The relationship between collective narcissism and group-based moral exclusion: The mediating role of intergroup threat and social distance

In our study, we investigated the relationship between collective narcissism and group-based moral exclusion. Since collective narcissists are motivated to see their group as unique and superior, and tend to show hostility towards outgroups threatening this presumed superiority, we hypothesized that perceived intergroup threat and social distance can mediate the relationship between collective narcissism and group-based moral exclusion. We tested this assumption in two intergroup contexts by investigating the beliefs of members of the Hungarian majority population about Muslim immigrants and Roma people. Our results showed that collective narcissism had a positive indirect effect on group-based moral exclusion in the case of both outgroups. Furthermore, both threat and social distance were significant mediators in the case of Muslim immigrants, but mostly social distance mediated the indirect effect of collective narcissism on moral exclusion of the Roma. These results indicate that collective narcissists tend to rationalize their intergroup hostility by the mechanism of motivated moral exclusion, and to find suitable justifications for doing so.


Group-Based Moral Exclusion
Moral exclusion happens "when individuals or groups are perceived as outside the boundary in which moral values, rules, and considerations of fairness apply" (Opotow, 1990, p. 1). When members of an outgroup are placed beyond this scope of moral concern, they are not considered to be entitled for just and morally appropriate treatment anymore, consequently, neither withdrawal of prosocial behavior, nor implementation of harmful and unjust behavior is viewed as morally wrong. Accordingly, targets within this scope of moral regard are indeed perceived to be entitled to positive treatment and considerations for their well-being, whereas negative treatment is more acceptable towards those who are excluded from this moral circle (Hadarics & Kende, 2018;Lima-Nunes, Pereira, & Correia, 2013;Opotow, 1990Opotow, , 1994Passini & Morselli, 2017).
People have a tendency to deny positive and fair treatment from members of outgroups that they cannot identify with, or perceive them endangering the ingroup's well-being, goals, or cultivated norms and values. Therefore, perceived characteristics of the outgroup (i.e., target perceptions) can contribute to their moral exclusion by justifying their moral disregard (Bandura, 2016;Opotow, 1990). Specifically, previous research has shown that psychological distance from outgroups and perceived intergroup threat are associated with moral exclusion.
When people engage in group-based moral exclusion (i.e., accept immoral behavior toward members of the outgroup), they nevertheless want to maintain a sense of personal moral integrity. On the one hand, this positive moral self-image is maintained if the outgroup is not regarded as worthy of moral concern (Lima- Nunes et al., 2013;Pereira, Vala, & Costa-Lopes, 2010). On the other hand, moral exclusion can be justified by serving the ingroup's goals as it is described by the dual-process model of prejudice (Duckitt, 2001). According to this theory, the source of prejudice is either the perceived threat to the ingroup's security or group-based dominance. Evidence shows that group-based moral exclusion is associated with both of these motivational goals, that is, members of outgroup can be morally excluded either because they represent a threat to the ingroup's sense of security and conventional values, and also to maintain existing social hierarchies Passini & Morselli, 2017). This connection suggests that the justification of immoral treatment and aggression toward outgroups can satisfy both of these social needs.

