblique rotation suggested one factor with an eigenvalue > 1 underlying these ten items, explaining 63.91% of the variance. The reliability of the scale was very good (Cronbach's alpha = .89) and the mean level was fairly high (see Appendix 2).
Finally, we included a measured of perceived discrimination for two reasons. First, the perception of group-based discrimination should predict collective action to the extent that it is perceived as unfair (e.g., Van Zomeren et al., 2008). Second, and relatedly, the perception of group-based discrimination is thought to increase group identification (e.g., Branscombe, Schmitt, & Harvey, 1999). Including this variable thus enabled a test of these predictions in the current context. We measured it with the Detroit Area Survey (DAS) Discrimination scale (Perez, Sribney, & Rodríguez, 2009) that consisted of eight items plus an additional item that specifically tapped ethnic group discrimination. Examples of items used were "You are treated with less courtesy than other people", "You receive poorer service than other people at restaurants and stores" and "You feel discriminated on the basis of your ethnic group membership" (which was the item specific to ethnic discrimination). Principal axis factoring with oblique rotation suggested one factor with an eigenvalue > 1 underlying these nine items, explaining 58.18% of the variance.
The reliability of the scale was very good (Cronbach's alpha = .90) and the mean level was fairly low (see Appendix 2) iv .

Results
Our analytic strategy was to first inspect the key correlations between the different indices of ethnic group identification and collective action tendencies, which we expected to be zero. Second, we tested whether individuals coped with low group status through seeking shelter in social relationships (abbreviated for ease of reference below to "social relationships") through regression analyses, for which we expected that social relationships would positively predict well-being.

A Rare Null Correlation Between Group Identification and Collective Action?
To prevent a reliance on one particular measure of ethnic group identification, we included different ones. We reasoned that if the pattern of results would converge, we would be in a better position to interpret any null correl- was significantly different from zero. The other correlations were even clearer null correlations (i.e., the Leach et al. self-definition measure, the single-item measure of group identification, and the ethnic pride measure). These findings support our expectation that stronger group identification, independent of how we measured it, is not positively predictive of collective action tendencies in the current cultural context. Thus, it seems that collective action was not a viable way to cope with low group status in this context.
Furthermore, and in line with our reasoning, social relationships were positively and significantly related to ethnic group identification. As can be seen in Appendix 2, this was the case across all identification indices (with rs ranging from .27 -.50, all ps < .05). Thus, stronger social relationships related to stronger identification with one's ethnic group (which was already found to be unrelated to collective action tendencies; and corroborating this notion, we also found a null correlation between social relationships and collective action; r = -.19, p = .10). In sum, seeking shelter in social relationships was positively related to group identification in this cultural context, but neither motivated collective action.

Identity Management and/or Relationship Regulation?
We then tested whether individuals coped with their low status group membership through identity management processes (with group identification predicting collective action) and/or relationship regulation processes (with social relationships positively predicting well-being). To do so, we conducted regression analyses to predict collective action tendencies, well-being, and, more exploratively, perceived discrimination.
We conducted a regression analysis to predict individuals' collective action tendencies from their group identification (all findings reported here refer to self-investment), their social relationships, and, as suggested by collective action research, perceived discrimination. This regression model was not statistically significant, F = 1.77, p = .16.
We then predicted individuals' well-being from their group identification, social relationships, and perceived discrimination. The regression model was significant, F = 4.73, p < .01, and inspection of the parameter estimates showed that, in line with expectations, only social relationships predicted well-being significantly (β = .44, p < .001; for group identification, β = -.07, p = .54; for perceived discrimination, β = .16, p = .16). This is in line with the idea that social relationships serve a protective function in coping with low group status (which is particularly striking given that group identification did not uniquely predict well-being).
In a more exploratory analysis, we predicted individuals' perceived discrimination from their group identification and social relationships. We found that the regression model was significant, F = 4.38, p = .02, with inspection of the parameter estimates showing that again only social relationships predicted perceived discrimination significantly (β = -.34, p < .01; for group identification, β = .17, p = .15). Thus, seeking shelter in social relationships was not only related to stronger well-being, but also to perceiving less group-based discrimination.

