The impact of regionalism on anti-immigrant attitudes: a multilevel international comparative study

ABSTRACT Regionalism, understood as affective attachment of people to place, is shifting into the focus of social scientific attention. An important aspect of this are the globally observable encounters of growing numbers of immigrants with regional identities. Empirical work concerning the impact of regionalism on anti-immigrant attitudes has been contradictory so far. Results range from a moderate to no influence. The paper, thus, analyses the intricate relationship between hostility towards immigrants and regionalism in detail. Merging country-level data on the extent of regional autonomy in a nation-state with individual and regional data on regionalism enables us to gauge the relevance of regionalism for anti-immigrant attitudes in a comparative perspective. The paper presents a multilevel regression analysis that shows a mitigating relationship of the presence of extensive regional autonomy in a nation-state on anti-immigrant attitudes at the country level. Regionalism, unlike nationalism, does not correlate with anti-immigrant attitudes at the regional or individual level, while nationalism is clearly significant at both levels. These results reflect general global prejudice trends and the call for further work on the effect mechanisms of institutionally granted regional autonomy on openness to immigrants.


INTRODUCTION
Rising global immigration figures, and ensuing growth in hostile attitudes toward newcomers, are urgently putting the affective attachment of people to space and place back onto the map of the wider social sciences. With nationalism (Antonsich, 2018;Pehrson et al., 2009) and regionalism (Agnew, 2001;Karolewski, 2007;Sindre, 2018) human geography discusses two possible facets of the correlation between anti-immigrant attitudes and place-related ties. Such ties are understood as people's attachment to 'imagined communities' (Anderson, 2016) that occur at both the national and regional scales. However, these concepts of nationalism and regionalism significantly differ in their impact on attitudes towards immigration, as well as in the frequency of their usage for empirical investigations.
Little work has been conducted to date on investigating to what extent regionalism, understood as affective bonding to subnational regions (Agnew, 2013), correlates with anti-immigrant attitudes (Curtis, 2014). This is all the more astonishing as, on the one hand, nationalism research speaks of a 'nationalist fallacy' (Bonikowski, 2016, p. 442)a misbelief in the assumption that people feel exclusively attached to only one nationand, on the other hand, social sciences show a history of deep engagement with forms and repercussions of regionalism in multilayered ways (Agnew, 1999(Agnew, , 2001(Agnew, , 2013Ciută, 2008;Entrikin, 1996Entrikin, , 2011Glass, 2018;Luukkonen & Sirviö, 2019;Massey, 1979;Sindre, 2018). Social geography and ethnic studies in particular show a huge lack of commitment in analysing the impact of regionalism on anti-immigrant attitudes and minorities in general. Existing work tends to focus on the positive aspects of regional identities for political identities, especially in the context of perceived threats through increased immigration (Agnew, 2001), but overlooks the conceivable correlations between exclusion attitudes and regionalism on different spatial levels (Escandell & Ceobanu, 2010).
Therefore, the main aim of the paper is to determine the relationship of regionalism with anti-immigrant attitudes by examining the correlation structure of predictors derived from multinational prejudice studies. From this basis, approaches examining the interconnectedness of regionalism, as an identification with the local cultural entity of the region (Agnew, 2013), and hostile attitudes toward newcomers can be further developed (Escandell & Ceobanu, 2010;Fitjar, 2010). Specifically, the paper examines the impact of regionalism, in combination with nationalism, to further assess its potential for reinforcing or mitigating exclusion attitudes. In so doing, this paper illuminates the potentially exclusionary or inclusionary side of regional attachment and the formation of regional reference groups. The paper adds, thus, to the social scientific knowledge 'of regional societies which now share a wide range of social and cultural characteristics' (Agnew, 2013, p. 15).
The database for the study is the International Social Survey Programme (ISSP) 2013 (ISSP Research Group, 2015) that contains representative samples of the adult population of the 29 different national states considered here. ISSP 2013 includes questions concerning national consciousness, national and regional identity as well as anti-immigrant attitudes. To assess the impact of regionalism more precisely, the ISSP data set is complemented with macrodata at the country level concerning migration (IOM Global Migration Data Analysis Centre, 2019), ethnic fractionalization , extent of regional authority , and the presence of an autonomous region in the respective country taken from the Database of Political Institutions .
