How “Us” and “Them” Relates to Voting Behavior—Social Structure, Social Identities, and Electoral Choice

Abstract

The last decades have seen the emergence of a divide pitting the new left against the far right in advanced democracies. We study how this universalism-particularism divide is crystallizing into a full-blown cleavage, complete with structural, political and identity elements. So far, little research exists on the identities that voters themselves perceive as relevant for drawing in- and out-group boundaries along this divide. Based on an original survey from Switzerland, a paradigmatic case of electoral realignment, we show that voters’ “objective” socio-demographic characteristics relate to distinctive, primarily culturally connoted identities. We then inquire into the degree to which these group identities have been politicized, that is, whether they divide new left and far right voters. Our results strongly suggest that the universalism-particularism “cleavage” not only bundles issues, but shapes how people think about who they are and where they stand in a group conflict that meshes economics and culture.

Introduction

Major disruptive shifts in electoral politics across advanced democracies—the rise of right-wing populism and the fragmentation of the mainstream left and right in many European countries—continue to intrigue political scientists: debates over whether the drivers of these changes are economic or cultural are by now giving way to discussions of how the two might interact (e.g., Cramer, 2016; Gidron & Hall, 2017; Kurer, 2020). In this paper, we show that integrating social identities into the study of electoral politics offers a comprehensive framework for studying the interplay of “economic” and “cultural” drivers of electoral behavior, and specifically of realignment with far right and new left parties. While socio-structural circumstances are relevant, their link to electoral behavior is less straightforward than narrow political economy models would have us think. Rather, individuals subjectively interpret their objective life conditions, and the ensuing group boundaries mesh economic and cultural elements. Furthermore, these interpretations need to be politicized to matter electorally. Hence, we need to know how voters belonging to particular socio-structural groups depict group boundaries between “them” and “us,” which in turn underlie the divide between the new left and the far right. Here, we study how objective socio-structural categories and subjective group identifications relate to each other and to electoral choice.

There is abundant evidence that objective social structural location continues to matter for electoral preferences even after the decline of class conflict, supporting the theory of electoral realignment rooted in an evolving social structure. Parties of the new left and the far right are located at opposing poles of a new divide that crystallized in the 1980s and 1990s throughout Western Europe (Bornschier, 2010; Hooghe & Marks, 2018; Rovny & Polk, 2019)—labeled here universalism-particularism divide. These terms acknowledge that this divide, while initially centering heavily on issues such as cultural liberalism and immigration, has also come to incorporate distributive preferences (Häusermann & Kriesi, 2015). This emerging divide is linked both to subjective perceptions of deprivation and status loss (Burgoon et al., 2019; Elchardus & Spruyt, 2012; Gidron & Hall, 2017, 2020; Kurer, 2020), as well as to objective socioeconomic positions. Regarding the latter, political sociologists have amply shown that the voters of new left and far right parties are characterized by specific socio-structural attributes. In particular, support for far right parties is concentrated within the manual working class and among those with intermediate levels of education. “New left” parties, on the other hand, are disproportionately supported by socio-cultural specialists, that is, qualified employees working in client-interactive settings (Gingrich & Häusermann, 2015; Häusermann & Kriesi, 2015; Oesch, 2013; Oesch & Rennwald, 2018). The urban-rural divide is also crucial for understanding support for far right versus new left actors and positions (Iversen & Soskice, 2019; Maxwell, 2019). These clear structural foundations of electoral alignment led many studies to infer voters’ electoral motives directly from their material life conditions in increasingly knowledge-based societies (see Iversen & Soskice, 2019; Manow, 2018 for two recent examples).

With the data typically drawn from large-scale surveys, however, we can only identify the socio-demographic profile of new left and far right electorates, as well as their attitudes, but we are unable to grasp subjective politicized group identities that underlie and stabilize electoral realignment. These identities are crucial, because they ultimately inform the programmatic demands these electorates have and the appeals they are likely to respond to (Huddy, 2001; Stubager, 2009). There is ample reason to think that voters’ self-identification does not simply mirror their ascriptive characteristics. The literature commonly characterizes voters of the far right as “losers” of modernization” (Betz, 1994), “low/medium educated” (Stubager, 2010), “(relatively) deprived” (Burgoon et al., 2019; Gidron & Hall, 2017; Gidron & Mijs, 2019; Kurer, 2020), “structurally threatened” (Mutz, 2018), or experiencing “declinism” (Elchardus & Spruyt, 2016). Political psychology would have us expect, however, that individuals construct their identities in more positive terms. Hence, our ascriptive categories may not grasp the social identities these respondents would name as relevant.