Discussion
Study 1 provided first evidence of a rare null correlation between group identification and collective action tendencies in the current context. Furthermore, it showed that seeking shelter in social relationships, as a relational form of coping with low group status, uniquely predicted stronger well-being (as well as weaker perceptions of discrimination). By contrast, ethnic group identification predicted neither variable. These findings support our expectations and are in line with the idea that individuals engaged in a relational form of coping, although this does not make with the positive relationship we obtained between social relationships and group identification.
However, Study 1 has a number of limitations that suggest some caution in interpreting its findings. First, with a sample size of 80 participants, statistical power was not ideal and thus we wanted to increase sample size in Study 2. A second limitation is that we sampled from a company in Study 1, whereas ideally we would like to sample from a broader population. This is what we did in Study 2. Third, Study 1 did not examine the possibility that individuals engaged in social creativity, an identity management strategy that revolves around seeking higher status on less central status dimensions (Tajfel, 1978;see Ellemers, 1993). We therefore included a measure of this in Study 2 to see whether individuals engaged in this identity management strategy. Finally, to gather more information about the viability of the underlying (social) identity management framework to predict coping with low group status in this context, we included measures of three theorized socio-structural antecedents of group identification and identity management strategies (Tajfel, 1978). This enabled a broader test of social identity theory predictions (e.g., Bettencourt et al., 2001;Ellemers, 1993;Mummendey, Kessler, Klink, & Mielke, 1999;Scheepers et al., 2006) in this particular context. Specifically, if identity management processes were relevant in this context, then we would expect to find that group identification predicted social creativity, and that group identification itself was predicted by this socio-structural triad.

Study 2 Method Participants
Participants were 133 Indonesian ethnic minority group members (57 male, 74 female, 2 did not report a gender; mean age = 32.53, SD = 12.89, range 17-88 years), who were asked by the second author to participate in the study at various locations in Jakarta. The second author collected the data via hard copy questionnaires, and the third author entered the data in SPSS. Throughout the whole study, anonymity was guaranteed, participation was on a voluntary basis, and participants indicated informed consent.
Participants came from twenty-six ethnic minority groups in Indonesia, with 14.8% from Ambon and 20.0% from Flores. The questionnaire was similar to the one used in Study 1, consisting again of two sections in a booklet in both Indonesian and English, and again starting off with the sentence: "The purpose of this survey is to examine individuals' sense of ethnic identity and its role in their life". The first part included demographic information such as gender, age and ethnicity; the second section included a set of questionnaire items tapping the key variables under study. Responses were scored on 7-point Likert-type scales ranging from strongly disagree (1) to strongly agree (7). The only difference with the Study 1 questionnaire was that we dropped some measures and added new ones (see the Measures section below). We note that the variables we used in both studies and for which we have not yet listed all items can be found in Appendix 1.

Procedure and Power
The sample was recruited among individuals attending church and mosque services and individuals from the second authors' work, university and neighbourhood in Jakarta. As such, the sample included not just employees of a company, as in Study 1, but also teachers, civil servants, students, housewives, and unemployed people. If individuals agreed to participate, the second author set a schedule for data collection, with some participants returning the questionnaires the same day, and some requesting to complete it at home and returning it to the second author at a later time. From the 160 questionnaires we distributed, 133 participants returned the questionnaire. One participant showed clear response tendencies and therefore this data was not analyzed further, which left us with a sample of 132 participants. Compared to Study 1, this provided us with stronger and sufficient statistical power to detect a positive medium-sized effect (r = .30) of group identification on collective action, as suggested by Van Zomeren et al.'s (2008) meta-analysis. Specifically, realized power to detect such an effect (Faul et al., 2009) with 132 participants and an alpha of .05 in a one-tailed test was .97.