The remainder of the paper is structured as follows. The next section outlines the background of the regionalism and anti-immigrant attitudes nexus in detail. The third section presents information on the data set and the analysis strategy. The subsequent section presents the relevant results of the multilevel regression analysis. It becomes clear from the models that regionalism at the individual and regional level shows no impact on anti-immigrant attitudes, whereas at the macrolevel regional authority in a nation-state has a mitigating influence on exclusion attitudes. The paper ends with a discussion of these empirical findings and draws further conclusions.

Regionalism
The notion of regionalism lacks an explicit definition as it subsumes rather different aspects under one single term (Middell, 2012). Much research on contemporary regionalism in wider social science emphasizes regional initiatives as responses to challenges related to globalization (Agnew, 2013;Iammarino et al., 2019;Väyrynen, 2003;Zimmerbauer & Paasi, 2013). This work focuses on notions of regionalism as entanglements of groups of people and specific territories in line with a main assumption of nationalism. Referring to identity conceptions in political geography, in a landmark paper on regionalism Knight identifies 'group politico-territorial identities' (Knight, 1982, p. 514) as a pivotal issue of geographic regionalism. Consequently, Knight defines regionalism according to this notion as: the awareness of togetherness among a people of a relatively large area. A regionalism becomes evident only at certain scales of abstraction, at certain scales of generalization. A regionalism thus is recognizable only when it represents but a part of a larger territorial unit, the latter being the areal extent of a political system. (p. 518) Regionalism has, thus, an important cognitive dimension. Regions are imagined with constructs such as trust, identity or shared values, which sediment in cognitive territorial entities that enable the demarcation and drawing of boundaries (Väyrynen, 2003). These regional imaginations use devices to transform particular specifics of place into reified entities that thereafter become efficacious. For Anderson (2016), local language is one of the most important of these devices.
Regionalism, understood as 'awareness of togetherness' (Knight, 1982, p. 518) on a specific regional scale, is thus not an entity existing in the world, but a possibility to interpret the world (Brubaker, 2004). With that said, regionalism as a form of plural identity also functions as an analytic rubric that focuses research efforts on situated actions, cultural idioms, common senses in respect of knowledge, routines, resources, or political projects and variable groupings (Brubaker, 2004). These could be aligned to subnational territorial links. As Paasi emphasizes, regions as cognitive territorial entities are processes characterized by their dependence on institutionalizations that secure the possibility of boundary-making as the initiation of the region itself (Paasi, 2009). The conceivabilities of the respective peer group that feels associated with the specific segment of space constituting the region is, thus, the main sphere of interest for social science efforts researching regionalism.
Agnew argues that peer-group identities are crucial for the realization of regions (Agnew, 2013), as regionalism manifests itself as territorially bounded relations in groups and the juxtaposition of these group relations on different scales (Knight, 1982). To tackle the complex process of regionalism as a specific characteristic of cognitive relations between peer groups and territorial entities, Brubaker and Cooper propose analysing the mechanisms underlying the emergence of the sociocultural imagination of identity in peer groups (Brubaker, 2004;Brubaker & Cooper, 2000). A sense of community (Buber, 1919) is presumed as jointly responsible for the emergence of a sense of regional identity as an imagination of togetherness among groups on the regional scale. However, such a culturalistic account for the explanation of group awareness is frequently criticized for the imprecise nature of its explanatory claims (Brubaker & Laitin, 1998), especially on the regional scale (Fitjar, 2010), as well as for the implicit assumption of a shared consensus of values among all members of the (regional) peer group (Wrong, 1961).
A second major strand of regionalism research in social sciences focuses on different peculiarities of regional identity. Regionalism as collective attachment to the region accompanied by a sense of connectedness (Knight, 1982) is expanded in this concept to a regional identification, comparable to a sense of national belonging. In contrast to national identity, which attracts much research (Esses et al., 2005;Esses et al., 2006;Yang, 2014), regional identity is often conceptualized as pre-existing. It often serves as evidence for a high quality of life or social cohesion in a particular region (Paasi, 2011). Regional identity is closely related to the conception of a peer group and combines this with a close identification of its members with a region, which is ordinarily distinguished by natural and cultural qualities (Paasi, 2012). This identification of people with regions can be regarded as nested in a Russian doll system of 'local, regional, national, and supranational identity' (p. 1453).