In this article, we therefore study voters’ own subjective perceptions of the social and political world surrounding them, and—in a second step—explore the politicization of these perceptions and how they contribute to predicting vote choice. To integrate social identity research into our study of partisan divides, we combine insights from social cleavage theory—which understands salient socio-political divides as being rooted in social structure, but translated into politics via collective identities—with applications of social identity theory in political settings (Helbling & Jungkunz, 2020; Huddy, 2001; Mason, 2018), and recent ethnographic studies (Cramer, 2016; Hochschild, 2016), which focus more on individual perceptions of social group belonging. We ask three main questions. First, how do key socio-structural variables (education, class and rural/urban residence) relate to social identities? Second, how do voters of new left and far right parties differ with regard to the social identities that are salient to them? And third, to what extent do subjective social identities mediate the relationship between (objective) socio-demographic categories and the vote choice for these parties?

We address these questions using data from an original public opinion survey, in which we extend established methods of survey-based comparative electoral research to encompass voters subjective group identities. The survey was fielded in German-speaking Switzerland, where electoral realignment has been underway for several decades, with the new left and far right today constituting the strongest party blocks. Hence, we look at a paradigmatic case of realignment in which the universalism-particularism divide may well have attained the quality of a fully-formed cleavage.

The article is structured as follows: the theory section explains why an appropriate understanding of electoral change requires the explicit integration of social identities. After describing the design and data, we discuss our findings in three steps. We first show how key socio-structural attributes relate to social identities. We then present evidence of party electorates’ subjective identities to single out those identities that have indeed been politicized. In a final step, we predict party preference in a multivariate model, including both objective and subjective group belonging. We conclude by reflecting on the significance of our findings for how we should think about long-term electoral transformations. Our results not only support theories of electoral realignment, they further suggest that realignment has led to the crystallization of distinctive, diametrically opposed collective identities. The universalism-particularism divide—beyond being a conflict over new issues—has come to structure how people think about who they are and about where they position themselves in an emerging group conflict that meshes economics and culture. This implies a durable new conflict structure, which may well underlie changing electoral dynamics in other advanced democracies, as well.

Theory

Social Structure, Social Identities, and Changing Cultural Conflicts

Successes of the far right in Europe and the US have reinvigorated inquiry into the social bases of the right-wing populist backlash against the new left. Ethnographic research (e.g., Cramer, 2016; Gest, 2016; Hochschild, 2016; Wuthnow, 2018), the political psychological literature (e.g., Huddy, 2001; Klandermans, 2014; Mason, 2018), and the literature on populism (e.g., Akkerman et al., 2017; Elchardus & Spruyt, 2016; Spruyt et al., 2016) all suggest that social identities are important for explaining recent electoral outcomes. Subjective identities affect vote choice both via informing programmatic policy demands, and via non-programmatic mechanisms such as group norms. Even if there is widespread evidence that for example, social class and education continue to matter for vote choice, neglecting subjective social identities that link social structure and partisan identities prevents us from understanding electoral choices (Huddy, 2001). Why do many far right voters doubt climate change, reject gay rights, or oppose “big government”? These non-obvious stances need to be understood via voters’ perceived group interests, which might be non-programmatic (e.g., demarcation from outgroups), as well as programmatic. Both types of interests are rooted in voters’ social identities as members of politically relevant collectives.

We contend that these insights complement the fundamental tenets of social cleavage theory, which acknowledges the crucial role of collective identities underlying both economic and cultural cleavages (Bartolini & Mair, 1990; Stubager, 2009). It posits that collective identities—as the link between social structure and politics—are central both for overcoming the collective action problem and in accounting for the durability of cleavages. What is more, cleavage theory in the tradition of Lipset and Rokkan (1967) conceives of the political space as multidimensional. It thus offers the ideal historical macro-sociological framework for studying how group identities render certain divisions more salient at the expense of others. In practice, both classic and newer cleavage research has tended toward equating group identities either with socio-structural categories (Kriesi et al., 2008; Lipset & Rokkan, 1967) or with more strictly political identifications (Bartolini & Mair, 1990). The most successful attempts to conceptualize cleavage identities as occupying a middle ground have borrowed from related strands of research, notably social identity theory (see Bartolini, 2005; Stubager, 2009).

