SOCIAL IDENTITY COMPLEXITY AND INCLUSIVENESS AS PREDICTORS OF INTERGROUP EMOTIONS2

As we belong to a host of groups, we have a multitude of social identities that are interdependent. Social identity complexity refers to the degree of overlap between cross-cutting group memberships while social identity inclusiveness to the range of people a person identifies with through shared group membership. In this paper, we explored the relationship between the complexity and inclusiveness of social identity, and feelings toward ethnic/religious in-groups, as well as feelings toward outgroup members. The research was conducted in two cities in Serbia: Belgrade and Novi Pazar (N = 178; average age 23), allowing for comparisons between young Serbs and Bosniaks, who belonged to the groups with a recent history of conflict. We found that social identity complexity was unrelated, whilst social identity inclusiveness was systematically and positively related to more positive feelings toward religious and ethnic out-groups. This effect was significant across different ethnic groups (Serbs and Bosniaks) and local contexts (Belgrade and Novi Pazar). Both social identity complexity and inclusiveness were unrelated to emotions toward members of the own group. We related these results with the existing data on the social identity structure and intergroup relations, and discussed the importance of inclusive identities for building tolerant societies.

Most people have multiple social identities, i.e. aspects of self-perception that derive from belonging to different social groups (Tajfel & Turner, 1986).Feelings originating from social identifications differ from personal emotions and can be strong motivators of intergroup behaviors (Mackie, Devos, & Smith, 2000;Mackie, Smith, & Ray, 2008).Consistent with Social Identity Theory (Tajfel & Turner, 1986), a social group membership influences intergroup emotions, because the desire for positive social identity makes people evaluate one's group more favorably in comparison to other groups.According to Social Dominance Theory, this in-group favoritism and biases toward out-groups are more likely to be expressed if a group one belongs to is dominant rather than subordinate (Pratto, Korchmaros, & Hegarty, 2007;Pratto, Sidanius, & Levin, 2006;Sidanius & Pratto, 1999).
Until recently, social identities and intergroup emotions were studied exclusively in the context of a single in-group-out-group categorization.In the last several years, more attention has been drawn to investigating the constellation of person's diverse identities, with the idea that relations between identities, rather than identities per se, might be more relevant for intergroup relations and biases.Two prominent constructs addressing crosscutting identities have emerged in the field: Social Identity Complexity -SIC (Roccas & Brewer, 2002) and Social Identity Inclusiveness -SII (van Dommelen, 2011(van Dommelen, , 2014)).
SIC refers to the mental representation of multiple in-group composition and boundaries (Brewer, Gonsalkorale, & van Dommelen, 2013).One way of operationalizing SIC is asking individuals to assess the perceived overlap between their different multiple in-groups (e.g.asking an Orthodox Serb, "When you think about people who are Serbs, how many are Orthodox?").People low in SIC should perceive the groups to which they belong as overlapping and similar (e.g. that all Serbs are Orthodox) resulting in a single convergent social identity, while those who have high SIC will tend to see their in-groups as nonoverlapping and dissimilar (e.g.there are Serbs who are not Orthodox), resulting in more loose boundaries among different group memberships.
SII refers to how inclusively or exclusively one defines their in-group from the combination of different potential social identities (van Dommelen, 2011(van Dommelen, , 2014;;van Dommelen, Schmid, Hewstone, Gonsalkorale, & Brewer, 2015).It is commonly explained as a range of people with a different combination of group memberships which a person identifies with.Those who have looser criteria for identifying with others have an inclusive and a highly diverse in-group.Those who are low in SII are on the other hand more likely to only relate to others who satisfy more rigid membership criteria, resulting in a single highly exclusive in-group representation constituted of others that are very similar to oneself.Such exclusive social identity is likely to be accompanied by the perception that any individual who is an out-group member on one dimension is also an out-group on all the others.
Understanding the constellation of multiple social identities is important because they determine both how we see ourselves and our in-groups, and, perhaps less intuitive, how we perceive and treat others.This process is driven by balancing between a need for similarity (feeling similar to other in-group members) and a need for distinctiveness (having an in-group that is distinctive from other groups).When the distinctiveness of one's group is threatened, people protect their identity by reducing its complexity and inclusiveness (Schmid, Hewstone, Tausch, Cairns, & Hughes, 2009).This in turn leads to more bias, less tolerance, and more negative attitudes toward excluded outgroups.However, when identity needs are met, people are more likely to see former out-group members as in-group members, and intergroup emotions and consequentially relations may improve.To this end, several studies also show that tolerance toward out-groups is higher for individuals high in SIC (e.g.Brewer & Pierce, 2005;Roccas & Brewer, 2002).
The main focus of the present research was to test and compare the power of SIC and SII in predicting intergroup emotions.We explored the constellations of three types of social identities: (1) ethnic (2) religious and (3) national among young Serbian citizens living in ethnically homogeneous (Belgrade) and heterogeneous (Novi Pazar) environments, and measured their feelings toward ethnic and religious in-group and out-group members.Both SIC and SII were proven to predict more positive attitudes towards out-groups and support for multiculturalism in previous research (Brewer & Pierce, 2005;Miller, Brewer, & Arbuckle, 2009;van Dommelen et al., 2015).However, as the former reflects the perceived relation between in-group identities, and the latter reflects readiness to include others in one's own group, they could be differently related to feelings towards in-group and out-group members.
Insights from the Balkan region, with its history of violent ethnic conflicts, could significantly help to generalize existing findings that stem either from other conflict regions (e.g.Israel), or from non-conflict regions (e.g.Belgium, Netherlands, Australia).We explored the ethnic and religious identities, as well as the possible overarching national identities.As the previous conflicts were drawn from the ethnic and religious divisions, these identifications have been still predominant in the region.Affiliation to an ethnic and religious group still provides a sense of meaning and importance for individuals, and has a tangible impact on everyday life.Our findings are of importance for designing policies aimed at improving intergroup relations in the region.