Measures
The questionnaire was similar to the Study 1 questionnaire, with a few exceptions. For group identification, we decided to drop Phinney's (1992) ethnic identity measure and Valk and Karu's (2001) ethnic pride measure because of the converging findings across identification indices in Study 1. We thus focused on the Leach et al. (2008) self-investment and self-definition measures, applied in this case to ethnic group membership, and the single-item measure of group identification (Postmes et al., 2013). As in Study 1, principal axis factoring with oblique rotation on the self-investment and self-definition items suggested three factors with an eigenvalue > 1 underlying the fifteen items, explaining 62.14% of the variance. Inspection of the pattern and structure matrices suggested that the first and second factors were in line with the self-investment and self-definition dimensions, with two remaining items forming the third factor. Because these were self-investment items as well, as in Study 1 we included them in this factor and thus calculated these two scales. The reliability of the self-investment and self-definition scales was very good (Cronbach's alphas of .92 and .87, respectively). As can be seen in Appendix 3, and as one could expect, correlations between the three different identification indices were all positive and significant. Moreover, means were above the midpoint of the scale, suggesting that ethnic group identity was psychologically meaningful and relevant to participants.
Collective action tendencies were measured with the same two items as used in Study 1, which correlated positively with one another (r = .41, p < .001) and were therefore combined into one scale. In line with Study 1, the mean level of collective action was fairly low (see Appendix 3) v .
As in Study 1, we measured seeking shelter in social relationships with the 12-item Multidimensional Scale of Perceived Social Support (Zimet et al., 1988). Principal axis factoring with oblique rotation suggested three factors with an eigenvalue > 1 underlying these twelve items, explaining 61.12% of the variance. Inspection of the pattern and structure matrices suggested that the three factors corresponded with the target of social support (a special person, family, or friends). However, follow-up analyses revealed correlations between the three scales of > .51 and few differences in their correlations with other variables. This led us, as in Study 1, to combine all items into one scale. The reliability of the 12-item scale was very good (Cronbach's alpha = .90) and the mean level of seeking shelter in social relationships was fairly high (see Appendix 3).
Also as in Study 1, we measured personal well-being through the 5-item Satisfaction with Life Scale questionnaire (Pavot & Diener, 1993). Principal axis factoring with oblique rotation suggested one factor with an eigenvalue > 1 underlying these items, explaining 58.96% of the variance. The reliability of the scale was very good (Cronbach's alpha = .86) and the mean level was fairly high (see Appendix 3). We also again measured perceived discrimination with the Detroit Area Survey (DAS) Discrimination scale (Perez et al., 2009) that consisted of eight items and the ninth item we added to it in Study 1. Principal axis factoring with oblique rotation suggested one factor with an eigenvalue > 1 underlying these nine items, explaining 56.02% of the variance. The reliability of the scale was very good (Cronbach's alpha = .92) and the mean level was fairly low (see Appendix 3). Therefore, testing whether higher identifiers with the group would endorse moral/warmth stereotypes of their group more strongly than competence stereotypes reflects a test of whether they engage in social creativity (which allows them to compete with higher-status groups on alternative status dimensions). Principal axis factoring with oblique rotation suggested two factors with an eigenvalue > 1 underlying these twelve items, explaining 64.10% of the variance. Inspection of the pattern and structure matrices suggested that the two factors were in line with the morality/warmth and competence dimensions, although two items (trust and honest) loaded on the competence factor and one item (agentic) loaded on the morality/warmth factor. For theoretical reasons, we included those items in the conceptually fitting scales. The reliability of the morality/warmth and competence scales was very good (Cronbach's alphas of .92 and .85, respectively) vi .
Second, we included measures of three socio-structural variables that social identity theory predicts to shape individuals' group identification and identity management strategies (e.g., Mummendey et al., 1999). Specifically, we measured status stability with two items (r = .42, p < .001; i.e., "In general, I think that there is little hope that my ethnic group will achieve a better position in Indonesian society in the near future"; "In general, I think that the ethnic group hierarchy in Indonesia is quite stable"), impermeability of group boundaries with two items (r = .61, p < .001; i.e., "In general, I think that it is difficult, as part of my ethnic group, to become part of the majority ethnic group in Indonesia"; "In general, I think there are closed boundaries between the ethnic groups in Indonesia."), and status legitimacy with two items that correlated only .26 (p < .01; i.e., "In general, I think the ethnic group hierarchy in Indonesia is quite legitimate", and "In general, I think that it is unfair that my ethnic group does not have a better position in Indonesian society."). Given this low correlation, we decided to use the item that had the strongest face validity in our judgment, which was the first item (also given the use of a double negative in the second item).

Results
As in Study 1, our analytic strategy was to first inspect the key correlations between the different indices of ethnic group identification and collective action tendencies. Second, through regression analyses we tested whether individuals coped with low group status through seeking shelter in social relationships (again abbreviated for ease of reference below to "social relationships"), as indicated by social relationships positively predicting well-being.