A Potential Link Between Collective Narcissism and Moral Exclusion
As we have seen, collective narcissists may be especially sensitive to intergroup threat and may be motivated to maintain intergroup distance. Therefore, we can assume that moral exclusion can serve as a strategic tool for collective narcissists to justify intergroup aggression. Since perceived social distance and intergroup threat are of the most important bases of group-based moral exclusion, it is reasonable to assume that the relationship between collective narcissism and moral exclusion is mediated by intergroup threat and social distance, two characteristics being very likely to be perceived by collective narcissists.
Our hypothesis about this mediation process is further supported by the assumed directions of the relationships between these variables both in the moral exclusion and the collective narcissism literature. In the former, perceived outgroup characteristics are often called as "precursors" (Hafer & Olson, 2003) or "antecedents" (Lima-Nunes et al., 2013;Olson et al., 2011;Opotow, 1990) of moral exclusion, and the same direction from threat and social distance to moral exclusion is also supported by experimental studies that found a stronger tendency for moral exclusion after manipulating these perceived outgroup characteristics (Brockner, 1990;Olson et al., 2011;Opotow, 1994). In their longitudinal study, Leite and colleagues (2019) also found that perceived threat and ingroup supremacy predicted moral exclusion over time, but not the other way around.
Meanwhile, in the collective narcissism literature, variables related to intergroup threat and social distance are much more considered as consequences of collective narcissism. In her extensive review, Cichocka (2016) mentions intergroup threat as a "consequence" of collective narcissism which induces an "exaggerated perception of threat to the in-group" (p. 291). Accordingly, both Cichocka and colleagues (2016), and Golec de Zavala and colleagues (2017) considered perceived intergroup threat as a result of collective narcissism, and as such, a mediator between collective narcissism and other intergroup beliefs. There is also indicating evidence about the direction of the relationship between collective narcissism and social distance. In their 4-wave longitudinal study, Golec de Zavala et al. (2020, Study 6) found that the longitudinal relationships from collective narcissism to social distance were always significant, but the reversed relationships were weaker and not always significant.

The Study
We investigated the mediating role of perceived intergroup threat and social distance between collective narcissism and group-based moral exclusion. These relationships were tested in two intergroup contexts: between ethnic majority Hungarians and Muslim immigrants on the one hand, and Roma people on the other. These two intergroup contexts can be characterized by important similarities and differences. Roma people have lived in Hungary for centuries, they are the largest ethnic minority group in the country, making up 5 to 8% of the population (Pásztor & Pénzes, 2013). Nevertheless, they continue to be severely discriminated and marginalized. Roma people are mainly rejected based on negative stereotypes about laziness and criminality (Enyedi, Fábián, & Sik, 2004), and the perceived lack of efforts they make about integration and assimilation (Cooper, 2002). In contrast, the number of Muslim immigrants living in the country remains extremely low. Nevertheless, immigration from Muslim countries have become the main topic of right-wing populist political propaganda since 2015, mainly presenting Muslim immigrants as a threat to security by equating immigration with terrorism (Simonovits & Bernát, 2016). Despite the differences, prejudice against the Roma and Muslim immigrants (previously measured as anti-Arab and general xenophobia) are and have been similarly high, shared by the majority of the population, and expressed in blatant and explicit ways (Bruneau, Kteily, & Laustsen, 2018;Kende, Hadarics, & Lášticová, 2017). According to a recent opinion poll by the Pew Research Center (2016) 64% of the majority Hungarians maintained unfavorable views about the Roma, and 72% did the same about Muslims, which were among the highest rates within the 10 investigated countries. Another recent Eurobarometer (2019) study showed that 22% of Hungarians would feel uncomfortable if they had a Roma colleague, and 34% in case of a Muslim colleague. Among the respondents, 44% answered the same if one of their children had a Roma partner, and the rate was even higher (49%) in case of a Muslim person. These rates were among the highest ones within the EU.

Method Sample
Our sample consisted of 1080 participants from an online participant pool using a multiple-step, proportionally stratified, probabilistic sampling method. Due to this method our sample was demographically similar to the Hungarian population in terms of age, gender, type of residence, and level of education. The recruitment was carried out by a professional public opinion company. Our survey was part of an omnibus survey where other research questions were tested parallelly.
i We report all measures and data exclusions related to the variables of the research question.
All participants reported Hungarian as their nationality, no participants indicated that they were Muslim, ten respondents reported that they were Roma, and 11 participants did not wish to report their ethnicity. These 21 participants were removed from the analysis, since our main focus was on the beliefs of the Hungarian majority population.
Our respondents did not receive any material reward for their participation, and they completed the anonymous questionnaire after accepting an informed consent form on the first page of the questionnaire. The sample was randomly split: half of the respondents completed the questionnaire related to the Roma (N = 512; 52.1% women; age: 46.60 ± 15.39), and the other half related to Muslim immigrants (N = 547; 52.3% women; age: 47.14 ± 15.49).
In terms of education level of the first group, 33% had a higher education degree, 44.8% had secondary education, and 22.2% lower than secondary education. In the second group, 35.1% had a higher education degree, 43.9% had secondary education, and 21% lower than secondary education. ii

Measures Social Distance
We used two standard items from the Eurobarometer (2015, 2019) survey, where respondents indicated how comfortable they would feel if they had a Roma/Muslim immigrant colleague or one of their children had a Roma/Muslim immigrant partner (1 = not at all comfortable; 10 = totally comfortable). Higher scores on this scale indicated a lower level of social distance, however as our predictions were phrased about distance, we reversed the scores of this scale so that higher means indicated larger social distance.