Participants and procedure
The sample included a total of 178 young people (90 males) of Bosniak (n = 94) or Serbian ethnicity (n = 84) from Belgrade (n = 78) and Novi Pazar (n = 100) (see Table 1 for a detailed structure of the sample).The participants ranged in age from 18 to 30 (M age = 23, SD = 2.81).They were recruited by the snowballing method, following the given quotas3 .The participation was completely voluntary.Participants were first asked to state their ethnic, national and religious affiliation.Participants whose self-declared identities differed from the expected ones (i.e. that those who declared themselves not to be religious) were removed from further analyses.
Participants filled the online questionnaire in prearranged facilities that had computers with Internet access.They were tested in groups of three to five.Upon providing basic demography data, they answered the SIC questions and completed the SII card sorting task, followed by the feeling thermometer.In the end, they were thanked and individually debriefed.

Instruments
Social Identity Complexity -SIC (Miller et al., 2009;Roccas & Brewer, 2002).We measured SIC with the standa rd set of questions, which we adapted to encompass the three main identities we focused on (i.e.ethnic, religious and national identity).The scale consisted of questions asking participants to assess perceived overlap between social groups they belonged to (e.g."When you think about people who are [Serbs], how many are [Orthodox]?").We asked respondents to assess this overlap between all possible pairs of their particular set of in-groups (national, ethnic and religious), a total of 6 possible combinations4 .The respondents answered the questions using a scale from 1 = none to 11 = all.We introduced the task with the following instruction that we adopted from Brewer et al. (2013) and provided them with two practice trials: "Now we are going to ask you some questions about how memberships in different social categories are related.Group memberships can be associated in various ways.For exam ple, "mothers" are all members of the cate gory "women", but only some of the people who are women are mothers.We are interested in your estimates of how many people in group X are also mem bers of group Y, rated on a scale from 1 (none) to 11 (all).In each case, we are asking for your subjective estimates based on YOUR OWN impres sions of the social groups.Please remember that we are interested in the groups as YOU think about them, not necessarily the actual numbers.There are no right or wrong answers." We calculated SIC as an average rating of overlap between six different combinations of in-groups.To ensure that higher SIC index indicated higher complexity and less overlap, we subtracted this score from the maximum average number of overlaps (i.e.11).The average SIC index was 4.21 (SD = 1.83).