The regionalism/anti-immigrant attitudes nexus This paper extends the discussed literature and scrutinizes to what extent regionalism influences anti-immigrant attitudes as a variety of 'enduring political and economic implications of those regional identities' (Johnson & Coleman, 2012, p. 874). In the context of global migration processes, the regional scale is gaining in importance for explanations of exclusionary and hostile sentiments towards newcomers because the region evokes affective attachment and practices in everyday life. The regionalism-resentment nexus as a topic of research starts around 2000 (Agnew, 2001). The relationship between regional identities, political identities and perceived threats to these identities due to immigration has been of particular interest here (Agnew, 2001). This relationship is accompanied by a hierarchical ordering of outgroups by an assumed cultural distance in terms of language, values or customs (Joppke, 2005).
However, the impact of regionalism on anti-immigrant attitudes is rarely tested in prejudice studies. In current work, addressing people's connection with regions tends to be conceptualized as a challenge for national identity (Johnson & Coleman, 2012;Sindre, 2018), rather than exploring the potential fostering or mitigating of hostile attitudes towards outsiders through regional identities. Curtis (2014) for instance does not find a correlation in the Eurobarometer data between the assertion of 'feeling regional' and anti-immigrant attitudes. She therefore concludes that '(r)egional identification had no effect, suggesting that subnational affiliations may be less salient for immigration attitudes than other identities' (p. 539). However, because Curtis's assessment roots in simply one variable for regional identification, she makes the case for further investigation of the relationship between anti-immigrant attitudes and regionalism with a more complex set of variables. Ceobanu and Escandell (2008) find regional peculiarities in anti-immigrant attitudes as well as the tendency for convergence of these characteristics in Europe. They explain their finding by mimetic tendencies between regions. Hackett (2009) comes to a similar result in her empirical validation of regional identity in Newcastle, which bases on an ascribed openness to newcomers. Her empirical evidence highlights mimetic tendencies towards immigrant exclusion in England despite the contrary regional identity in Newcastle (Hackett, 2009). With that said, it becomes noticeable that regional identities, which are based on narratives including stereotypes, ethnicity, or community demarcations, are in the position to draw boundaries between insiders and outsiders (Paasi, 2003). According to Paasi, particularly refugees and immigrants are targets of these identity politics of regional exclusion (Paasi, 2012).
Two directions of the effect of regionalism as attachment to the region resp. regional identities on anti-immigrant attitudes can be found in the literature. The first hypothesis is that regional identification of this kind reinforces anti-immigrant attitudes, since the region is a social entity that should in principle be protected against strangers (Paasi, 2012). On the other hand, the second hypothesis is that regionalism promotes positive attitudes towards immigrants, namely when immigration shapes regional culture and contributes to the possibility of differentiating regional identity from the nation-state (Fitjar, 2010).
In accordance with the first hypothesis, Escandell and Ceobanu (2010) develop an approach involving 'individual self-association with regions' (p. 160), which postulates that individuals with a solid regional or ethnic-regional identification are more disposed to showing anti-immigrant attitudes. Escandell and Ceobanu (2010) offer as an explanation at the aggregate level that contexts with high levels of regional identification among the population evoke high levels of immigrant exclusionism. Individual prejudice is therefore traced back to collective values of regional peculiarity. van der Zwet (2016) analyses the potential relationship of regionalism and anti-immigrant attitudes with a shift of nationalist tendencies into a more positively evaluated regionalism by regionalist parties. Accordingly, rather hostile attitudes, for instance towards immigration, can be disguised under the umbrella of regionalism or regional identity. Right-wing or regionalist political parties can therefore circumvent negative associations of nationalism and are electable for broad strata of the population (van der Zwet, 2016).
