Multicultural Attitudes in Europe: A Multilevel Analysis of Perceived Compatibility Between Individual and Collective Justice

Contemporary political philosophers debate the degree to which multiculturalism, with its emphasis on collective justice principles, is compatible with Western liberal societies’ core ideologies based on individual justice principles. Taking on a social psychological perspective, the present study offers a cross-national, multilevel examination of the asymmetric compatibility hypothesis, according to which majority and ethnic minority groups differ in the association between support for individualized immigration policies (based on individual justice principles) and support for multiculturalism (based on collective justice principles). Using data from Round 7 of the European Social Survey (N = 36,732), we compared minority and majority attitudes across 1) countries with stronger versus weaker equality policies at the national level (a Migrant Integration Policy Index [MIPEX] sub-dimension indicator), and 2) Western and post-communist European countries. In line with the asymmetric compatibility hypothesis, ethnic minorities perceived significantly less incompatibility between individual and collective justice than majorities. This majority-minority asymmetric compatibility was stronger in Western countries compared to post-communist European countries. Moreover, in Western countries and in countries with stronger equality policies, ethnic minorities generally supported multiculturalism to a greater extent than majorities. Overall, these findings suggest that deep-seated ideological orientations of national contexts shape minority and majority justice conceptions and hence, also, multicultural attitudes. Implications and future research directions are discussed.

Attitudes towards immigration and multiculturalism by majority and minority groups have recently received ample attention from social and political psychologists (Deaux & Verkuyten, 2014;Green & Staerklé, 2013;Guimond, de la Sablonnière, & Nugier, 2014;Ward, Gale, Staerklé, & Stuart, 2018). Much of this research has demonstrated that the presence of immigrants and culturally diverse groups in society is differentially valued by majority and minority groups (see Just & Anderson, 2015;Sarrasin, Green, Bolzman, Visintin, & Politi, 2018), depending on a variety of individual and contextual factors (for some overviews, see Ceobanu & Escandell, 2010;van de Vijver, Breugelmans, & Schalk-Soekar, 2008). Recent work suggests that these factors include belief systems implying differential or asymmetric sensitivity of majorities and minorities for individual and collective principles of justice (Gale & Staerklé, 2019). The current study investigates this asymmetry across countries.
In a social psychological view, multiculturalism is based on collective, group-based justice principles, emphasizing categorical distinctiveness and conceptualizing society as composed of clearly identifiable subgroups (Gale & Staerklé, 2017;Guimond et al., 2014;Verkuyten, 2005). As a political theory, multiculturalism moreover values the presence of distinct and culturally diverse groups and endorses normative recognition of differences between them, for example through legal protection of cultural practices or language rights (Licata, Sanchez-Mazas, & Green, 2011;Moghaddam, 2008;Taylor, 1992). Political philosophers debate, however, the degree to which multiculturalism, with its emphasis on group rights, is normatively compatible with Western liberal societies' emphasis on individual justice principles, ex pressed in ideological values of individual freedom and personal responsibility (see Barry, 2001;Joppke, 2004;Kymlicka, 1995). This compatibility debate has occurred predominantly within Western contexts such as North America and Europe, while its pertinence has been questioned in the context of post-communist or authoritarian regimes (Kuzio, 2005;Kymlicka & Opalski, 2001).
The present study tested cross-nationally the asymmetric compatibility hypothesis according to which national majorities and ethnic minorities differ in the degree to which they perceive group-based multicultural principles and individual justice principles to be compatible with each other. Indeed, prior research has shown that in an economically liberal national context (Switzerland), individual responsibility beliefs and multicultural attitudes were negatively asso ciated (and hence perceived as incompatible) by majorities, but not by minorities who perceived more compatibility (Gale & Staerklé, 2019). The present study extends this research and is to our knowledge the first to examine the extent to which this asymmetric compatibility hypothesis holds across a variety of Western European and post-communist national contexts.
More specifically, using European Social Survey (ESS) 2014 data, this study looks at how majority and minority groups differ in the association they demonstrate between their support for individualized immigration policy (based on individual justice principles) and their support for multiculturalism (based on collective justice principles). Using a multi-level design, we make a novel contribution by examining whether (geo)political country-level factors (equality-en hancing policies and Western vs. post-communist countries) moderate this individual-level asymmetric compatibility for majority and minority groups. In the following, we first elaborate on what we mean by individualized immigration policy and its foundation in dominant Western individual justice principles. We then explain how multiculturalism-as a collective justice principle-can be conceptualized at the same time as an individual attitude and as a contextual policy indicator that shapes individual attitudes and beliefs.