Social Identity Inclusiveness -SII (van Dommelen et al., 2015).
To measure SII, we adapted the Triple Crossed-Categorization Task (TCCT) from van Dommelen et al. (2015) to match the three main identities we focused on (i.e.ethnic, religious and national identities).At the beginning of the TCCT task, we gave participants the following instruction (adapted from Maloku-Berdyna, 2012): "We will now present you a series of cards.On each card, you will see a picture/figure of a shadowed profile, together with the name of the person, his ethnicity, citizenship and religious affiliation.Your task is to take a look at these cards and answer by clicking the button US or THEM at the left and right sides of the screen.If you feel that this person is someone who belongs to your group or is US, you click button US.In contrast, if you feel that this person does NOT belong to your group, click the THEM button on the screen.We remind you that there are no correct or wrong answers.What matters is that you evaluate the Identity Cards based on how YOU THINK or FEEL that they belong to.It also does not matter how many cards you click as US or NOT US.The time for sorting the cards is not limited, so please take as much time as you need when making your choice.These persons and their characteristics are taken (with their permission) from the National statistical office5 ." After the instructions we presented a set of identity cards to participants and asked them (as described in the instruction) to assess whether the person from the identity card presented on the screen belongs to US or THEM.The identity cards included a shadowed profile (head and shoulders) together with the name and surname of the person, his/her ethnicity, citizenship and religious affiliation (see Figure 1 for an example of a card).In addition to the explicitly written national, ethnic and religious cues on the identity cards, the names on each card were typical ethnic names providing cues about the ethnic and religious characteristics of the target on the card.There were two sets of cards: male and female; the gender of persons on the identity cards matched the respondent's gender.In total, each respondent sorted 24 cards to categories of US or THEM.We varied the overlap in group memberships for each of the characteristics to create targets on the cards that shared three, two, one or none of the characteristics with the respondent.Six targets on the cards represented triple in-groupers -persons who shared all three characteristics with the respondents (they had the same ethnicity and religion and were citizens of the same country), six double in-groupers, who shared two characteristics (e.g.ethnicity and religion, but were from a different country), six single in-groupers who shared only one characteristic (e.g. the same ethnicity, but different religion and nationality).Finally, there were six cards with triple out-groupers that shared no characteristics with the respondent.The cards were displayed to the participants in random order.SII score was operationalized as the number of targets that were categorized as in-group members in the card-sorting task.Higher values on SII index indicated greater inclusiveness of the social identity.The average SII index was 13.16 (SD = 4.98).
Feeling thermometer -FT (Converse & Presser, 1986).Participants assessed their feelings towards two religious (Muslim and Orthodox Christian) and two ethnic (Bosniaks and Serbs) groups by sliding a bar on the scale from 0 to 100 degrees Celsius according to how warm or how cold they felt toward the specific group.Higher number indicated positive feelings, and lower number indicated negative feelings.The value of 50 meant that feelings were neither positive nor negative.
Social identifications.Participants rated the extent to which the affiliation to national, ethnic and religious group is important to them personally on 5-point Likert-type scales.