Complementing cleavage theory, approaches from social psychology offer a theoretical micro-foundation for understanding why identities have psychological value, why they may foster affective dimensions of voting, and why they affect behavior. Social identity theory (and self-categorization theory) argues, first, that group identification is driven by a desire to positively distinguish one’s in-group (Tajfel, 1981; Tajfel & Turner, 1979). Second, individuals harbor multiple, potentially conflicting identities, the salience hierarchies of which are to some extent malleable; and third, salient social identities are crucial for shaping behavior (see Haslam et al., 2011; Huddy, 2001; Stubager, 2009). The fact that social identities shape behavior is hardly surprising in light of their psychological functions: Social psychology emphasizes that group membership entails feelings of belonging, participation in a shared reality, and group-related emotions (Baumeister & Leary, 1995). These may be feelings of pride and deservingness (e.g., Slothuus, 2007; Van Oorschot, 2006), or of resentment and deprivation (Elchardus & Spruyt, 2012; Gidron & Hall, 2017; Teney et al., 2014). The more identities are aligned (as opposed to cross-cutting), the more likely they are to foster affective polarization (Roccas & Brewer, 2002). Such affective group ties shape political behavior via, as well as beyond rational programmatic positions, because they reflect individuals’ desire for self-esteem. Furthermore, shared group norms set parameters for what is appropriate in a specific group (evident from ethnographic research describing “normative communities” within advanced societies, e.g., Lamont, 2000; Wuthnow, 2018). Group conformity and norm adherence have been shown to be powerful motivators of behavior (Akerlof & Kranton, 2010; Cialdini & Goldstein, 2004).

Beyond social identity theory and social psychology, recent sociological and ethnographic studies have revealed the crucial political significance of identities. Even so-called “losers” of economic and social change have distinctive, often positively connoted understandings of group belonging, having to do with such things as the arduousness of their work, adherence to moral standards providing non-economic definitions of success, or with their geographical distance to urban centers of power and prosperity (Cramer, 2016; Hochschild, 2016; Lamont, 2000; Wuthnow, 2018). These, as well as important sociological studies of changing class structure in the 21st century and related identities (Savage, 2015; Savage et al., 2013; Vester et al., 2001)—many of which build on Bourdieu’s sociology—highlight social markers and milieus that might be increasingly relevant for understanding contemporary electoral politics. Bourdieusian cultural sociology and ethnographic research more broadly is equipped to study the processes of boundary drawing and social closure in ways that neither work in cleavage theory nor in social identity theory have been able to do. They provide valuable indications of specific group boundaries that are appropriated and constructed by individuals. While often milieu-specific, these studies provide many insights that can be integrated into the wider framework of political sociology that cleavage theory offers. They show that if we really want to comprehend what voters want (and what resonates with them), we have to develop an understanding of who they are in their self-identification. Without this effort, we cannot understand their electoral motives (affirmation of group membership or demarcation from other groups vs. e.g., protest voting, issue-voting or personalized electoral choice), assess the durability of electoral realignment, or understand which programmatic appeals and policy decisions are likely to resonate with different electorates.

In this paper, we apply these insights to the study of the electoral divide between the new left and the far right in Europe. Specifically, existing research in the cleavage tradition suggests that shifts of aggregate lower-class support from the left to far right parties, or upper middle-class support for left-wing parties reflect mobilization on the basis of new collective identities. Since 1968, the traditional cleavages characteristic of the Lipset and Rokkan (1967) world have been complemented—and to some extent transformed—by new cultural divisions in advanced democracies. In a first wave of mobilization, the new social movements of the late 1970s and 1980s found expression in the emergence of the Green party family, as well as in a new left transformation of many established Social Democratic or Socialist parties (Inglehart, 1984; Kitschelt, 1994; Kriesi, 1998). Politicizing issues such as gender equality, the free choice of lifestyles, or solidarity with the Third World, this “new” left advocated a strong equality principle rooted in universalistic values. Subsequently, and to a large extent in reaction to the agenda of the new left, the far right put issues of community and national sovereignty on the political agenda (Betz, 2004; Ignazi, 1992). These issues represent polar normative ideals to the universalism of the new left (Bornschier, 2010). They find expression in the politicization of immigration, as well as in the insistence on the primacy of democratic majority decisions over rulings of courts or supra-national bodies (Hooghe & Marks, 2018).

After the first transformation of cultural conflicts by the new left had given rise to a libertarian-authoritarian divide (Flanagan & Lee, 2003; Kitschelt, 1994), this mobilization of the far right reshaped cultural conflicts yet again. The resulting divide has been variously labeled as opposing Green-alternative-left (GAL) and traditional-authoritarian-nationalist (TAN) positions (Hooghe et al., 2002), libertarian-universalistic and traditionalist-communitarian values (Bornschier, 2010), or as a divide exhibiting distinctive “grid” and “group” components (Kitschelt & Rehm, 2014). Finally, scholars emphasizing the struggle over borders prefer the terms “integration-demarcation” (Bartolini, 2005; Kriesi et al., 2008) or “cosmopolitanism-communitarianism” (de Wilde et al., 2019; Strijbis et al., 2020). Here, we use the terms “universalism versus particularism” (Beramendi et al., 2015), since it remains an empirical question whether the new divide is exclusively cultural in nature, given that recent research shows that attitudes regarding some social policies align more with socio-cultural conflicts than with the traditional economic state-market cleavage (Häusermann & Kriesi, 2015). This is particularly relevant for the present context because we are interested in subjective group boundaries that may combine cultural and economic elements to shape partisan alignments.