Perceived group threat also explains the potentially positive correlation between regionalism and anti-immigrant attitudes. The group threat thesis (Blumer, 1958) draws on the phenomenon that the reference group's perception of an outside group's prerogatives is construed as a peril. Members of the dominant group are then inclined to raise dislike against members of the outside group (Quillian, 1995). For instance, immigration is seen as responsible for rising social costs in a region which triggers negative interest among the dominant regional group (Küng, 1981). The threat approach emphasizes the perceived danger for regional identity and supremacy due to the influx of immigrants (Hainmueller & Hopkins, 2014) and interprets attitudes of rejection as efforts to maintain social control and the status quo by the dominant group (Olzak & Shanahan, 2014). Moreover, anti-immigrant attitudes foster solidarity and cohesion within the reference group itself (Murphy, 1957), and, thus, regional identities surge (Paasi, 2012). Väyrynen interprets identification with the region as a peer group affine form of rejection of globalization and associated modifications of individual conducts of life in the region (Väyrynen, 2003). Hence, a perceived group threat leads to negative attitudes towards members of outgroups as a means to shape regional group identity and to compete for privileges under threat (Semyonov et al., 2006). Woods explains this by taking the example of the reactions of the Lega Nord in Italy in the late 1990s to the constraints of globalization (Woods, 2009). Promoting prejudices towards immigrants 'pays greater dividends for the Lega than other issues that it links to globalization' (p. 172). Anti-immigrant attitudes are driven by individual apprehensions that their personal group 'will be put at systematic disadvantage to the other' (Quillian, 1996, p. 820) immigrant group due to the forces of globalization. In Italy, perceived group threat was regionally instrumentalized by declaring regionalism under siege and was turned against newcomers.
Fitjar (2010) recognizes a fundamental difference between regional identity, which he does not see as positively correlated with anti-immigrant attitudes, and national identity, which is based on cultural factors that are threatened by immigration (Fitjar, 2010). Regional identity is based less on notions of territorially framed cultural homogeneity, this is more the idea of nation, than on certain preferences of lifestyles, which usually fit into the region. This idea is much more open to the outside world than the defence of cultural-territorial homogeneity (Fitjar, 2010). If the self-description of regional culture is openly cosmopolitan, then immigration can contribute positively to this regional culture and, consequently, identity. The connection between regionalism and anti-immigrant attitudes is therefore either negative or non-existent (Fitjar, 2010). Fitjar (2010) suggests that immigration strengthens regional identity and, consequently, anti-immigrant attitudes correlate negatively with regionalism.
Taken together, the discussed literature on regionalism and regional identity reveals that, on the one hand, there are only few empirical studies on the interplay of regionalism and anti-immigrant attitudes. On the other hand, this empirical research arrives at inconsistent results, suggesting both that regionalism and regional identities have either no impact on anti-immigrant attitudes (Curtis, 2014) or a reinforcing one (Escandell & Ceobanu, 2010) and also a positive impact by strengthening a cosmopolitan regional culture (Fitjar, 2010). Hence, the present study is motivated by these inconclusive empirical results. It focuses on correlations between regionalism and anti-immigrant attitudes in a comparative perspective on individual, regional, and country level, taking 29 different countries worldwide into account. The key aim of this approach is to analyse the relationship of regionalism with anti-immigrant attitudes at these different spatial levels, controlled by nationalism. In so doing, comprehensive correlations can be estimated.

Data
The study is based on the International Social Survey Programme -National Identity III (ISSP) for data on the individual scale. ISSP is a collaborative cross-national survey programme covering selected countries in the Global North and South. It has been conducted since 1985, using simple or multi-stage stratified random samples of the respective populations, aged 18 years or older, except for Finland starting with 15 years or older and Japan and South Africa starting with 16 years or older (GESIS, 2015). Data collection employs face-to-face interviews, both as paper and pencil and computer assisted. Additionally, fixed form self-administered and interactive self-administered questionnaires were used as well as computer-assisted telephone interviews in Denmark (GESIS, 2015). The sample sizes vary between 904 for the UK and 2739 for South Africa. The original data file contains 45,297 individuals nested in 33 countries.
The ISSP data were complemented with country-level data on the net migration rate in 2015, data on ethnic fractionalization, the regional authority index, and the presence of an autonomous region in the respective country as well as with aggregated individual scores on regionalism and nationalism on the regional level. The migration variable was obtained from the IOM (2019), the ethnic fractionalization data from Alesina et al. (2003), the regional authority index from Hooghe et al. (2016), and presence of an autonomous region in the respective country was derived from the database of political institutions 2017 .