Intergroup Threat
Threat was measured by 3 items adapted from Kteily, Bruneau, Waytz, and Cotterill (2015). ("Roma people / Muslim immigrants pose a health threat to Hungarians", "The cultural values of Roma people / Muslim immigrants are in opposition with Hungarian values", "Roma people / Muslim immigrants endanger the physical safety of Hungarians."; 1 = strongly disagree, 7 = strongly agree).

Moral Exclusion
We measured moral exclusion of the Roma and Muslim immigrants with Opotow's (1993) Moral Exclusion Scale to Muslim immigrants / Roma people too"; "I am willing to make personal sacrifices to help or foster Muslim immigrants' / Roma people's well-being"; "I am willing to allocate a share of community resources to Muslim immigrants / Roma people"; 1 = strongly disagree, 7 = strongly agree). Higher scores on this scale indicated a higher level of moral inclusion, however as our predictions were phrased about moral exclusion, we reversed the scores of this scale so that higher means indicated higher level of moral exclusion.

Collective Narcissism
We measured collective narcissism with the 5-item version of the Collective Narcissism Scale with reference to the national ingroup as "Hungarians" (Golec de Zavala, Cichocka, & Bilewicz, 2013). As this measure was not available in Hungarian, we applied a translation-backtranslation technique (Brislin, 1980) to obtain it. Answers were measured on a 7-point scale in this case too (1 = strongly disagree, 7 = strongly agree). iii

Analytic Strategy
First, we checked the adequacy of our measures by setting up a measurement model in the case of both groups.
These measurement models were estimated using confirmatory factor analysis. To test the mediating role of perceived intergroup threat and social distance between collective narcissism and moral foundations we applied structural equation modeling (SEM) with the SPSS AMOS 25.0 software (Arbuckle, 2017). The amount of missing data was low in both samples (0.77% in Muslim group and 2.97% in the Roma group), and missing values were replaced with imputed values calculated with the regression method (Byrne, 2010).