Indeed, even if the new conflict has for the most part been characterized as cultural, the structural potentials that nourish it have been variably linked to cultural as well as economic modernization, and to the multi-faceted process of globalization (see Bornschier, 2018 for a review). Correspondingly, an extensive literature has analyzed the social structural basis of new left and far right parties, most notably class (Evans, 1999; Kitschelt, 1994; Kriesi, 1998; Kriesi et al., 2008). The finding that the manual working class is over-represented within the electorate of the far right is very robust (Bornschier & Kriesi, 2013; Kurer, 2020; Minkenberg & Perrineau, 2007; Oesch, 2013; Oesch & Rennwald, 2018). Following Allardt (1968), others have delved into the role of higher education in fostering new group divisions both in terms of values and group identity (Ivarsflaten & Stubager, 2013; Stubager, 2008, 2009, 2010; Waal et al., 2007). In a pioneering study, Stubager (2009) revealed that in Denmark, educational groups exhibit antagonistic collective identities related to their educational achievement. This suggests that education indeed has the potential of becoming part of a full-fledged cleavage in Bartolini and Mair’s (1990) sense.

Beyond education and class, the spatial foundations of political divides have recently received greater attention, as divergences between “cosmopolitan” cities and “nationalist” towns and rural areas become apparent (e.g., Iversen & Soskice, 2019; Maxwell, 2019, 2020). This urban-rural divide is discussed mainly in light of territorial disparities in prosperity that emerge in knowledge-based economies (Hobolt, 2016; Jennings & Stoker, 2016; Rodríguez-Pose, 2018), but important research in sociology and urban studies also indicates the emergence of cultural identities based in rural communities (Cramer, 2016; Wuthnow, 2018) or rather cosmopolitan urban environments (Cunningham & Savage, 2015; Florida, 2012; Savage et al., 2018).

While we thus know that class, education and urban-rural residence matter for the vote along the universalism-particularism divide in objective terms, it is not clear whether these categories are salient for voters themselves. Voters usually do not self-identify as “low-educated” or as “modernization loser.” How do voters themselves depict the group boundaries between “them” and “us” that underlie the universalism-particularism divide? We hypothesize that voters indeed perceive themselves as members of their “objective” groups. Hence, we do not think that social identities are entirely unrelated to structural foundations. However, we suggest that the link between these structural foundations and other, more culturally connoted social groups should be at least as strong. In other words, even non-economic, cultural identities should be structurally rooted. Showing this link should help us overcome the false dichotomy between economic and culturalist explanations of electoral preferences.

Social Identities and Electoral Preferences

After theorizing the link between structure and identities, a second step in relating structural electoral potentials to actual electoral outcomes lies in the politicization of these identities. We conceive of the link between collective identities and parties as interdependent. In other words, we do not make a directional argument in favor of a strong bottom-up versus top-down mechanism of group identity formation. In line with the argument by Bornschier (2010, see Figure 1 below), we would think that both mechanisms are plausible and at work. ¹ The interaction of top-down and bottom-up processes is the reason why the politicization of social identities is a separate analytical step from studying the structural roots of these identities. ²

Figure 1.

Theoretical framework (adapted from Bornschier, 2010).

Which social identities would we expect to matter? Substantively, we seek to tap into respondents’ sense of belonging to social categories already identified as electorally relevant in existing research (e.g., education, nationality, nature of work, or residence in cities vs. peripheral areas). Thereby, in this paper we cover social groups commonly associated with the economic interpretations of the divide (such as educational and occupational groups), as well as groups commonly associated with the cultural interpretation of the divide (such as nationalism vs. cosmopolitanism).

People might identify with their educational group, as education has been robustly shown to shape where people position themselves along the universalism-particularism divide (e.g., Ivarsflaten & Stubager, 2013; Kriesi et al., 2008; Stubager, 2008, 2009, 2010; Waal et al., 2007). Further, far right parties recruit their voters over-proportionally from people with intermediate levels of education (i.e., vocational training) in manual occupations (e.g., Bornschier & Kriesi, 2013). Class is hence also likely to be relevant as a source of identity. Class schemes developed for advanced post-industrial societies refine vertical divisions by drawing horizontal distinctions based on a differentiation between organizational, technical, and interpersonal work logics (Kitschelt & Rehm, 2014; Kriesi, 1998; Müller, 1999; Oesch, 2006). As the final set of potential social identities directly related to social structure, we ask respondents about closeness to people in urban and rural environments. The antagonism between urban centers and the rural periphery has been highlighted especially by ethnographic research (Cramer, 2016; Hochschild, 2016; Wuthnow, 2018).