Analysis strategy
Only cases that include all items from the respective scale of the dependent variable and that are covered in the regional authority index were considered in the analysis. Germany was included twice, differentiated between West and East, due to long-lasting socio-historical differences in attitudes and immigration (Weins, 2011). Hostility towards immigrants shows significantly higher values for East Germany (t-value 4.224; p < 0.001). Thus, the final data set structure is individuals (n = 31,245) nested in regions (n = 421) and nested in countries (n = 30). Regions are presented in Table A1 in the supplemental data online. The number of context units at the country level suggests the use of restricted maximum likelihood (RML) with the Kenward-Roger approximation for optimizing significance tests of fixed effects (Kenward & Roger, 1997) as estimation method. The right number of random slopes in multilevel regression is the subject of controversial debate in the literature on multilevel models (Bryan & Jenkins, 2016;Heisig et al., 2017;Schmidt-Catran & Fairbrother, 2016). To meet the minimum standard formulated in this discourse for a balance between flexibility and economy in the models (Heisig et al., 2017), random slopes were included for education, regionalism and nationalism. The net migration rate is centred on its grand mean; all other interval scaled variables at the country and individual levels are standardized by calculating the z-scores across all countries. This guarantees that the intercept can be understood as an anticipated value of the resulting variable, which means zero has to be meaningful (Hox, 2002). Five multilevel linear models were calculated to assess the impact of regionalism on anti-immigrant attitudes as an individual and macro-predictor. Because the general association between the country and regional level and anti-immigrant attitudes is of specific interest, only fixed effects of the respective models are reported.

Dependent variable
The dependent variable is the individual-level attitude toward immigrants. The ISSP data set includes eight questions concerning attitudes toward immigrants in the respective country (see The impact of regionalism on anti-immigrant attitudes: a multilevel international comparative study Table 1 for the exact wording of the questions). The answers range on a five-item scale from 'agree strongly' to 'disagree strongly'. The answers are recoded in such a way that high figures represent high rejection of immigrants. The final variable is calculated as the sum of these eight items (α = 0.79).

Independent variables
Individual level. As 'people have varied conceptions of the relations between different territorial identities, but can also have a different understanding of the content of those identities' (van der Zwet, 2016, p. 1245), with nationalism and regionalism two different concepts of attachments to place and territory are added. Nationalism is derived as a factor score out of 12 items from an exploratory factor analysis using the principal component extraction method with promax rotation, which means that extracted factors can correlate. The nationalism factor subsumes four items representing a superior position of one's own nation (α = 0.68), namely 'I would rather be a citizen of [COUNTRY] than of any other country in the world'; 'The world would be a better place if people from other countries were more like the [COUNTRY NATIONALITY]'; 'Generally speaking, [COUNTRY] is a better country than most other countries'; and 'People should support their country even if the country is in the wrong' (GESIS, 2015). The factor depicts a chauvinist dimension of nationalism. The response categories range from disagree strongly to agree strongly, whereby the response categories were recoded in such a way that high agreement corresponds with high values. Regionalism is a scale (α = 0.76) subsuming the questions 'How close do you feel to your town or city?' and 'How close do you feel to your county?' with a four-point answer scale. The resulting variable was standardized by calculating the z-scores across all countries.
Controls. Five demographic control variables from studies concerning nationalism, regionalism and anti-immigrant attitudes are added because empirical research reveals an impact of sex, age, education, immigration history and urban/rural living (Ariely, 2011;Bonikowski, 2016;Curtis, 2014;Fetzer, 2000;Pehrson et al., 2009). Sex is indicated by 0 (female) and 1 (male). Missing values for age are imputed with regression imputation (0.3% of values imputed) (Baltes-Götz, 2013). The ordinal scale for education is recoded into 'years in formal education'. For instance, the level 'no formal education' was recoded as zero years of school education and the level 'still in college' was recoded as 14 following Weins (2011). Missing cases were imputed with regression imputation (1.7% of values imputed) (Baltes-Götz, 2013). The five-item scale of location was recoded as a dummy for rural/urban. Immigration background is indicated by 0 (yes) and 1 (no) as individual migration history is a strong predictor for a weaker anti-immigrant attitude (Fetzer, 2000). Fractionalization (z-scores) Regional authority index ( Country level. As the main research purpose is figuring out the relationship between anti-immigrant attitudes and regionalism in particular, four macro-variables are added to control for context effects. As Pehrson et al. (2009) emphasize in their analysis of the relationship between nationalism and anti-immigrant attitudes, both are subject to an endangerment scenario that interprets immigration as a threat to national or regional identity. Immigration is seen as competition for resources (Borjas, 2003) and quality of life in general that results in nationalistic and xenophobic attitudes (Bonikowski & DiMaggio, 2016;Jeong, 2013;Pehrson et al., 2009). Net migration rate and ethnic fractionalization as a proxy for ethnic conflict  control for potential threat effects. The regional authority index and the presence of an autonomous region, which is a part of a country that 'must be constitutionally designated as "autonomous" or "independent" or "special"' (Cruz et al., 2018, p. 20), for instance Scotland (UK), Greenland (Denmark) or Corsica (France) are autonomous regions in this sense, control for the impact of institutionalized regional identity on anti-immigrant attitudes. Regional authority is defined as 'the capacity to make legitimate and binding decisions for a collectivity' (Hooghe et al., 2016, p. 16).