Hypothesis Testing
We set up a structural model for the two subsamples separately to test how collective narcissism is related to moral exclusion, and to what extent this relationship is mediated by social distance and intergroup threat. All these variables were set up as latent variables, and we controlled both models with the effects of respondents' gender, age, education level, subjective financial status, and left-right political preference. Political preference was incorporated in the models because both collective narcissism and negative beliefs about culturally different outgroups are more closely associated with the political-ideological right Golec de Zavala et al., 2009). Our goal was to avoid a potential spurious correlation between collective narcissism and outgroup attitudes that is only due to their mutual connection to political preferences. This was especially important in the Hungarian context, since anti-Roma rhetoric and anti-Muslim rhetoric had been an integral part of Hungarian political communication in the years preceding this study (Glózer, 2013;Vidra, 2017). Control variables were built into the models as predictors for each latent variable. iv Results of the hypothesized models testing (see Figure 1) revealed that both models showed an acceptable fit   To reveal the extent to which the relationships between collective narcissism and moral exclusion is mediated by social distance and threat, a series of mediation analyses was conducted with the bootstrapping technique as suggested by Macho and Ledermann (2011), where we requested 95% confidence intervals using 5000 resamples.
We found that collective narcissism had an indirect effect on moral exclusion mediated by social distance in both models ( We also tested the equivalence of each pathway between the main variables across the two models. For this purpose, we set up a multigroup model, and constrained the pathways one by one to be equal across the two groups. After that, based on Δχ 2 -tests and AIC fit values, we checked whether the fit indices of these constrained models were significantly worse than the fit of the original non-constrained multigroup model (fit of the non-constrained multigroup model: χ 2 = 558.83; df = 208; CFI = .964; RMSEA = .040; SRMR = .034; AIC = 898.83). The results of these analyses can be seen in Table 2. These show that the relationships between collective narcissism on the one hand, and intergroup threat and social distance on the other were significantly stronger in the Muslim group compared to the Roma, just like the pathway from intergroup threat to moral exclusion. The relationship between collective narcissism and moral exclusion turned out to be invariant, just like the pathway from social distance to moral exclusion. Finally, we tested whether it is more reasonable and parsimonious to consider perceived threat and social distance as mediators between collective narcissism and moral exclusion than to treat either collective narcissism or moral exclusion as mediators. We set up three models in both groups for this purpose. In the first pair of models, perceived threat and social distance were mediators between collective narcissism and moral exclusion but the direct relationship between these latter two variables was erased to test the importance of the mediator variables. The fit of these models was only slightly worse than the fit of the original ones reported above. In the second pair of models, collective narcissism was set as mediator between the two perceived outgroup characteristics and moral exclusion, and direct relationships were erased again. These models showed a much worse fit to the data compared to the fit of the original ones, and even to the fit of their counterparts with the outgroup characteristics as mediators (Muslim: χ 711.08; Roma: χ 2 = 534.52; df = 106; CFI = .893; RMSEA = .089; SRMR = .115; AIC = 700.52). Finally, in the third pair of models, we set moral exclusion as a mediator between the two perceived outgroup characteristics and collective narcissism, and direct relationships were erased again. Once again, these models showed worse fit than the models with the outgroup characteristics as mediators (Muslim: χ 2 = 381.74; df = 106; CFI = .953; RMSEA = .069; SRMR = .064; AIC = 547.74; Roma: χ 2 = 325.13; df = 106; CFI = .945; RMSEA = .064; SRMR = .051; AIC = 491.13). These findings support that it is the most parsimonious model to assume that perceived threat and social distance mediate the effect of collective narcissism on moral exclusion.