Beyond identification with these straightforwardly structural groups, we investigate potential group identities more strongly related to the universalism-particularism conflict itself: Moving one step up the ladder from rural and urban identities, this primarily cultural divide is intimately related to identification with international communities (“cosmopolitans”) versus identification with members of the nation state (“nationals”) (Hooghe & Marks, 2018; Inglehart, 1977). With respect to migration as a key issue related to the universalism-particularism divide, we are interested in how close respondents feel to “people with a migration background.” We expect group belongings at the opposing universalistic pole to be structured not only by cosmopolitanism, but also by other elements of the left-wing liberal (upper) middle-class milieu. Inspired by Bourdieu’s (1979) sociology of “distinction” and by more recent investigations of transforming “cultural capital” that build on it (Flemmen et al., 2019; Friedman & Reeves, 2020; Savage et al., 2013), we probe into this milieu by asking about closeness to “culturally interested people,” in terms of a lifestyle-related group identity. Finally, we include “men” and “women” to cover gender identities or roles, the politicization of which was the core element of the women’s movement that fed into the transformation of the left. The overall hypothesis underlying the role of these culturally connoted identities is straightforward: We expect them to be relevant in mediating the effect of “objective” structural group membership on electoral preferences. ³

Case Selection, Data and Measurement

Economic Realignment in Switzerland: Context and Expectations

Oesch and Rennwald (2018) show that partisan competition in Western Europe has become tripolar, with the left advocating statist and culturally liberal, universalistic positions, the parties of the traditional moderate right advocating relative fiscal conservatism, and the radical right offering a socio-culturally conservative, communitarian program to voters. Each of the three poles mobilizes a distinctive socio-structural “electoral stronghold.” For this reason, the authors expect this tripolar order to reflect a new, relatively stable configuration in most countries where the far right has been in parliament for several terms. Importantly, the main antagonism in this tripolar space is between the left’s cultural liberalism and the radical right’s national-conservatism. This is consistent with findings by Rovny and Polk (2019) on the rising importance of the universalist-particularist divide especially in continental Europe.

Switzerland constitutes a model case of such an electorally realigned, tripolar political space. The country witnessed a comparatively early and strong transformation of the traditional conflicts the party system was built upon, similarly to what occurred in France with the transformation of the Parti Socialiste and the early emergence of the Front National (Kriesi et al., 2008, pp. 98−101). For our purposes, it thus constitutes a most likely case in which group identities that underlie the universalism-particularism divide should indeed be salient. This divide is particularly strong in Switzerland: for one, the new left impetus was forceful. While it initially resulted in the formation of a number of new parties, such as the Greens, an established party, the Social Democrats, ultimately absorbed much of this electoral potential (Ladner, 2007). In other words, the Greens and Social Democrats, which together rallied almost a third of the vote in the last parliamentary elections in 2019, both adopt staunchly universalistic value positions (Nicolet & Sciarini, 2010). As a result, the Swiss Social Democrats have also lost more of their working-class support than the mainstream left elsewhere (e.g., Rennwald & Evans, 2014). Hence, although an established party, the Swiss Social Democrats have over time become decidedly “new left” in terms of their program and voter base. On the other hand, Switzerland also harbors the strongest far right party in Western Europe. The Swiss People’s Party (SVP) reached its peak in 2007 at almost 30% of the vote, and has remained more or less stable at that level since. Similar to what the Social Democrats did on the left, the SVP also meshed several minor parties and organizations into a broad movement. The party and its voters’ programmatic preferences are very similar to those of far right parties elsewhere in Western Europe (Lachat, 2008; McGann & Kitschelt, 2005; Skenderovic, 2009). Like other exponents of the far right, the SVP argues that the new left’s universalistic convictions clash with established cultural practices, and opposes immigration and cooperation with the EU. The parties of the traditional center-right—notably the Liberals and the Christian Democrats—are caught between the two poles of the universalism-particularism divide. Not least because they have found it difficult to define their position with respect to the new divide, they have lost considerable shares of their electorates.

The programmatic profiles of the parties and their configuration vis-à-vis each other make Switzerland a model case of electoral realignment. Hence, we expect group categories related to nationality and origin, education, urban-rural residence, work logic and cultural lifestyle to be important aspects of (antagonistic) collective identities held by the electorates of the new left and the far right.

Data, Concepts and Measurement

Integrating social identities into the study of electoral behavior requires that we combine data on voters’ party preferences and sociodemographic characteristics with data on their identities. To this purpose we implemented an online survey including standard measures of party preferences and socio-demographics, and measures of subjective social identities and their salience. In the introduction to the questionnaire, respondents were told that the survey was about social groups and politics in Switzerland, and prompted with the statement that everyone has different ideas about the groups that make up society. The survey was conducted by a social research company (GfK) in the German-speaking part of Switzerland in September 2018. The sample, which includes 1000 completed interviews, is representative of German-speaking Swiss citizens including quotas for education, age, and gender. In what follows, we describe the most important concepts and the measures of social identity used in the analysis.