Regional level. To control for Escandell and Ceobanu's (2010) 'individual self-association with regions' (p. 160) approach, which postulates that contexts with high levels of regional identification among the population evoke high levels of immigrant exclusionism, the aggregated nationalism and regionalism scales are added using the 421 country-specific regions defined by the ISSP included in the data set. It is assumed that threat effects foster anti-immigrant attitudes, whereas regional authority and autonomous regions as concessions of national governments have a mitigating effect (Escandell & Ceobanu, 2010). Table 1 summarizes the variables and data sources involved in the analysis.

RESULTS
The descriptive statistics give a first indication of the structure of the data in the sample (Table 2). Russia has the highest average level of anti-immigrant hostility in the sample combined with marked regionalism and nationalism. The highest value on the nationalism scale is found in Turkey, followed by Japan and the Philippines. In order to scrutinize the relationship between anti-immigrant attitudes and regionalism, five multilevel regression models were calculated, as displayed in Table 3. First, an empty model was calculated that only contains an intercept and error terms for the individual, regional and national level of the variance in the anti-immigrant attitude scale. This baseline model serves as a reference for calculating the explained variance (R²) for all following models. It reveals that 10.0% of the variance in the anti-immigrant attitude scale can be traced back to differences between the countries and 4.5% to differences between the regions. This amount of explained variance at the macrolevel warrants the usage of multilevel regression for investigating the correlation structure of anti-immigrant attitudes and regionalism in a comparative perspective.
Model 1 contains the control variables for demographics on the individual level and threat effects on the national scale, explaining 13.5% (R²) of variance overall. As its most significant result, model 1 replicates the finding that being male, higher aged, with lower education and rural living reinforce anti-immigrant sentiment (Ariely, 2011;Bonikowski, 2016;Curtis, 2014;Pehrson et al., 2009), whereas immigration history has a mitigating impact on anti-

TERRITORY, POLITICS, GOVERNANCE
The impact of regionalism on anti-immigrant attitudes: a multilevel international comparative study immigrant attitudes (Fetzer, 2000). The two country-level proxies for threat effects are either non-significant (net migration rate) or reinforce these attitudes (fractionalization). Migration figures as opposed to demographic controls cannot contribute significantly to hostile attitudes yet. Model 2 tests the baseline effect of regionalism at the individual and regional level. This improves the explained variance of the model by about 3% (R² = 16.5% overall), indicating the limited significance of regionalism for explaining anti-immigrant attitudes. Both independent variables of regionalism at individual and regional level are not significant in accordance with Curtis's results for Europe (Curtis, 2014). Demographics remain important as in model 1.
In model 3 the dummy for the presence of an autonomous region and the regional authority index are added at the country level as well as the threat effects on the national scale, which improves the model fit only slightly (R² = 17.4%). As expected, high ethnic fractionalization fosters anti-immigrant attitudes, whereas a high score on the regional authority index has a mitigating effect (Escandell & Ceobanu, 2010). Regional authority as 'the capacity to make legitimate and binding decisions' (Hooghe et al., 2016, p. 16) at the regional level correlates with low antiimmigrant sentiment. The granting of regional autonomous possibilities for political decisions in a nation-state seems to support amicable attitudes towards outsiders. The index is highly significant, unlike the simple presence of an autonomous region. Model 3 also confirms the results of the previous model 2, revealing that immigration background is the strongest predictor at the individual level.
Model 4 controls for the impact of nationalism (R² = 25.4%), revealing that nationalism is significant even when aggregated at the regional level, while regionalism is not. The impact of nationalism on anti-immigrant attitudes in the data is, thus, much stronger than regionalism even at the regional level, suggesting that a nationalistic mindset opposes immigration strongly, because nationalism is more exclusive in the assumption of the superiority of one's own nation. Regionalism as pure attachment to a region does not show this exclusionary side.