Replicating a Rare Null Correlation Between Group Identification and Collective Action?
The results can be seen in Appendix 3. First, replicating Study 1, none of the three ethnic minority group identification indices correlated significantly with collective action tendencies (ranging from r = -.08 for self-investment to .11 for the single-item measure and .12 for self-definition). Thus, we replicated the rare null effect of group identification on collective action tendencies, suggesting once more that collective action was not a viable way to cope with low group status in this context.
As in Study 1, social relationships were positively and significantly related to ethnic group identification. As can be seen in Appendix 3, this was the case for self-investment (with r = .23, p = .01), although slightly weaker so for self-definition and the single-item measure (with rs of .16 and .17, ps < .08). Thus, stronger social relationships related to stronger identification with one's ethnic group (which was already found to be unrelated to collective action tendencies; and, corroborating this notion, we also found a null correlation between social relationships and collective action; r = .07, p = .42). All these findings replicate Study 1.

Identity Management and/or Relationship Regulation?
We then tested whether individuals coped with their low status group membership through identity management and/or relationship regulation processes, like in Study 1. To do so, we conducted regression analyses to predict collective action tendencies, well-being, and, again more exploratively, perceived discrimination. Furthermore, to test whether identity management processes were at all likely in this context, we tested whether group identification positively predicted social creativity and whether the socio-structural triad positively predicted group identification (as in Study 1, all reported findings below refer to self-investment).
We used regression analysis to predict individuals' collective action tendencies from their group identification, social relationships, and perceived discrimination. The regression model was statistically significant, F = 3.02, p = .03, but this was only because perceived discrimination predicted collective action tendencies (β = .23, p = .01).
We then predicted individuals' well-being from their group identification, social relationships, and perceived discrimination. The regression model was significant, F = 12.02, p < .001, and inspection of the parameter estimates showed that, as in Study 1, only social relationships predicted well-being significantly (β = .46, p < .001; for group identification, β = .10, p = .22; for perceived discrimination, β = .00, p = .98). This replicates support for the idea that social relationships serve to protect well-being in this context.
We subsequently predicted individuals' perceived discrimination from their group identification and social relationships. The regression model was not significant, F = 0.79, p = .45. Thus, although social relationships did not predict weaker perceived discrimination (as we found in Study 1), this variable uniquely predicted stronger wellbeing (replicating Study 1).
Furthermore, if social creativity reflects the strategy by which individuals managed their group identity in this context, then we would expect group identification to predict it. In terms of our measure, this would mean that group identification predicted the moral/warmth ratings, but not (or at least less so) the competence ratings. We therefore first predicted individuals' ratings on the moral/warmth scale from their group identification, social relationships, and perceived discrimination. of the parameter estimates showed that both social relationships and group identification predicted moral/warmth significantly (β = .39 and .32, respectively; both ps < .001); perceived discrimination showed a significant negative effect (β = -.16, p = .05).
However, when we ran the same analysis on individuals' ratings on the competence scale, the regression model was significant, F = 8.79, p < .001, but inspection of the parameter estimates showed that again both social relationships and group identification predicted competence significantly (β = .30 and .26, respectively; both ps < .01); perceived discrimination did not show a significant effect (β = .02, p = .79). Thus, inconsistent with a social creativity account, the findings of both analyses were very similar.
To more specifically test for the social creativity strategy, we compared whether the correlation between group identification and morality/warmth (r = .45) was significantly stronger than the correlation between group identification and competence (r = .38). Using Lee and Preacher's (2013) test for the calculation of the difference between two dependent correlations with one variable in common, we found that this difference was not significant (z = 1.34, p = .18). This does not support the possibility of social creativity in this context. Instead, these findings indicate that higher identifiers with the group claimed both stronger competence and morality/warmth, which highlights the salience and relevance of group identity in this question.
Finally, we tested whether the three theorized socio-structural antecedents of group identification actually predicted group identification in this context (Mummendey et al., 1999). If identity management processes were at work, then, all else being equal, increased status illegitimacy and instability should be associated with increased group identification (as they signal hope and scope for social change). Thus, we predicted group identification from status stability, status legitimacy and impermeability of group boundaries. However, the regression model was not significant, F = 1.60, p = .19, with all three parameter estimates being non-significant.
Intriguingly, when we exploratively predicted social relationships from the same socio-structural variables, the regression model was significant, F = 6.54, p < .001, and inspection of the parameter estimates showed that status stability predicted social relationships significantly (β = .45, p < .001; for impermeability, β = -.12, p = .16; for status legitimacy, β = -.15, p = .25). Thus, stronger status stability, which is a key feature of the Indonesian context of ethnic hierarchy and associated with a low hope and scope for social change, positively predicted seeking shelter in relationships. This is consistent with the new relational form of coping we propose and the protective function we believe it serves.
Finally, we exploratively predicted perceived discrimination from the three socio-structural variables. The regression model was significant, F = 5.43, p < .01, and inspection of the parameter estimates showed that only impermeability predicted perceived discrimination significantly (β = .31, p < .001; for stability, β = -.18, p = .16; for status legitimacy, β = .13, p = .33). Thus, the perceived impermeability of group boundaries, which is another key feature of the Indonesian context of ethnic hierarchy, was positively related to perceiving discrimination.