Results
We present means and standard deviations for feeling thermometer ratings toward different out-group and in-group members, and for different subsamples in Table 2.We could observe that "warmer" feelings are expectedly expressed towards the members of one's own ethnic and religious group: the participants of Serbian ethnicity expressed more positive feelings toward Orthodox Christians and Serbs than toward Muslims and Bosniaks; the opposite is true for the participants of Bosniak ethnicity.
All the measures (SIC and SII indices and FT measures) deviated from normality and were normalized by using Blom's ranking method.We performed all subsequent analysis on normalized and standardized variables.To explore relationships between social identity complexity and inclusiveness, and feelings toward different in-groups and out-groups, we correlated these variables separately on subsamples of Serbs and Bosniaks (Table 2).The SII was significantly correlated to the feelings toward both religious and ethnic out-groups (Muslims and Bosniaks for Serbs and Orthodox and Serbs for Bosniaks).Serbs who reported more positive feelings toward Muslims and Bosniaks tended to have greater social identity inclusiveness; the same was true for Bosniaks when they rated Orthodox Christians and Serbs.SIC measure was only related to one of the FT variables: Serbian participants with higher complexity expressed less warm feelings toward their ethnic in-group.To directly test the power of SIC and SII to predict intergroup emotions, we conducted eight multiple regression analysis in which we regressed SIC and SII variables on four different FT measures (feelings toward: (1) Orthodox Christians (2) Muslims (3) Serbs and (4) Bosniaks).Before running the analysis we dummy coded two variables: (1) majority/minority status on local, town level (status_local; Serbs/Bosniaks in Belgrade/Novi Pazar) and ( 2) majority/ minority status on global, country level (status_global; i.e.Serbs/Bosniaks).We entered the first variable into the tested models to control for the influence of objective group membership on FT ratings (as opposed to the subjective nature of SIC and SII constructs).To explore and control for the influence of local ethnic/religious context on intergroup feelings, we entered the local minority/ majority status as a second predictor.In addition, we entered the interaction terms between these measures and measures of SIC and SII to estimate the magnitude of their moderating effects on the relationship between SIC/SII and FTs (Table 3 and 4).As expected, the main predictor of intergroup feelings was the ethnic group affiliation -i.e. each group expressed in-group favoritism.More specifically, majority members (i.e.Serbs) had more positive feelings than minority members (i.e.Bosniaks) toward Orthodox Christians and Serbs, while minority members (Bosniaks) had more positive feelings toward Bosniaks and Muslims primenjena psihologija, str.359-378 (Table 2).In addition, feelings toward Orthodox Christians and Serbs were related to the local context -local minority members (i.e.Serbs from Novi Pazar and Bosniaks from Belgrade) expressed warmer feelings toward these groups.Most importantly, however, SIC was not a significant predictor of FTs toward different groups.In contrast to SIC, SII turned to be a significant predictor of FTs (Table 4).However, the effect of SII depended on the global majority/minority status (interaction terms between SII and global majority/minority status were significant).To explore the nature of this interaction, we conducted multiple regression analysis separately for subsamples of the state majority members (i.e.Serbs) and minority members (i.e.Bosniaks).In these, we regressed the SII index and local majority/minority status on four different FT measures (Table 5).In case of Serbs, SII was a significant predictor of feelings toward Bosniaks and Muslims over and above their local majority/minority status, or the city they were living in.The same was true for Bosniaks -SII was a significant predictor of feelings toward Serbs and Orthodox Christians over and above their local majority/minority status.In other words, people with more inclusive identities expressed more positive feelings toward out-groups independently of the ethnic environment or the context they were living in (i.e.Belgrade or Novi Pazar) or the ethnic group they belonged to (i.e.Serbian or Bosniak).To further explore the predictive power of SII for feelings toward outgroups over and above the singular in-group identifications, we conducted a hierarchical regression analysis.The first step was to enter strength of national, ethnic and religious identification as predictors, while SII was entered in the second step.We tested whether introducing SII as predictor would contribute in explaining additional variance, compared with the strength of individual identifications.The findings are presented in Table 6.We could observe that SII explains additional variance in 3 of 4 cases: in predicting feelings of Serbs towards Bosniaks (about 6% of additional variance), as well as in predicting feelings of Bosniaks towards Serbs and Orthodox Christians (10-12% of additional variance).The findings demonstrate added predictive power of SII, particularly in the state minority group.