Individual Justice-Based Immigration Criteria
Predominant in Western Europe, civic integration policies are designed to ensure that newcomers are on a level playing field compared to citizens, for example in terms of their employment opportunities and language skills, but also in terms of their respect for the host country's way of life (Kymlicka, 2012). These policies are often implemented at the initial stage of entry of the immigration process, in the form of individualized selection criteria such as level of education and language competences. Such policies, however, favour high status immigrants who are more likely to have these valued skill sets (Goodman, 2014). Such immigrants are considered valuable for society, increasing cultural diversity while at the same time supporting the national economy.
This civic integration perspective is the foundation of an "individualized immigration policy" (Gale & Staerklé, 2021), which uses individual criteria to determine if a person is allowed to enter and reside in a given country. Reflecting principles of equity theory (Walster, Walster, & Berscheid, 1978), individualized immigration policies "reward" valued in dividual contributions and assets of immigrants by granting them entry into the host nation. Consequently, individuals supporting civic integration policies expect newcomers to follow dominant norms and practices, by encouraging them to be self-supporting and behaving in public space according to generally accepted social and cultural norms (Green, 2009). Cultural maintenance (e.g., celebrations, spoken language) is often only tolerated in the private sphere and therefore, cultural group membership is, in principle, irrelevant for individualized immigration policies (Arends-Tóth & van de Vijver, 2004;Montaruli, Bourhis, Azurmendi, & Larrañaga, 2011).
Individualized immigration policies are designed to assimilate immigrants by requiring them to conform to dominant (Western) values and practices (Bourhis, Moïse, Perreault, & Senécal, 1997;Testé, Maisonneuve, Assilaméhou, & Perrin, 2012). They involve the merit-based expectation that immigrants acquire certain skills or behaviours (see Pehrson & Green, 2010) and as such, individual justice principles underlie their endorsement. At the same time, they also represent a political strategy of "gatekeeping" and thus of restricting immigration. Indeed, the endorsement of such policies is associated with prejudiced attitudes towards immigrants and perceived threat of immigration (Green, 2007(Green, , 2009, suggesting that the emphasis on individuals may imply the rejection of cultural diversity and its underlying group-based worldview. It has been shown that such rejection is especially put into practice by members of dominant, national majority groups. For minority and other subordinate groups, in turn, this incompatibility between principles of individual and collective justice is less clear-cut. Indeed, Kymlicka (2012) suggests that individualized immigration policies do not have to be incompatible with multiculturalism and acceptance of diverse cultural groups, especially if these individualized policies entail the expectation that immigrants conform to superordinate, national values that actually support multiculturalism.

Collective Justice for Minorities Versus Majorities
Research shows that justice at a macro, group-based level is particularly important to minority members, given their heightened sense of group identification and sensitivity to intergroup inequality (Azzi, 1992(Azzi, , 1998. This sensitivity to group-based justice partly explains why minorities support multiculturalism more strongly than majorities (see Green & Staerklé, 2013;van de Vijver et al., 2008;Verkuyten, 2005; see also Sarrasin et al., 2018, for a comprehensive overview of the immigrant perspective). However, when minority members simultaneously desire to become fully accepted members in the larger society, principles of individual justice become necessarily important to them as well, especially in liberal, Western societies, where such principles are pervasive.
The necessity for minorities to simultaneously focus on individual and group-based determinants of civil, social and political rights in order to become fully accepted members in society may explain why minorities perceive greater compatibility between individual and collective justice principles than majorities, especially in national contexts where principles of individual justice are normatively valued. This is consistent with the asymmetric compatibility hypothesis. Indeed, for majorities, membership in a dominant group defined by individualist norms leads them to self-categorize as individuals (Deschamps, 1982;Lorenzi-Cioldi, 1998), to be carefree about group belongingness (see Simon, Aufderheide, & Kampmeier, 2001), and to demonstrate concern for individual rather than for collective justice (Azzi, 1992(Azzi, , 1998Staerklé, 2009). While awareness of their privileges may lead majorities to express greater concern for minorities and collective justice (Leach, Iyer, & Pedersen, 2006;Powell, Branscombe, & Schmitt, 2005), this does not mean they have experienced what it means to be a member of a disadvantaged minority group. Hence, for majorities, endorsing collective justice is likely to come at the expense of individual justice.