In a series of closed-ended questions, we ask respondents about perceived closeness to a number of social groups (“How close do you feel to the following groups? By “close” we mean, who is likely to resemble you with regard to their attitudes, circumstances, and sentiments?”). While cleavage theory guides our selection of group categories and ethnographic research informs our wording of them, we draw on social psychology for our measure of identity. This survey item is adapted from work guided by social identity theory (Mason & Wronski, 2018). We consider it a good starting point to draw on such a tried and tested item as we strive to extend research in cleavage theory to encompass group identities. While our measure may be extended to capture more demanding aspects of identification in future research (such as group consciousness, intergroup conflict or emotional connection), in this survey, we prioritized including a breadth of potential categories of identification over measuring such various dimensions of identity (in contrast to e.g., Stubager, 2009). Especially in a context where identities may only be taking shape, the closeness item seems suitable for gaining a basic understanding of “who voters are” in their subjective self-understanding. Finally, additional results presented in Appendix 1 underline that the measure we use indeed taps identification with groups, rather than attitudes. While our group identification variables correlate with certain issue preferences in expected ways, the correlations remain very moderate. For example, identification with cosmopolitan people is uncorrelated with EU support, while the correlation between identification with people with a migrant background and anti-immigration attitudes is rather low at −0.29. In other words, migration attitudes are not entirely explained by the closeness/remoteness that respondents feel toward individuals with a migrant background. This suggests that while group identifications play a role in linking socio-structural positions to partisan alignments, they are conceptually and empirically distinct from attitudes.

The main results shown in this paper are based on this closeness question, in which we presented voters with a list of 17 specific groups (in randomized order), asking them to indicate how close they feel to each group on a four-point scale ranging from “very close” to “not at all close.” Respondents were presented with the following selection of groups (theorized above):

• Education: We use three categories, namely people with higher education, people with vocational training, and people without a degree.

• Occupational class: To tap into the traditional vertical class dimension, we asked respondents how close they felt to wealthy people, people from the middle class, and people with humble financial means ⁴ We then went on to selectively tap into work logics, asking about people who work with and for other people (i.e., people with an interpersonal work logic, which are heavily over-represented in the new left electorate), people who produce a concrete product in their job (probing into the group identifications that may underlie the well-known over-representation of manual workers among the voters of far right), as well as (more generally) people with a similar job as you have.

• Rural/urban residence: We asked respondents how close they felt to people in the countryside and urban people.

• Universalism-communitarianism: We use three categories to measure identities linked to this aspect of the universalism-particularism divide, namely, Swiss people, people with a migration background, and cosmopolitans.

• Milieu theory: We probed into this by asking about culturally interested people, in terms of a lifestyle-related group identity.

• Gender: Men and women cover gender roles.

This choice of categories is not only theoretically grounded, but also validated by two separate analyses. One validation refers to a separate survey question on how “important” different socio-structural attributes are to respondents. This battery—already in the pretest—confirmed the relevance of, for example, education, class, residence, or gender as relevant categories, but also confirmed that religion, language or age were less important attributes, which we thus did not pursue. Second, we analyzed voters’ responses to open-ended questions about perceived in-groups and out-groups, asked at the very beginning of the survey (Zollinger, 2020). Unprompted, respondents refer to educational groups (“students,” “interested,” “educated”), occupation or work more broadly (“hardworking,” “self-reliant”), residence (“urban,” “countryside”), universalism-communitarianism (“Swiss,” “down-to-earth,” demarcation from “foreigners” vs. “open/cosmopolitan”), and in various ways also to lifestyle or milieu (“normal,” “simple,” vs. “adventurous/enterprising,” “unconventional,” “open”). These open-ended questions also provide insights into how respondents interpret categories that we formulated rather openly in our group battery, notably cosmopolitans and culturally interested. We expected the former category to be associated with a self-image as broad-minded, globally oriented, and open to diversity; Bourdieusian mappings of the changing “space” of lifestyles led us to expect that cultural interest or sophistication would be increasingly linked to “cosmopolitan,” urban, diverse patterns of consumption (Flemmen et al., 2019; Savage et al., 2013). Responses to our open-ended questions indeed show that terms such as “open,” “internationally oriented (weltoffen),” or “tolerant” are most frequently and characteristically mentioned by respondents who report closeness to cosmopolitans in the closed-ended question. Respondents who report closeness to culturally interested people characteristically describe themselves as having wide-ranging (cultural) interests, being open-minded, tolerant, cosmopolitan and outgoing. These self-descriptions suggest that our categories—while open to different interpretations—indeed resonate differently and in expected ways across socio-demographic and electoral groups.