DISCUSSION
The results of the multilevel regression models arrive at separate effects at the individual and aggregated levels. The regional authority index in a nation-state shows a significant mitigating impact on anti-immigrant attitudes at the macrolevel, but regionalism has no significant relationship with anti-immigrant attitudes at the individual and regional levels. Moreover, regionalism and nationalism behave in a divergent fashion at the individual and regional level in international comparative perspective. While nationalism in chauvinist variant as belief in the superiority of one's own nation is associated with anti-immigrant attitudes in a reinforcing manner at both levels, regionalism is unassociated.
The comparative perspective across 29 different nation-states makes clear that demographic controls of age, education, and location contribute to a systematic explanation of the variance in the anti-immigrant attitude scale. The theoretically expectable effect of age, representing life experience (Dustmann & Preston, 2001), on anti-immigrant attitudes is a boosting one, because higher age means lower availability of life alternatives (Stichweh, 2017) that causes more prejudice (Tir & Singh, 2015), and can be found in the models. However, Stawarz and Müller (2019, p. 17) find for Germany that anti-immigrant attitudes tend to decrease with 'increasing age'. For East (r = 0.19; p < 0.01; [CI = 0.095-0.278]) and West Germany (r = 0.08; p < 0.05; [CI = 0.019-0.147]) the correlation (Pearson) between age and anti-immigrant attitude is positive, immigrant rejection in Germany therefore increases with age in the data. Education mostly has a mitigating influence on anti-immigrant attitudes, especially in richer societies in the West. Education is an important asset in these societies when competing for jobs. Yet, most immigrants do not have a university degree, or their degrees are not recognized, which means they are excluded from better Table 3. Three-level regression models predicting anti-immigrant attitudes.
The impact of regionalism on anti-immigrant attitudes: a multilevel international comparative study Table 3. Continued.   (2015) jobs. Thus, higher education correlates with low anti-immigrant attitudes especially in Western and Northern societies (Fetzer, 2000). Moreover, schooling leads to a perpetuation of values regarding openness as much as to contacts between unequal people. Thus, in the four models, education mitigates anti-immigrant attitudes. Immigration history is the most important demographic predictor for lower levels of anti-immigrant attitudes in the models. Fetzer also reports a mitigating effect of immigration history for France, Germany and the United States (2000). Especially in social contexts that are shaped by high immigration figures in combination with high pressure to achieve like Singapore, this effect of personal migration experience on mitigating immigrant prejudice is important (Dirksmeier, 2020).
Differences between the country samples explain 10.0% of the total variance. According to these results, fractionalization as a proxy for ethnic conflict fosters anti-immigrant attitudes, while immigration figures do not, revealing the contextual composition of anti-immigrant prejudices in general (Pottie-Sherman & Wilkes, 2017). Migration figures and ethnic fractionalization can be regarded as a proxy for perceived threats (Laurence, 2014) as opposed to real threats, for instance decreasing salaries due to a newly established pool of immigrant workers willing to work for lower wages (Borjas, 2003). Perceived threat evokes symbolic boundary-making (Bail, 2008). Members of the majority interpret common virtues, like religion or ethnicity, as symbolic boundaries that are useful for defending privileges (Bail, 2008) by establishing stakeholder groups that are directed against newcomers. Anti-immigrant attitudes could, thus, be a consequence of a perceived threat scenario that is uncoupled from empirical facts like decreasing wages or unemployment, and rather directed towards an abstract category of commonality, which is traceable cross-culturally.
Nationalism correlates significantly with anti-immigrant attitudes on the individual and regional level, replicating common findings in comparative international work (Ariely, 2011;Bonikowski & DiMaggio, 2016;Gordon, 2017;Heath et al., 2019;Pehrson et al., 2009;Salari Rad & Ginges, 2018). German sociologist Rudolf Stichweh explains this prominent interrelation of anti-immigrant attitudes and nationalism by understanding citizenship of all legal members of a nation as a resource for the individual (Stichweh, 2000). Anti-immigrant attitudes become explainable as a recourse to this imagined civic equality of all members of a nation, in order to retain the privileged life situation of the legal citizens of the nation (Stichweh, 2000). Citizenship thus becomes a symbolic boundary for nationalists (Bail, 2008), who see their nation as supreme. Stichweh (2000) combines the excluding and comparing character of nationalism (Hobsbawm, 1992) with a perceived threat scenario (Stichweh, 2000), which makes the impact of nationalism at the individual and regional level in the models plausible.