Collective Justice at the National Level
There is reason to believe, however, that the extent of this minority-majority asymmetric compatibility hypothesis var ies across countries. Indeed, while individual justice-based policies like civic integration are pervasive in Western liberal societies, governments in some countries have incorporated group-based, equality-enhancing policies as a supplement to them (Banting & Kymlicka, 2013). The degree to which these policies are present within a national context likely has implications for how majorities and minorities conceptualize justice. Different hypotheses can be advanced.
Recent research has shown that attitudes of national populations reflect to some extent the respective institutional settings (see Eisner, Turner-Zwinkels, & Spini, 2021), and multicultural policies are no exception in this respect (e.g., Guimond et al., 2013Guimond et al., , 2014. On the one hand, the asymmetry between minorities and majorities in terms of perceived compatibility between individual and collective forms of justice could be less pronounced in contexts with strong multicultural, equality-enhancing policies. Societal norms in such contexts allow collective justice-based policies to coexist with individual justice-based ones, suggesting all members of society should perceive less incompatibility between the two (i.e., less pronounced asymmetric compatibility).
On the other hand, not everyone abides by social norms: collective justice policies that enhance equality may be empowering for minorities and reinforce their perceived compatibility between individualized immigration policy (individual justice) and multiculturalism (collective justice). For majorities, however, such policies may change little beyond their individual attitudes and perceived norms: While they may support multiculturalism more strongly when multicultural policies are present (Guimond et al., 2013(Guimond et al., , 2014, they may still struggle to reconcile individual and collective forms of justice. In this way, the asymmetric compatibility hypothesis would be more pronounced in contexts with strong equality-enhancing policies. It should be noted, however, that multiculturalism as a political theory and a national policy has a longer history in Western countries than in former communist countries (Kymlicka & Opalski, 2001). Even though post-communist countries in Europe each have their own political history and their own ethnic composition, they have been descri bed, by and large, as "nationalizing states" that attempt to incorporate ethnic minorities into a majority-dominated nation-state (Brubaker, 1996(Brubaker, , 2011. While the generalizability of such analyses is necessarily contested (Kuzio, 2005), it is nevertheless important to take into account this geopolitical dimension in cross-national analyses on the expected minority-majority asymmetric compatibility (see Staerklé, Sidanius, Green, & Molina, 2010), in addition to the degree to which equality-enhancing policies are present.
We suggest that the hypothesized minority-majority asymmetric compatibility between individual and collective justice should be anchored in contexts defined by individualist norms. The compatibility debate has occurred predomi nantly within these contexts, and so perceptions are likely to be most polarized. Moreover, this polarization should be particularly marked between minorities and majorities, as they may both be sensitive to the dominant individualist norms in these contexts, but with different implications for their multicultural attitudes. We therefore expect the asym metric compatibility hypothesis to be supported in particular within Western European countries that are characterized by a historically strong liberal, individual justice orientation, compared to post-communist European countries.

The Present Study and Hypotheses
The purpose of the present study was first to test the asymmetric compatibility hypothesis using nationally represen tative, international survey data. The second objective was to examine whether this asymmetric compatibility could be generalized across European countries. We thus conducted a simultaneous comparison between countries with stronger versus weaker multicultural, collective justice policies, and between Western versus post-communist European countries. Our focus on comparing these regions is crucial because it both acknowledges relevant historical distinctions (e.g., the reign and fall of the former Soviet Union and the "nation-building" that occurred thereafter in post-communist countries; see Kuzio, 2001) and takes into account some important present-day heterogeneity.
To test whether the minority-majority asymmetric compatibility varies as a function of multicultural, collective justice policies, we use a sub-dimension of the Migrant Integration Policy Index (MIPEX; Huddleston, Bilgili, Joki, & Vankova, 2015) called "Equality policies" as a dimension differentiating countries according to their multicultural policy orientation. The MIPEX in general is a numerical country-level indicator representing the degree to which policies within a given country are inclusive and tolerant towards immigrants. This composite indicator assesses a variety of policies in domains such as family reunification, political participation, antidiscrimination, and access to citizenship and permanent residency for immigrants. Its sub-dimension of Equality policies, however, specifically measures the preva lence of governing bodies that are responsible for listening, responding, and taking action in relation to minority claims. These policies demonstrate willingness to institutionally recognize minority groups and to compensate disadvantages experienced by their members.
The general, composed measure of the MIPEX has been successfully used in prior research to explain cross-nation al variation of attitudes towards immigration and cultural diversity (see Ariely, 2012;Green, Visintin, Sarrasin, & Hewstone, 2020;Hooghe & de Vroome, 2015;Schlueter, Meuleman, & Davidov, 2013;; see also Callens & Meuleman, 2017, who used some, but not all sub-dimensions). This research shows, for example, that more inclusive policies are associated with more positive immigration attitudes for majorities, presumably because policies are closely intertwined with shared beliefs at the national level which in turn shape individual attitudes (Guimond et al., 2013(Guimond et al., , 2014Tankard & Paluck, 2016). To our knowledge, however, the equality policy sub-dimension of the MIPEX has never been used in previous research. This sub-dimension is particularly relevant for studying issues related to multiculturalism, as it is visibly a measure of collective justice at the national level. Indeed, this indicator goes beyond issues of individual (civic) integration of minorities by explicitly incorporating multicultural principles of group-based, collective justice.
Our first hypothesis was that majorities would perceive significantly greater incompatibility between support for individualized immigration policies and multiculturalism, compared to members of minority groups (H1). For the reasons stated above, alternative hypotheses were advanced regarding the moderating role played by collective justice policies implemented at the national level: the asymmetric compatibility could be either less pronounced (H2a) or more pronounced (H2b) in such contexts. As for the geopolitical region, our third hypothesis was that the asymmetric compatibility should be greater in Western European countries, compared to post-communist countries (H3). To test these hypotheses, we conducted a multilevel study on 20 European countries. Support for multiculturalism and for individualized immigration policies were measured at the individual level. National collective justice-based equality policies as well as the distinction between post-communist versus Western European countries were located at the contextual level to explain possible country level variation.