Discussion
We found that collective narcissism was related to group-based moral exclusion, and perceived intergroup threat and social distance were important mediators in this relationship. We collected evidence from two intergroup contexts, between the Hungarian majority population on the one hand, and the Hungarian Roma and Muslim immigrants on the other. Although some differences could be observed between these two cases, similarities in the results suggest that collective narcissists tend to perceive outgroups in a more negative way what makes their moral exclusion more probable.
It has already been known that collective narcissists tend to treat outgroups with a certain extent of hostility (e.g. Golec de Zavala et al., 2009;Golec de Zavala et al., 2016). Nonetheless, the relationship between collective narcissism and moral exclusion implies that they tend to justify this negative treatment by placing these outgroups without the boundaries of moral regard. Furthermore, we were able to identify mediated effects between collective narcissism and moral exclusion through social distance and intergroup threat, which suggests that these perceived group characteristics enable collective narcissists to rationalize their moral disregard and their subsequent hostile behavior.
Beside the generally similar pattern of relationship between the variables, we also found some differences between the intergroup contexts. Most importantly, three of the four relationships involved in the mediations were stronger in the case of the Muslim model, and both social distance and intergroup threat were important mediators in that model, whereas it was mainly social distance that mediated the indirect effect of collective narcissism on moral exclusion in the Roma model. These differences fit with the dominant stereotypes about these two target groups in the Hungarian contexts. While Roma people are generally treated with hostility on the basis of stereotypes of laziness and petty criminality, and a preference for low contact with them (Enyedi et al., 2004;Kende et al., 2017), Muslim immigrants are mostly talked about as posing a security threat to the nation, and a cultural threat to European way of living and values.
The most important difference between these two intergroup contexts, that can be relevant to our results, is probably their salience within the Hungarian public thought. It is worth highlighting that immigration from Muslim countries has become a central topic within the Hungarian domestic political discourse since the beginning of the European Refugee Crisis started in 2015, and it was also the most emphasized topic by the Hungarian government party (Fidesz) during the parliamentary elections campaign in 2018 (Bíró-Nagy, 2018). It is also worth noting that the media framing on Muslim immigrants is rather negative, and it emphasizes the threat related to this outgroup (Vidra, 2017). The salience of this issue is also shown by the fact that anti-Muslim attitudes were high but stable until 2015, but started an increase related to the political events in the post-2015 period. On the other hand, prej-udice against the Roma has been high but much more stable over time (Eurobarometer, 2015(Eurobarometer, , 2019Simonovits & Bernát, 2016).
To highlight the differences between the two intergroup contexts in terms of salience at the time of data collection, we checked the number of hits for the expressions "roma" and "cigány" for Roma people and "bevándorló", "migráns", and "muszlim" for Muslim immigrants on Google within the period from 1 January to 31 March, when we conducted the data collection. This test estimated 51.600 hits for the Roma, and 503.000 hits for Muslims. This is in accordance with the approach of ideological thinking as motivated social cognition (Jost, Glaser, Kruglanski, & Sulloway, 2003), which argues that any kind of individual motivation can only lead to political preferences that are offered by the political environment. Since immigration from Muslim countries dominates the Hungarian political discourse, Muslim immigrants serve as a more available target for collective narcissists to satisfy their self-protective and self-enhancing needs. Nevertheless, the same relationships could be observed also in the case of the Roma group, although in a weaker form.
The salience and the framing of the issue of Muslim immigration can also offer a possible explanation to our finding that both perceived outgroup characteristics mediated the effect of collective narcissism on moral exclusion in case of the Muslim outgroup, but it was mainly social distance that served the same function in the Roma model. Although previous research on moral exclusion emphasizes the importance of social distance and intergroup threat as antecedents of moral exclusion (Opotow, 1990), this does not necessarily mean that both of these perceived characteristics are equally relevant in all intergroup contexts. Dominant public discourse can prime those target characteristics that subsequently direct people's behavior (Semetko, 2007;Valentino & Nardis, 2013), and consequently, it can prime justifications for moral exclusion. Specifically, in a media environment where the threatening behavior of Muslims and the importance of defending the nation's boundaries is continuously emphasized, perceived intergroup threat can easily get highlighted as a plausible justification for moral exclusion.
Nonetheless, this extensive media framing was absent in the case of the Roma, which can explain why intergroup threat was not a strong predictor of moral exclusion in their case. On the other hand, in the case of social distancebased moral exclusion, members of the excluded outgroup do not do anything specific to make others exile them from the boundaries of moral regard, but they have not been considered within those boundaries in the first place (Lima-Nunes et al., 2013;Olson et al., 2011). This latter mechanism might be less sensitive to media priming and framing effects, since in such a case, the media is not required to frame the behavior of the outgroup as threatening or immoral in some other ways. This moderating role of media coverage and exposure should be tested by further research in the future.
After interpreting our results, it is necessary to highlight some limitations in our study. We have to stress that our study was a cross-sectional survey study, and as being such, it is not suitable to conclude on the directions of the investigated relationships. Although our model comparison procedure supports that it is reasonable to handle social distance and intergroup threat as mediators in the relationship between collective narcissism and moral exclusion, based on a correlational study we cannot say it with certainty that it is collective narcissism per se that makes people perceive outgroups as either threatening or being psychologically distant.
Although as we saw, intergroup threat and social distance is mostly considered as antecedents of moral exclusion (e.g. Hafer & Olson, 2003;Lima-Nunes et al., 2013) and consequences of collective narcissism (Cichocka, 2016),  Leite et al., 2019;Olson et al., 2011;Opotow, 1994), other causal directions are also worth considering. For example, the literature of moral justification suggests that people tend to rationalize their immoral behaviors not just before but also after their occurrence (Shalvi, Gino, Barkan, & Ayal, 2015). We can therefore assume that perceived outgroup characteristics can function as posterior justifications for moral exclusion, and in such a case the direction between perceived outgroup characteristics (e.g. perceived threat and social distance) and moral exclusion might turn around, because people can be motivated to see the excluded target group in a more negative way in order to justify the exclusion itself. This would result in a model where moral exclusion is placed in the mediator position.
In another possible model, collective narcissism can be the mediator between intergroup threat and moral exclusion.
This can be based on the fact that one important antecedent of collective narcissism is a frustrated need for personal control (Cichocka, 2016;Cichocka et al., 2018). According to the relevant theoretical explanation, if one lacks control over private life events, a narcissistic social identity can be constructed for self-defensive purposes to restore the sense of control. The source of this sort of frustration is mostly considered unrelated to intergroup relations, but if it is a specific outgroup that seemingly frustrates these personal goals, then perceived intergroup threat, as a source of this frustration, might also function as an antecedent rather than a consequence of collective narcissism. This assumed mechanism would implicate a model where collective narcissism is a mediator between intergroup threat and moral exclusion.
Future research might verify these other possible directions between the investigated constructs, but we have to stress that these potential alternative directions and mechanisms do not falsify the mechanism described in our study, and our mediation tests based on different models did not support these alternatives.
Furthermore, we have to note that some indicators of secure or positive national identity is also measured in several studies about the influence of defensive national identity on intergroup relations, but our study did not include such a measure. However, these positive indicators are typically not related or negatively related to negative outgroup attitudes if their negative or defensive counterpart is controlled for (e.g. Cichocka, 2016;Roccas, Klar, & Liviatan, 2006;Schatz, Staub, & Lavine, 1999). Because of this, we assume that the revealed positive relationships between collective narcissism and the measured outgroup attitudes wouldn't be essentially different by incorporating positive national identity, or they might have been even stronger by partialling out the effect of positive national identity. Besides, there is a good chance that the remaining negative direct (non-mediated) relationship between collective narcissism and moral exclusion in both models is due to the suppressing effect of positive national identity (see Cichocka, 2016), and these relationships wouldn't have even appeared with positive national identity incorporated into the models.
Finally, we conducted our study in the context of Hungary which has some unique characteristics that are relevant to the current research. Although both anti-Roma and anti-immigrant attitudes are prevalent in all East-Central European countries, and these issues can be found in the Western European context as well, democratic institutions and media freedom have been severely weakened in the country resulting in an uncontrolled level of governmental propaganda. This context may therefore be unique, and generalizations limited. However, with the growing populism and growing anti-immigrant movements around the world, insights from Hungary may carry important messages outside the country and the region, and highlight the specific importance of understanding the consequences of an inflated but fragile national identification.