Discussion
The Study 2 findings replicated and extended the Study 1 findings with the benefits of increased statistical power and a more diverse sample. Specifically, we replicated the rare null correlation between group identification and collective action in this context, and rather strong support for a relational form of coping that revolves around seeking shelter in social relationships and serves to protect well-being. As such, it seems that coping with low group status in this context revolved around relationship regulation rather than around identity management processes.
Indeed, Study 2 extended the Study 1 findings by including two theoretically relevant measures that enabled tests for indications of identity management processes. First, we found little evidence for social creativity in this study vii ; and second, none of the three socio-structural variables theorized to predict group identification actually predicted group identification in this context. By contrast, and in line with a relational coping account, status stability positively predicted seeking shelter in social relationships, which was positively related to well-being. These findings offer a strong pointer to the protective function of this relational form of coping with low group status in this context.
Interestingly, most of the findings for perceived discrimination differed between the studies. Although we used the same measure as in Study 1 (where it did not correlate with almost any other measure), in Study 2 it significantly and positively predicted collective action, and was in turn significantly and positively predicted by the impermeability of group boundaries. We discuss potential reasons for this inconsistent pattern in the General Discussion.

General Discussion
The current set of studies among Indonesian ethnic minority group members had two aims. First, we aimed to replicate a rare null correlation between group identification and collective action (for a meta-analysis see Van  in a cultural context in which group identities and stable power and status relations between groups would be valued. Indeed, both studies showed that ethnic minority group identification did not positively predict collective action intentions. Although it is always difficult to 'prove' a null effect, we can be quite confident that in neither study the typical medium-sized positive effect of group identification on collective action was obtained.
Thus, in the purposefully selected cultural context in which the two studies took place, collective action did not appear to be a viable identity management strategy to cope with low group status. Study 2 further suggested that social creativity did not seem a more viable identity management strategy, and that the theorized socio-structural antecedents of group identification did not predict it, thus casting doubt on whether individuals engaged in any type of identity management in this context.
Our second aim was to test a relational form of coping with low group status based in the notion of seeking shelter in social relationships. This relational form of coping serves to protect well-being (Lazarus, 1991), which is consistent with attachment theory and other work that accords a major motivational role to relationship regulation (Bowlby, 1969;Fiske, 1992;Mikulincer & Shaver, 2007;Van Zomeren, 2016). In support of this approach, in both studies we found that seeking shelter in social relationships uniquely and positively predicted well-being (whereas group identification did not). This is in line with the idea that social relationships provide safe havens in which individuals can seek shelter to protect well-being. Further in line with this approach, Study 2 found that status stability, which implies little hope and scope for social change, positively predicted seeking shelter in social relationships.
In the remainder of this section, we discuss the theoretical and practical implications of our findings, as well as 2008), most studies of collective action are conducted in contexts that are already very much conducive to it (e.g., hope and scope for social change, coupled with impermeable group boundaries; Van Zomeren, 2013; see also Tajfel, 1978). The current studies purposefully selected a context that lacks hope and scope for social change, within which we found that group identification and collective action were virtually unrelated. Nevertheless, group identities were quite salient and relevant, thus suggesting that strong group identification does not have to imply a stronger motivation to engage in collective action. Rather than focusing merely on the notion of group identification, this points to the importance of taking into account the specific change-or stability-oriented content of group identities (Turner-Zwinkels, Van Zomeren, & Postmes, 2015). In a similar vein, our findings suggest the value of focusing on contexts in which collective action is not already normative, likely, or ongoing (Van Zomeren, 2013).
Second, our findings have implications for social identity theory, which includes three socio-structural predictors of group identification and identity management strategies such as collective action and social creativity (Tajfel, 1978; see also Bettencourt et al., 2001;Ellemers, 1993;Mummendey et al., 1999;Scheepers et al., 2006). Although the replicated rare null correlation between group identification and collective action fits to some extent with social identity theory, Study 2 actually suggested little support for the viability of an identity management framework in the current context (as we found that group identification was unrelated to the three socio-structural factors). Instead, across the studies we found rather strong support for a relational form of coping with low group status (and indeed Study 2 found that status stability was positively related to seeking shelter in social relationships). All these findings raise the question whether social identity theory should not include a relational form of coping that protects individuals' well-being under conditions of low hope and scope for social change.