Method Participants
We used data from the ESS (wave 7, 2014) comprising 36,732 participants. All countries with relevant data were included in the analyses (20 countries in total; see Table 1; Israel was the only exclusion both because it is a non-European country and because there were missing data). This relatively small number of level-2 units is a limitation in multilevel modelling, especially with respect to the stability of cross-level interactions. According to Kreft and De Leeuw (1998), 20 level-2 groups is the minimum number required, though large sample sizes within each group does to some extent offset the limited statistical power of level-2 analyses. The sample was composed of 52.90% women (coded −1; men coded 1) and the overall average age was 49.33 (SD = 18.63; grand-mean centred across countries). The overall average number of years of full-time education was 12.91 (SD = 3.95; also grand-mean centred across countries), and respondents were primarily citizens of their respective countries (94.7%; coded 1; non-citizen coded −1). Sex, age, years of education, and national citizenship were used as individual-level control variables. Coding allowed for the intercept in the regression analyses to represent a hypothetical average person.

Ethnic Majority and Minority Group Membership
Classification into majorities (coded 1) and minorities (coded −1) was done using a single variable where respondents were asked to indicate whether or not they belonged to a minority ethnic group. During interviews in each country, the term "belong" was defined and clarified when necessary with the social psychological constructs of "attachment" and "identification. " All respondents who indicated "yes" to the question were considered minority members, whereas those who responded "no" were considered majority members. 1 Because the proportion of majorities to ethnic minorities differed substantially between countries, this proportion was controlled for in the final two-level model. The number of self-declared minority and majority members in each country is summarized in Table 1. 2 In addition, to verify our findings with a more objective (rather than self-declared) measure of ethnic group membership, preliminary analyses were also conducted using national citizenship as a proxy (for countries where the dataset included non-national respondents). Similarities and differences between these analyses are reported in the Supplementary Materials (see Point 1 and Table A2).

Individual-Level Variables
Support for individualized immigration policy was measured using four items. On an 11-point Likert scale, respondents were asked to indicate the degree to which they considered four criteria to be extremely unimportant (0) to extremely important (10) when determining whether someone born, brought up and living outside the country should be allowed to come in and live there. The four items referred to educational qualifications, speaking the country's official language, work skills that the country needs, and commitment to the way of life in the country, and were collapsed into a single score. 3 Internal consistency of the measure was satisfactory across all countries, ranging from .69 in Slovenia to .82 in Sweden. Higher scores indicated support for individualized immigration policy. Average scores across countries showed fairly strong support in general (M = 6.91, SD = 2.07), with the weakest support in Norway (M = 5.48, SD = 2.27) and the strongest in the UK (M = 7.58, SD = 1.77). The variable was group-mean centred to prepare for cross-level interactions (described below; see Enders & Tofighi, 2007). Support for multiculturalism was assessed with six items investigating perceived benefits of cultural diversity resulting from immigration. Respondents were asked to indicate the degree to which they believed the country's economy, cultural life, job market, taxes and services, and crime problems are undermined or enriched by immigrants, as well as the degree to which they felt immigrants made the country a better or a worse place to live. All items were coded on an 11-point Likert scale ranging from 0 (bad for the economy, cultural life undermined, take jobs away, generally take out more, crime problems made worse, worse place to live) to 11 (good for the economy, cultural life enriched, create new jobs, generally put in more, crime problems made better, better place to live) and were collapsed to create a single score. Internal consistency of the measure was adequate across all countries, ranging from .79 in Switzerland to .90 in 1) Despite its inherent limitations (e.g., ambiguity concerning the nature of actual minority group membership, differential meaning of the indicator across countries, social desirability issues, see Burton, Nandi, & Platt, 2010;Hoffmeyer-Zlotnik & Warner, 2010), this subjective measure of minority group membership was an adequate indicator for our research question given its social psychological nature. Indeed, to self-declare membership in an ethnic minority group, individuals need to identify with that group and feel they belong to that group, thus implying processes of social identification assumed to underlie the effects predicted.
2) To provide additional external validity to this subjective measure, cross-tabulations and Chi-Square tests were used to examine objective criteria explaining self-identification as an ethnic minority group member. In all countries, non-citizens were more likely to self-identify as an ethnic minority group member than citizens (p < .001). Moreover, in all countries, respondents whose mother and/or father were not born in the country were more likely to self-identify as an ethnic minority group member than respondents whose mother and/or father were born in the country (p < .001). While linguistic (national) minorities such as French-speakers in Belgium, French-, Italian-, and Romansh speakers in Switzerland, and Swedish-speakers in Finland were more likely to self-identify as members of an ethnic minority group than linguistic (national) majorities in these contexts, most of these linguistic minorities did not self-identify as ethnic minority group members. However, most Russian-and Polish-speakers in Lithuania, and Russian speakers in Estonia did self-identify as members of an ethnic minority group-these national minorities were also more likely to report foreign ancestry. Therefore, while national minority status occasionally explained self-identification as an ethnic minority group member, the most common explanation was foreign ancestry.
3) The criterion of speaking the country's official language has been considered by previous literature non-univocally as an acquired (i.e., "civic"; Green, 2009) or as a "cultural" (Reijerse, Van Acker, Vanbeselaere, Phalet, & Duriez, 2013) criterion to accept immigrants. Therefore, we also tested our models using an alternative three-item measure excluding this item (see Supplementary Materials, Table A3).
Austria. 4 Higher scores indicated support for multiculturalism, and average scores across countries were around the midpoint of the scale, if not just below, showing moderate levels of support (M = 4.77, SD = 1.71). Support was weakest in Hungary (M = 3.84, SD = 1.68) and strongest in Sweden (M = 5.87, SD = 1.60). Since the "cultural" component of multiculturalism is most evident in two of the six items (country's cultural life enriched by immigrants and immigrants make the country a better place to live), multilevel analyses were also replicated using this restricted measure of support for multiculturalism. Similarities and differences are reported in Supplementary Materials.