Conclusion
In our study, we integrated the literatures of collective narcissism and moral exclusion, and collected empirical evidence to the link between these constructs. Our findings support the idea that the darker side of national identification does not only increase the willingness to commit hostile acts towards certain outgroups as previous research suggested, but it also enhances the chances of the moral justification of these acts. Nonetheless, these justifications do not emerge randomly, but they are the results of motivated social cognition, as it is shown by collective narcissists' tendency to use perceived social distance and threat -two characteristics they are especially sensitive to -to rationalize their moral indifference.

Notes
i) Study 1 from the publication of Kende, Lantos, and Krekó (2018) was carried out by analyzing data from the same omnibus database, but the hypotheses we tested and the theoretical constructs they were based on were different and unrelated to those tested by Kende and colleagues, since they investigated how the ethnic vs. civic perception of citizenship predicts group-based collective action preferences. The only construct that was used in both studies was intergroup threat, what was considered as a mediator between ethnic perception of citizenship and collective action preferences in the cited study.
ii) For more information on the demographic characteristics of the Hungarian population see: http://www.ksh.hu/mikrocenzus2016/kotet_3_demografiai_adatok iii) Descriptive statistics and correlations between the items can be found in the Supplementary Materials (Table S1).
iv) All correlations, pathways, and factor loadings within these models can be found in the Supplementary Materials (Tables  S2-S4).
v) We also have to note that many hits for "roma" included items related to the capital of Italy (Rome), what might easily led to an overestimation of relevant hits for the Roma.

Funding
The study was funded by the National Research, Development and Innovation Office (Grant No.: NKFI-K119433).