Our main interest in this paper is how social identities (measured using the closeness question) contribute to explaining electoral choice between parties of the new left (Greens, Social Democrats, AL, Solidarités, PdA), and the far right (Swiss People’s Party, EDU, SD). Center parties were aggregated into a distinct group that we do not focus on here (CVP, FDP, GLP, BDP, EVP). We measure party preference by asking which party respondents feel closest to.

We operationalized objective occupational class following the Oesch (2006) 8-class scheme on the basis of ISCO-3d codes. The ISCO-3d codes were derived from answers to three open questions regarding (i) respondents’ occupational tasks, (ii) the characteristics of the firm and (iii) respondents’ hierarchical function. Occupational information is available only for working respondents who gave meaningful answers to these questions (i.e., not their spouses, non-employed, pensioners and students, creating 451 missing values). ⁵ Education is recoded in three categories (below secondary, secondary, and tertiary degree) and we have a measure of objective urban versus rural residence. For the analyses, data is weighted by sex, age and education, and in addition by party preference for the analyses of electoral choice.

Results and Discussion

Social Structure and Social Identities

In this first step of the empirical analysis, we study the structural “rootedness” of social identities. For this purpose, we regress perceived closeness to each of the 17 social groups theorized above on education, class and urban-rural residence, controlling for age and gender. This leaves us with 17 regression tables that we present in full in Appendix 2. ⁶ Tables 1, 2a and 2b serve to summarize these findings. Table 1 provides the regression coefficients in an exemplary way for “Swiss people,” one of the culturally connotated groups. Table 1 thus allows us to showcase how we derive the structural divides underlying these social identities from the regression tables. We summarize findings from the full set of regressions in Table 2a and 2b.

Table 1.

Ordered Logistic Regression Relating Education, Class and Urban-Rural Residence to Feelings of Closeness toward “Swiss People.”

	M1 education	M2 class	M3 rural/urban	M4 full
Education: below sec.	0.140 (0.77)			–0.012 (–0.04)
Education: secondary	r			r
Education: tertiary	–0.725*** (–5.28)			–0.318 (–1.49)
<30	r			r
30-50	0.109 (0.58)	–0.283 (–1.12)	0.102 (0.55)	–0.285 (–1.13)
>50	0.084 (0.48)	–0.424 (–1.54)	0.208 (1.21)	–0.482 (–1.73)
Female	r	r	r	r
Male	–0.176 (–1.41)	–0.186 (–0.96)	–0.239 (–1.95)	–0.184 (–0.94)
Small bus owners		0.447 (1.12)		0.230 (0.55)
Technical profs.		0.722* (2.02)		0.702 (1.95)
Prod workers		1.550*** (4.53)		1.384*** (3.86)
Managers		r		r
Clerks		0.739** (2.78)		0.582* (2.05)
Socio-cult profs.		–0.018 (–0.06)		–0.000 (–0.00)
Service workers		0.775* (2.48)		0.564 (1.63)
Urban			r	r
Rural			0.405** (2.84)	0.250 (1.28)
R2	0.021	0.031	0.007	0.035
BIC	2104.345	1080.287	2126.255	1095.200
N	997	530	997	530

^*p < .05. **p < .01. ***p < .001.

Table 2.

Summary Table of the Main Socio-Structural Correlates of Feeling Close toward Different Identity Groups (for Regression Tables, See Appendix 2).

(a)
	Educational group identities			Class group identities
	“People with tertiary education”	“People with apprentice-ship”	“People without a degree”	“Wealthy people”	“Low income people”	“People who work with other human beings”	“People who manufacture concrete products”
Education group closest	High	Medium	Low	High	Low		Low
Class closest	MNG & SCP	PW	PW	MNG & SBO	PW	SCP	PW
Class most distant	PW & SW	SCP	SCP	PW	TECH	TECH	MNG & SCP
Territorial group closest	Urban	Rural					Rural
Sex closest				Men		Women	Men

Dark gray cells indicate that the socio-structural groups predict closeness to identity groups at p = .001; light gray cells indicate prediction of closeness to identity groups at p = .05−.01. Abbreviations of the class coding: SBO = small business owners; TECH = technical professionals; MNG = managers; SCP = socio-cultural professionals; PW = production workers; SW = service workers.

Table 1 shows that closeness to “Swiss people” discriminates between educational groups, occupational classes, as well as residents in rural and urban areas. In other words, this cultural social identity has clearly identifiable structural foundations, the strongest (most robust) of which seems to be class (robust also in the multivariate model): we see that production workers (just as lower-educated and rurally based individuals) feel closest to “Swiss people,” followed by the other working-class categories (service workers /clerks).