Country-Level Variables
The Equality Policy Index, extracted from the 2014 MIPEX (Huddleston et al., 2015) was used as the first country-lev el moderator. The Equality Policy Index (EQPol) assesses the prevalence of national government bodies responsible for ensuring equal treatment of all groups, especially those who experience discrimination on grounds of "race/ethnicity, religion/belief and/or nationality. " It also implies the existence of legal infrastructure that supports "positive action measures" and where minority claims are appropriately handled (Huddleston et al., 2015). The variable thus represents an indicator of collective justice policy at the national level. Scores ranged from 0 to 100, with higher scores represent ing the presence of group-based equality policies. Country scores for the years 2014 are provided in Table 1, revealing that Portugal and Sweden had the highest EQPol scores, while Czech Republic and Germany had the lowest scores. The variable was mean centred in preparation for analyses.
Western European versus post-communist countries were compared by coding post-communist countries (Czech Republic, Estonia, Hungary, Lithuania, Poland, Slovenia) as −1 and Western countries (all remaining ones in Table 1) as 1. This variable served as the second country-level moderator. The relatively low correlation between the two country-level moderators (r = .349, p = .132, country level) suggests that multicollinearity was not an issue.

Data Analysis
Data preparation, descriptive statistics, and country-by-country regressions were conducted using SPSS version 25. Multilevel analyses were conducted using Mplus version 8.2 (Muthén & Muthén, 1998. To test H1, we calculated an interaction at the individual level between majority/minority group membership and support for individualized immigration policy on support for multiculturalism. To test H2 and H3, we calculated independent cross-level inter actions between the respective context-level variables (Equality Policy Index; geopolitical region) and the previous individual-level interaction, again predicting support for multiculturalism in a level 2 model.
Post-stratification weights were applied for all analyses to ensure the use of nationally representative samples (espe cially of majorities, although the proportion of minorities in most countries increased slightly once these weights were applied). For missing data, we used listwise deletion for sociodemographic variables (0.50% missing data) and applied Maximum Likelihood estimation for support for individualized immigration policy and support for multiculturalism items (2.65% missing data). No data were missing for the country-level indicators.
To ensure robustness of the findings a number of additional analyses were conducted using alternate individual-level measures (mentioned above), including country-level predictors independently, including country-level controls (Gross Domestic Product [GDP] and income inequality [GINI] indices), and running part of the model with ESS 2002 data (Round 1). These analyses are reported in the Supplementary Materials (Tables A2-A8). 4) A Principal Component Analysis with Varimax rotation was run in each country to ensure that the items composing the support for multiculturalism indicator and the support for individualized immigration policy measure loaded onto two separate constructs. This two-factor solution was indeed found in all 20 countries. Table 1 shows country-by-country zero-order correlations between support for individualized immigration policy and multiculturalism among minorities and majorities. Across all observed countries, majorities show a significantly negative association between the two measures. Correlations are also generally negative among minorities, albeit less consistently, ranging from significantly positive only in Portugal, to significantly negative in the Czech Republic (interpretation of effect sizes is relatively congruent with this pattern).

Preliminary Country-by-Country Analyses
Prior to multilevel analyses, we conducted multiple regressions for each country, with support for multiculturalism as the dependent variable (see Supplementary Materials, Table A1). Predictors included ethnic minority and majority group membership, support for individualized immigration policy, the interaction between the two, as well as control variables. The interaction effect tested our first (asymmetric compatibility; H1) hypothesis, suggesting that support for individualized immigration policy would be negatively linked to support for multiculturalism for majorities, but not necessarily for minorities.
The interaction was statistically significant, in the expected direction, in eight of the twenty countries, namely Belgium, Switzerland, Denmark, the United Kingdom, the Netherlands, Norway, Portugal and Sweden. These are all Western European countries, with Portugal and Sweden having the highest scores on the Equality Policy Index. These findings provide preliminary evidence of a potential moderating role of the two country-level variables on the asymmetric compatibility hypothesis. Indeed, for all other countries, the effect was non-significant, with the exception of the Czech Republic where it was significant (p = .015), but in the opposite direction.
In general, these results show that in many European national contexts, majorities perceive greater incompatibility than minorities between support for individualized immigration policy and multiculturalism. However, results also suggest that there is considerable variation between countries, both when comparing post-communist versus Western countries and in terms of national equality policies. Multilevel analyses were conducted to determine if country-level predictors account for such variation.