Our position is that it should, particularly because the notion of a relational basis for coping with low group status is not mutually exclusive with a social identity theory account of coping with low group status. Indeed, because the social relationships individuals seek shelter in are typically ingroup members (although this does not exclude the possibility of cross-group relationships), a relational approach is quite compatible with a social identity approach to coping under such conditions (see Easterbrook & Vignoles, 2013;Postmes et al., 2005). For this reason, it is possible that the notion of seeking shelter in social relationships is also relevant to coping with low group status under different socio-structural conditions. For instance, we would not be surprised to find in future research that activists may particularly be in need of social shelter in order to maintain their enduring activism in the face of adversity (and potential repression), which suggests a link between seeking shelter in social relationships and collective action (and thus a relational form of coping among activists). This is a novel and exciting direction for future research.
Third, our findings with a focus on relationship regulation (Rai & Fiske, 2011), which is consistent with notions about individuals' need for a safe haven through their need for attachment (Ainsworth, 1989;Bowlby, 1969;Mikulincer & Shaver, 2007) and about social support as being an absolutely pivotal coping resource with respect to protecting well-being (e.g., Lazarus, 1991). However, seeking shelter in social relationships may also have more agentic effects. Specifically, attachment theory posits that individuals form and maintain social connections to provide safety in times of trouble, but also to explore the world once they experience such safety (Ainsworth, 1989;Bowlby, 1969). As such, we speculate that when social relationships reflect a safe haven, this may motivate individuals to seek social stability; but when they reflect a secure base, this may motivate them to seek social change. Future research can examine this prediction. A more general implication of our findings is that there is considerable potential for stronger theoretical and empirical integration of relational and group perspectives on coping with low group status. Although we observe a trend within and outside of social identity toward a stronger appreciation of the role of social relationships in forming group identities (e.g., Brewer & Chen, 2007;Easterbrook & Vignoles, 2013;Postmes et al., 2005;Swann, Jetten, Gómez, Whitehouse, & Bastian, 2012), much of this work departs from non-relational principles and assumptions. We believe that a much stronger level of theoretical integration is required to do justice to both approaches (see Fiske, 1992;Rai & Fiske, 2011;Van Zomeren, 2014, 2015, 2016. For instance, Van Zomeren (2016) has recently proposed an integration of a number of major theories on the basis of relational principles, including Bowlby's (1969) attachment theory, Fiske's (1992) relational models theory, and Lazarus' (1991) cognitivemotivational-relational theory (about coping). The current findings suggest that there is no reason to focus on either group identity or relationship processes, and that the real question to pursue is how these processes can be integrated and thus understood in conjunction.
Fourth, we note that perceived discrimination was unrelated to most other variables in Study 1, whereas it positively predicted collective action, and was positively predicted by the impermeability of group boundaries, in Study 2.
As such, our findings provide no support for the notion that perceived discrimination increases individuals' group identification (Branscombe et al., 1999), and mixed support for the notion that perceived discrimination predicts collective action . We believe that this may tell us something important about the difficulty of perceiving discrimination as unfair within cultural systems that value collectives and intergroup power and status differences. Indeed, only when individuals can actually see any unfairness in group-based discrimination, such as in Study 2, it can predict collective action . One reason for why this may have been harder for participants in Study 1 than in Study 2 is that it took place in a work environment, in which ethnic group discrimination may have been a particularly sensitive issue, leading individuals to be more uncomfortable and cautious when filling out the discrimination items (e.g., for fear of repercussions). In any event, we certainly do not want to attach much value to these inconsistent findings regarding perceived discrimination.