Although regional scientists perceive a danger for social cohesion in regional disparities in economic development, especially in Europe (Iammarino et al., 2019) and, therefore, accentuate the meaning of the regional scale, regionalism at the individual and regional level does not correlate with immigrant hostility. Contrary to nationalism, regionalism is not associated with material resources of a particular kind, like for instance citizenship or a social welfare system. In line with this assertion, Williams explains the first appearance of the term regionalism in the English language closely adjunct to an incompleteness of centralization and state formation (Williams, 1985). Regions do not bestow citizenship or offer social welfare. Regionalism, thus, lacks important material resources beyond identity and belonging that are worth fighting for. Moreover, regional boundaries are contingent and fluid and mostly not congruent with ethnic boundaries (Hjerm, 2007). According to Brubaker's (2009, p. 28) concept of 'groupism', social groups identify with regions but commonly in a loose way, which does not affect individuals' attitudes as enunciated in the lack of significance of the regionalism item in the models.
On the contrary, regionalism correlates with anti-immigrant attitudes at the regional level (r = 0.25; p < 0.01; [CI = 0.244-0.265]). But the regional authority index as representing the 'capacity to make … binding decisions' (Hooghe et al., 2016, p. 16) in regions in a nation-state mitigates anti-immigrant attitudes. This indicates the state-specific meaning of regionalism. Following Paasi (2003) and Johnson and Coleman (2012), regions as social constructs are shaped by social processes (Paasi, 2003;Johnson & Coleman, 2012), which can be interpreted as mostly contested. The most extreme form of this contestation is separatism (Knight, 1982;Sindre, 2018). The multilevel regressions reveal that regionalism in the sample is rather an expression of a specific attachment to place, which is not exclusionary. This finding is different from the 'individual self-association with regions' approach of Escandell and Ceobanu (2010, p. 160), which predicts that 'strong regionalist identifications … would report more immigrant exclusionism' (Escandell & Ceobanu, 2010, p. 174) and is partly confirmed in their study conducted in Spain. However, the presence of political regional authority can be interpreted as a concession of the state towards regional ties. This concession correlates with lower anti-immigrant attitudes. The institutional support of regional decision-making power fosters a more openminded attitude towards newcomers in the national population. Far-reaching regional possibilities to decide instead of regionalism mitigates anti-immigrant attitudes cross-culturally.

CONCLUSIONS
Regionalism does not promote anti-immigrant attitudes. Moreover, institutionally granted rights to regional decision-making even mitigate these hostile mindsets. These findings cannot replicate assumptions from potential threat scenarios for regional identity (Escandell & Ceobanu, 2010), but unveil a demarcation between the individual, regional and country level, confirming the application of multilevel regression. Subsidiarity supports a positive attitude towards immigrants. The difference between chauvinistic nationalism, which goes hand in hand with a hierarchization of nations, and regionalism as a positively conceived attachment to one's own (home) region becomes very clear in the models. This is the most important finding of the study. However, unsatisfactorily little is known about the effect mechanism of institutionally granted regional political authority on openness to immigrants. This is also a methodologically reasoned shortcoming of this study. These possible symbolic and practical mechanisms are hard to capture with quantitative models and call for qualitative research that is able to explain the structure and reasons for this interdependency. Like Antonsich's (2018) work on everyday uses of nationalism in social interactions, regional research needs to employ case studies to scrutinize exactly how regional political authority is expressed in everyday life and, subsequently, what the reasons are for its mitigating impact on anti-immigrant attitudes. As people live first and foremost in 'their' region, the regional scale as a specific expression of place is of particular importance.
In view of the fact that 'regional dynamics' (Paasi, 2009), and a 'programmatic shift toward regionalism' (Sindre, 2018, p. 23), right up to secessionism, are still current in today's societies, Applegate's call for 'a renewed engagement with the regional level of experiencean engagement sensitive to the interactions of society, identity, and place' (Applegate, 1999(Applegate, , p. 1182) must still be taken seriously. This should result in research efforts to tackle attitude/place relations in their widest sense (Berger & Engzell, 2019).