Multilevel Analyses: Random Intercepts (Fixed Slopes) Models
As a first step, we calculated an intercepts-only model on support for multiculturalism in order to serve as a benchmark to which subsequent models are compared (see Hox, 2010). The intraclass correlation was equal to .080, meaning that 8% of the variance of support for multiculturalism was located at the country level and could therefore be explained by level-2 variables (country-level variance b = 0.235, SE = 0.085, p = .006; individual-level variance b = 2.705, SE = 0.119, p < .001).
As a second step, we calculated a random intercepts (fixed slopes) model to bring preliminary analyses together into a single model (see Table 2, first column). Individual-level predictors and control variables (see table notes) were incorporated into this model and slopes were not allowed to vary across countries. Individualized immigration policy and group membership main effects were significant, suggesting that increased support for individualized immigration policy was generally associated with decreased support for multiculturalism, and that majorities generally supported multiculturalism significantly less than minorities (identical interpretation when the level-1 interaction was omitted from the model). Moreover, the overall interaction between ethnic minority and majority group membership and support for individualized immigration policy was significant (p = .002), in the expected direction. 5 Consistent with the asymmetric compatibility hypothesis (H1), the decomposition of the interaction, which can be seen in Figure 1, revealed that the negative association between support for individualized immigration policy and support for multiculturalism was stronger among majority respondents, B = −0.21, SE = 0.02, p < .001, compared to minority respondents, B = −0.12, SE = 0.03, p < .001. (Point 2 and Tables A2, A3, and A4) for converging evidence using alternative variables and datasets. Step

Level 2 Model
Step 1

Level 2 Model
Step 2

Level 2 Model
Step 3

Multilevel Analyses: Random Intercepts and Slopes Models
The final models brought country-level variation into the equation. First, a level-1 model testing random slopes was calculated to provide information on the degree to which the associations between predictors and support for multiculturalism varied across countries (see Table 2, second column). The reduced Akaike information criterion (AIC) value for this model in comparison to the random intercepts model indicated a better model fit (Hox, 2010). Control variables were still included in the model and their effects were not allowed to vary. Results in this model showed that the statistically significant interaction term between ethnic minority and majority group membership and support for individualized immigration policy highlighted above (p = .002) did not significantly vary across countries (p = .143). Therefore, compared to majorities, minorities overall perceived less incompatibility between support for individualized immigration policy and multiculturalism. The support for individualized immigration policy and group membership main effects did however vary significantly across countries. Three two-level models were then calculated, each building on the previous one.
Step 1 incorporated the level-2 variables (post-communist vs. Western European countries and equality policies at the national level) as main effects predicting support for multiculturalism.
Step 2 incorporated two-way cross-level interactions to determine if these level-2 variables could explain some of the country-level variance in the main effects of support for individualized immigration policy and of group membership.
Step 3 incorporated three-way cross-level interactions. Because our theoretical reasoning as well as preliminary country-by-country analyses suggested the level-1 interaction term should vary, we decided to continue to let this interaction vary in the final models (see Aguinis, Gottfredson, & Culpepper, 2013), in particular to examine whether the level-2 predictors would explain any potential variation (in line with H2 and H3). Indeed, while the random slopes model showed non-significant country-level variation in the level-1 interaction term (p = .143), this is likely due to the small number of level-2 units. 6 Only the full models are reported here for space reasons. However, simpler models were also tested where the main level-2 predictors were inserted individually (see Tables A5 and A6 in Supplementary Materials). 7 This is important because the limited number of level-2 cases increases the risk of biased results when too many contextual variables are included in the model (see Stegmueller, 2013). Similarities and differences with these simpler models are therefore reported where relevant.
First, main effects showed that support for multiculturalism was significantly stronger in Western European coun tries than in post-communist countries (p = .008; see Step 1 in Table 2). Second, a marginally significant cross-level interaction was found between post-communist versus Western European countries at level 2 and support for individual ized immigration policy at level 1 (p = .061; see Step 2 in Table 2). 8 Figure 2 shows that the negative relationship between support for individualized immigration policy and support for multiculturalism was marginally more pronounced in Western European countries, B = −0.19, SE = 0.02, p < .001, compared to post-communist countries, B = −0.12, SE = 0.03, p < .001. Third, we found a marginally significant cross-level interaction between the presence of equality policies at level-2 and minority/majority group membership at level-1 (p = .087; see Step 2 in Table 2). 9 Figure 3 shows that the same difference between minorities and majorities in support for multiculturalism was marginally stronger in countries with more extensive (+1 SD) equality policies, B = −0.28, SE = 0.04, p < .001, than in countries with less extensive (-1 SD) equality policies, B = −0.15, SE = 0.07, p = .028. Fourth, a marginally significant cross-level interaction was found between post-communist versus Western European countries at level-2 and minority/majority group membership at level-1 (p = .069; see step 2 in Table 2). 10 Figure 4 shows that the difference in support for multiculturalism between minorities 6) As Aguinis et al. (2013) suggest, "given that […] multilevel modelling is usually conducted with L2 sample sizes that are much smaller [than same-level research], it is possible that in many situations there may be an incorrect conclusion that τ is not different from zero due to insufficient statistical power […]. To balance Type I and Type II error considerations, our recommendation is to proceed with the cross-level interaction test even when the null hypothesis of no slope variance is retained when there is a strong theory-based rationale for a particular hypothesis" (p. 1502).