Fifth, the current findings offer a practical pointer to the power of social relationships in effectively facilitating how individuals cope with low group status to protect well-being. Our findings point to the relational nature of group life and its trials and tribulations (e.g., Fiske, 1992), which may not necessarily be the first aspect of human motivation that springs to mind for many social engineers, politicians, and practitioners of collective action in the western world, who often assume that individuals are first and foremost individuals who respond to instrumental and pragmatic incentives (e.g., free transport to a demonstration; Oegema & Klandermans, 1994). This is important because seeking shelter in social relationships provides individuals with a stronger psychological buffer against threats to their well-being. Given the importance of social relationships for happiness as well as health (e.g., Bowlby, 1969;Holt-Lunstad et al., 2010), this may be a very basic and important benefit for members of low status groups, especially under (but not limited to) conditions that limit chances of individual mobility and social change.
Finally, two clear limitations of the current set of studies is that the correlational data have low internal validity and that it remains unclear whether similar results may be found in comparable contexts viii . However, we believe that our research aims were met even despite these limitations. The first aim was to repeatedly identify a 'black swan' in a world seemingly full of (and almost exclusively focused on) 'white swans' (Popper, 1959) -that is, a rare null correlation between group identification and collective action. The second aim was to acquire a better understanding of how individuals cope under such conditions by revealing the importance of seeking shelter in social relationships.
This is not to say, of course, that follow-up research should not seek to increase the internal validity of our findings Notes i) We test for the possibility of social creativity in Study 2.
ii) Indonesia is a large country with a population of over 230 million people. It includes at least 300 ethnic groups and more than 700 living languages. The Ethnic Javanese number around 42% of the Indonesian population, followed by the Sundanese of around 15%. The remainder of the population comes from a multitude of ethnic minority groups that range in size between 0 and 4% of the population. Thus, the Indonesian population can be described as very diverse and multi-ethnic. Moreover, ethnic minority groups face regular discrimination in society (e.g., Bertrand, 2004; see also http://www.minorityrights.org/4430/indonesia/indonesia-overview.html).
iii) Some may wonder why one item refers to the group and the other to the individual. Although originally devised to capture this differentiation, we found in both studies that participants did not make this distinction (see also Endnote 6), possibly because of the collectivist cultural context in which both studies were embedded. We interpret this as meaning that the items are both about tendencies to engage in social protest. v) To keep the questionnaire as similar as possible between the studies, we retained the exploratory items reported in Endnote 4 except for the two language items. We added two new items ("I would like to do something on my own to improve my personal living conditions" and "I would like to do something with other group members to improve my ethnic group's living conditions"), which, together with the two collective action items used in Study 1, should have differentiated collective from individual action. However, factor analysis suggested otherwise: The Study 1 items formed one factor and the new items formed a different factor, together explaining 46.87% of the variance. Although this new scale showed a positive correlation with self-investment (r = .26, p < .01), it also correlated positively with status stability and legitimacy (rs = .27 and .20, p < .01 and = .03; according to social identity, these correlations should have been negative). Furthermore, a regression analysis with group identification (self-investment), social relationships and perceived discrimination showed that only social relationships significantly predicted this "improvement" scale, which therefore makes it hard to interpret as collective action.
vi) Analyzing these items in different ways did not change the pattern of findings. Therefore we settled for creating the scales as they conceptually should be.
vii) It is true that there may be many alternative status dimensions and therefore there may be many ways to engage in social creativity. This is precisely why we decided to focus on the general dimensions of competence and morality/warmth, on which we found little evidence for social creativity. Together with the failure of the socio-structural variables to predict group identification and the rather strong support for the relational form of coping, our interpretation is that identity management strategies, if any, did not appear dominant in coping with low group status in this context. viii) Although the current research specifically focused on ethnic identity, some may argue that class identity may also have been relevant to participants. Indeed, the questionnaire in general and for example the identity and collective action questionnaire items were focused on the ethnic minority group; furthermore, ethnic group identification seemed quite relevant to individuals, as judged by the mean scores in both studies, which fits with observations of Indonesian mainstream culture as strongly collectivist (defined as whether individuals are interdependent on others and loyal to their in-groups) and high on power distance (defined as whether those low in power legitimize, accept, and respect intergroup power relations). We further note that although the Study 2 sample had a much more diverse sample, our key findings replicated. This suggests that class identity is unlikely to explain our findings around coping with the low status of one's ethnic minority group.

Funding
The authors have no funding to report.