Discussion
The aim of the present study was to explore the relationship of two relatively new and underinvestigated constructs in social identity research: social identity complexity/social identity inclusiveness with emotions toward ethnic and religious in-group and out-group members.Findings demonstrated that individuals sharing an (objective) membership in the same ethnic, religious and national groups could differ in their subjective representations of these memberships, as reflected in the individual variations in social identity complexity/inclusiveness.These differences were further reflected in their perceptions and feelings toward different ethnic and religious groups.
By including three dominant social identifications -ethnic, religious and national, this study answered the call from the previous research saying that the two latter social categories remained overlooked in many studies on minority identification (van Dommelen et al., 2015).Specific sample design enabled us to compare minorities by their status in the local community, as well as on state level, and to compare ethnically homogeneous and heterogeneous social environments.
More inclusive social identity uniquely predicted more positive feelings toward what was traditionally conceived of as ethnic and religious out-groups6 .This effect replicated across different ethnic groups (Serbs and Bosniaks), and across different local contexts (Belgrade and Novi Pazar).Furthermore, social identity inclusiveness demonstrated an additional predictive power over and above the strength of singular identifications (ethnic, religious and national).
To understand the feelings towards other groups, it is not only important to know how strongly one identifies himself/herself with the in-groups but also how one construes relations between different identities.Social identity inclusiveness has proved more predictive for the state minority group members, suggesting differences in minority and majority group perspectives in multiple identity construal.
On the other hand, social identity complexity was not predictive of feelings toward other groups.These findings indicate that, despite being similar in nature, two multiple identity measures just partially overlap.The measure of SII takes into account whether respondents identify with their in-groups when their memberships do not overlap completely.It seems to be more conative in nature and, therefore, more predictive of intergroup relations.In addition, SII measure is based on more trials, and therefore it is more reliable.It is also less intrusive and less susceptible to socially desirable responding.Importantly, unlike SIC, it is not sensitive to the objective overlap among the groups.
Finally, the constructs are not related to emotions toward in-groups, with one exception: Serbian participants with more complex identities have expressed less warm feelings towards the in-group.Although both SIC and SII reflect perceptions of the in-group structure and composition, thus they are not related to warmer feelings toward the in-group.This finding is in line with the previous research (Brewer & Pierce, 2005;van Dommelen et al., 2015), and it primenjena psihologija, str.359-378 is partly due to the fact that the out-group evaluation is the variable with more individual variations.This issue should be further addressed in future studies.
There are several limitations to the present study and its findings should be replicated by using larger samples, as well as more varied measures of criterion variables.The future research could try to further disentangle between the two constructs and relate them to a larger range of potential outcomes.It is plausible to assume that SIC would be a better predictor of a more cognitive measure of intergroup relations.It would also be interesting to replicate the findings in similar contexts, taking into account both the type of minority (either religious or ethnic), and its size (both on global and local level) (following Laurence, 2015).Other cross-cutting group memberships could also be investigated, e.g. between ethnicity and social class or profession, as well as their possible effects on inter-group perceptions and emotions.
This research aimed to demonstrate that the complexity of the social world around us can be mirrored (at least to a higher extent that usual) in the research design, and that membership in important social groups and their interrelations do make a difference when it comes to embracing others.Our findings suggest that by fostering a more inclusive perception of one's ethnic or religious identity, one also builds foundations for better relations with other groups -more inclusive identities lead to more inclusive societies (Bodenhausen, 2010).Previous research shows that in the region of Western Balkans ethnic and religious identities are conceived of as inherited and largely unchangeable (Milošević Đorđević, 2007;Turjačanin & Majstorović, 2013).Given this fact, the efforts (i.e. in the educational system or media) targeted at raising awareness of cross-cutting identities might prove a more useful approach to reducing inter-group tensions in comparison to a direct promotion of tolerance.We believe that this path should prove to be particularly applicable for young people who are still in the process of structuring their multiple identities.

Figure 1 .
Figure 1.Example of a stimulus card from the Triple Crossed-Categorization Task.

Table 1
Sample structure (a number of participants within each quota)

Table 3
Regression analyses for predicting feeling thermometer measures with social identity complexity and majority-minority status variables

Table 4
Regression analyses for predicting feeling thermometer measures with social identity inclusiveness and majority-minority status variables

Table 5
Regression analyses for predicting feeling thermometer measures with social identity inclusiveness and local majority-minority status on subsamples of Serbs and Bosniaks * p < .05. ** p < .01.

Table 6
Regression analyses for predicting feeling thermometer measures in a hierarchical regression, with strength of individual identification and SII measures Ethnicity/religion/national ID = strength of ethnic, religious and national identification on 5-point scale; SII = social identity inclusiveness.* p < .05. ** p < .01.