Cross-Level Interaction: Equality Policies at Country Level Associated With Increased Minority Support for Multiculturalism
10) See Supplementary Materials (Point 5, Tables A3, A4 and A6) for converging evidence using alternative models.

Cross-Level Interaction: Western Countries Associated With Increased Minority Support for Multiculturalism
A significant cross-level (three-way) interaction was found between post-communist versus Western countries and the level-1 interaction (p = .016; see Step 3 in Table 2). The direction of this three-way interaction was negative, suggesting that Western countries are characterised by greater minority-majority asymmetric compatibility, in comparison to post-communist countries. This finding was in line with H3 as well as the direction of preliminary country-by-country analyses. 11 When decomposing this three-way interaction, simple slope analysis (Preacher, Curran, & Bauer, 2006) showed that the negative association between support for individualized immigration policy and support for multiculturalism was the strongest among majority members in Western European countries, B = −0.23, SE = 0.02, p < .001, while the associa tion was weaker among minorities in Western European countries, B = −0.12, SE = 0.04, p = .001, among majorities in post-communist European countries, B = −0.14, SE = 0.03, p < .001, and among minorities in post-communist European countries, B = −0.13, SE = 0.05, p = .008. 12 There was no significant cross-level (three-way) interaction comprising the presence of equality policies in this model. H2a and H2b (that involved competing predictions) thus did not receive support. 13

Discussion
A key debate in the political philosophy of multiculturalism concerns the relationship between individual and collective justice principles that provide normative justification to immigration policies (Kymlicka, 1995(Kymlicka, , 2012. In this view, rights are often granted to immigrants on an individual basis as a function of their own unique skills, competences, and motivations. At the same time, multiculturalism and the arrival of immigrants involves a group-based worldview whereby the cultural and ethnonational categories to which these newcomers belong, differing from the majority, are salient and viewed positively. The present research examined how these individual and collective justice principles relate to each other as a function of minority and majority group membership across different national contexts. In line with the asymmetric compatibility hypothesis (which comprised our first hypothesis), we expected that individual and collective conceptions of justice were more mutually exclusive among majorities (principle of incompatibility) than among minorities. Extending previous research that focused on Switzerland, that is a single, liberal and Western European country (Gale & Staerklé, 2019), the current findings provide new cross-national evidence for this asymmetric compatibility hypothesis. This key finding was largely consistent across the main and alternative analyses, taking into account two indicators of minority and majority group membership (subjective and objective, respectively), two measurements of support for multiculturalism (six and two items, respectively), and two separate datasets (ESS wave 7, 2014, andESS wave 1, 2002, respectively).
Echoing early findings by Azzi (1992), our findings highlight that attitudes towards multiculturalism emerge in an intergroup context in which majorities and minorities develop differentiated justice conceptions. Considering that both individual and group-based forms of justice have intrinsic moral justification-one granting priority to individuals in justice decisions, the other one to groups-attitudes towards cultural diversity and immigration imply a trade-off that is resolved differently by individuals as a function of their majority or minority membership.
Majorities tend to pit individual against collective justice principles while favouring individualized immigration poli cies over multiculturalism (incompatibility). This emphasis on individual merit inherent to individualized immigration policy is based on pervasive individual justice principles (Azzi, 1998;Deschamps, 1982). Importantly, prior research (Green, 2007(Green, , 2009 has shown that such emphasis is associated with rejection of cultural diversity among majorities who are prejudiced against immigrants. For minorities, however, the current findings suggest that the group-based justice conception underlying endorsement of multiculturalism is less irreconcilable with individual justice principles, thus potentially extending the scope of justice to both individuals and groups (see Gale & Staerklé, 2019).
The chief novelty of this study was to examine country-level variation in this asymmetric compatibility hypothesis. Across all multilevel analyses, we found support for our third hypothesis showing that the effects were stronger in Western European countries compared to post-communist European countries. In Western European countries, majori ties perceived significantly greater incompatibility between individual and collective justice principles. This occurred not only when Western European majorities were compared with minorities there, but also when they were compared with both minorities and majorities in post-communist European countries. This finding suggests that the differential perspectives involved in the debate between multiculturalism (based on groups) versus liberalism (based on individuals) (Barry, 2001;Joppke, 2004;Kymlicka, 1995) are indeed more polarized among the public in Western European countries characterized by a historically pervasive liberal, individual justice orientation (Kuzio, 2005;Kymlicka & Opalski, 2001). In other words, perceived incompatibility between individual and collective justice principles may not be generalizable to all countries (and groups) around the world, but may instead be context-specific and mainly anchored in Western, classically liberal societies.
Our second hypothesis, however, did not receive support. Indeed, the majority-minority asymmetric compatibility did not vary as a function of country equality policies (i.e., the degree to which governmental bodies ensure minority claims are heard and attended to). This may be due to opposing forces at play, as reflected by our competing hypotheses for this effect. It is also possible that the equality policy indicator did not capture relevant country variation. A more concrete measure of the way in which equality policies are implemented in each country (e.g., a quantification of the proportion of successful cases, where minority claims were indeed addressed and actively rectified, rather than an abstract indicator of the presence of such policies) might have been a more appropriate indicator to test this hypothesis. Nevertheless, the presence of such policies played a marginal role in explaining basic support for multiculturalism by minorities: Minorities were more supportive of multiculturalism than majorities in countries with stronger equality policies (as well as in Western countries). This finding adds an important contribution to research showing that multicultural policies often have an impact on majority opinions towards multiculturalism and immigration in general (Green et al., 2020;Hooghe & de Vroome, 2015;van de Vijver et al., 2008), and that such majority attitudes are shaped by shared beliefs which are closely intertwined with policies (Guimond et al., 2013(Guimond et al., , 2014Tankard & Paluck, 2016). Our (marginal) result, in turn, suggests that such policies may, in fact, predominantly affect minority opinions (see also Bloemraad & Wright, 2014) and that multicultural, group-based, equality policies do not necessarily shape shared beliefs of all groups in society. These collective justice policies at the national level are especially meaningful to minorities, possibly empowering them (see Vorauer & Quesnel, 2017), without negative consequences for majorities in terms of support for multiculturalism.
This study has a number of limitations, but also points towards directions for future research. First, the dataset used in the present study was composed of only 20 countries, which is arguably the minimum for multilevel research (Kreft & De Leeuw, 1998). The results may thus have been driven by specific post-communist and Western European countries, possibly obscuring heterogeneity both within and between these regions. More countries (i.e., a higher number of level-2 units) would have allowed for more reliable results. Relatedly, while an intraclass correlation coefficient (ICC) of .08 is common in multilevel research involving countries (see Stegmueller, 2013), this implies that effect sizes are fairly small. In the case of the present study, it is worth noting that in the final model, 2% of the individual-level variance was explained by our individual-level predictors (R 2 = .022), and that 44% of the country-level variance was explained by our country-level predictors (R 2 = .441). To put this into context, the latter number represents 4% of the total variance in support for multiculturalism, given that 8% of the variance was located at the country-level as indicated by the ICC (see Snijders & Bosker, 2012).
Second, the use of secondary survey data brings up a number of issues, including that the measures used in this study were not always ideal. For example, the items used to represent support for multiculturalism had qualitatively different meanings depending on their low or high scores. These items have been interpreted in existing research as perceived (material and symbolic) threat of immigration, in line with their low scores (see for example Schneider, 2008). This threat is nevertheless rather consistent with our current interpretation. Indeed, perceiving the presence of culturally diverse groups as threatening can be a means of opposing principles of collective justice.
Moreover, in some countries, the minority groups were very small, and the measure used to identify them was based on a subjective self-definition as an ethnic minority member. Future research should investigate ethnic groups more closely, ideally with data for which minorities have been oversampled. These minorities could include immigrant minorities (citizens or non-citizens who have close ties to another national culture after emigrating) and national minorities (citizens who are members of a native subgroup making claims for official recognition, self-governance or special representation; see Kymlicka, 1995), who are often subordinate both in terms of power and status in relation to the national majority. The perspective we have adopted in the present paper indeed suggests that minorities are de facto disadvantaged relative to majorities. It would also be interesting to examine high status minorities such as German-speaking Italians in the South Tyrol region of Italy or Catalan-speakers in North-Eastern Spain. Indeed, despite their objectively "advantaged" societal position, high status minorities may still be subjectively aware of their minority status and would thus be sensitive to principles of collective justice (while at the same time supporting dominant individual justice principles).
Third, a valuable direction for future research would be to look at an even broader array of national contexts including Asian, African, and American countries, thereby allowing for more encompassing comparisons and more fine-grained analyses of the dimensions organizing cross-national variation.
Finally, the present study has used correlations to interpret compatibility between individual and collective justice. To complement this method, future research should directly measure perceived compatibility between individual and collective justice.

Conclusion
This study offered a social psychological perspective on a contemporary debate in political philosophy that questions the compatibility between collective justice principles as expressed in multicultural ideology and individual justice principles inherent to Western liberal societies. Results of the present study suggest that this question of compatibility between individual and collective justice is meaningful especially within classically liberal, West European societies, and that it may be less meaningful in post-communist, "nationalizing" countries (Brubaker, 1996(Brubaker, , 2011 that have more recently been confronted with processes of nation-building and democratization (Kuzio, 2001(Kuzio, , 2005. This conclusion is critical for research on multicultural attitudes as it shows the importance of taking into account the cultural and geopolitical context in which these attitudes are shaped.
In Western societies, individual justice and its implementation in terms of civic integration policies has important value for both majorities and minorities (Joppke, 2004). However, these policies do not seem to suffice when it comes to the minority perspective on justice. Indeed, civic integration policies are arguably the foundation upon which collective justice and respect for minority rights is built. Kymlicka (1995Kymlicka ( , 2012 argues that multiculturalism is compatible, and is thus not in competition, with civic integration. When collective justice-based multicultural policies are implemented as a complement to civic integration policies at the national level, they should not be harmful for majorities. However, establishing such a balance between individual and collective justice in liberal societies should, indeed, be particularly empowering for minorities.

Funding:
The authors have no funding to report.