When replicating the interpretation of these regression findings for all 17 social groups, ⁷ we find two main insights: first, the fact that membership in the “objective” structural categories consistently relates to subjective closeness to the corresponding categories validates the analytical value of the (subjective) social groups we intend to measure. This is obvious when it comes to education, territorial residence and sex, but it also holds for both the vertical and the horizontal dimensions of class: Closeness to wealthy people divides managers from production workers and, vice versa, closeness to people with low income divides workers from technical professionals. Regarding work logics, we find that socio-cultural professionals identify most closely with a social group defined by the interpersonal work logic, while technical professionals feel least close to this logic. Similarly, production workers identify significantly more strongly with the social group of people in product manufacturing jobs than managers and socio-cultural professionals do. These divides correspond to the expectations and confirm that these occupational class groups rely on criteria that are meaningful not only in the eyes of scholars, but also of individuals.

The second insight, however, is that social identities of structural “objective” groups go far beyond the immediate corresponding groups. Highly educated people also differ from less educated respondents in their closeness to urban people and—most importantly—cosmopolitans, culturally interested people and people with a migration background. These links do not simply reflect composition effects (as they hold in the multivariate estimation). Similarly, people living in rural areas feel significantly closer to people in manual labor, to people who hold an apprenticeship as highest degree, and to Swiss people. Finally, objective class divides are tellingly reflected in subjective identity divides regarding education, urban-rural residence, cosmopolitanism, lifestyle, nationality and migration background. This network of relationships already indicates that specific aspects of group identity tend to cluster in systematic ways. Here, it is particularly noticeable that “objective” categories significantly predict closeness to the more culturally connoted social groups in every respect we examined and in reinforcing, rather than cross-cutting ways. This provides evidence for the hypothesis that the structural foundations of the universalism-particularism cleavage are mobilized to a large extent through cultural identities.

Figure 2 illustrates the finding that cultural social identities are at least as strongly rooted in objective social conditions as the straightforwardly corresponding groups themselves. The figure shows predicted probabilities of feeling “very close” to social groups by levels of education, class and residence. For each of these structural categories, we compare substantive effect sizes for the directly corresponding social identity and for “Swiss people” as a culturally constructed social identity (remember that all our respondents were “objectively” Swiss). The effects for the culturally connotated identity are as strong if not stronger than for the directly corresponding social identities. We interpret this as evidence that social structure may translate into attitudes and behavior via social identities that depart from the focus on material life conditions. In other words: Knowing how a person’s education level relates to political attitudes is insufficient for understanding individuals’ perceived interests, motivations and choices.

Figure 2.

Predicted probabilities of different “objective” socio-structural groups (x-axes) to feel “very close” to the corresponding socio-structural group and to “Swiss people.”

The Politicization of Social Identities: Identities and Electoral Preference

We have so far established that socio-structural groups differ significantly in terms of their social identities. The next step is to investigate which of these identities are politicized in the sense that they link structural and political divides. Hence, we explore how these identities contribute to predicting vote choice.

We first present descriptive evidence on how party electorates in Switzerland differ in terms of their social identities. In other words, we switch perspectives by looking at social identities through the lens of party electorates. Figure 3 shows how the different party electorates diverge in terms of identification with the same 17 different groups discussed above. It presents mean responses to the closeness questions by partisan preferences in terms of divergence from the sample mean. The point where the y-axis meets the x-axis represents the mean score for the entire sample (partisans and non-partisans).

Figure 3.

Identity divergence between supporters of the far right, new left, and center (-right).

Note. For each identity category, the figure shows how electoral groups diverge from the sample mean (represented by a horizontal line) in terms of mean perceived closeness (1 = not at all close/4 = very close).

Most relevant are the partisan differences between far right and new left that emerge from Figure 3. Looking at the dark grey (far right) and light gray (new left) bars, we see that new left voters reportedly feel much closer than far right voters to cosmopolitans, to people interested in culture, to people with a migration background, and to urban people, while they feel much more distant from Swiss nationals and people living in rural areas. Looking at Figure 3, it is striking how new left and far right voters are most distinctive in terms of their identification with culturally connoted groups.

To test for the significance of these differences between party electorates, we computed bivariate regressions of perceived closeness on partisan preference (Appendix 3): new left and far right voters differ in how close they feel to no less than 14 out of 17 groups, indicating that most of these identities are to some extent politicized. ⁸ In line with the descriptive evidence, differences are strongest regarding culturally connoted groups. This suggests that the politicization of structural group divides has occurred primarily in terms of these cultural groups, more so than through the politicization of the actual occupational class identities and educational identities. For instance, an (objective) interpersonal occupational work logic is one of the strongest predictors of the new left vote in Switzerland, but “working with people” is not the group that divides new left and far right voters most clearly. One structural divide that is explicitly politicized in its own terms is the urban-rural divide, and plausibly this has occurred through cultural frames at least as much as through economic ones.

While some specific identities appear distinctly more politicized than others, it becomes clear in Figure 3 that new left and far right voters differ from each other regarding entire sets of identities. Relatedly, the previous step of the analysis showed that certain sets of identities share similar roots in social structure. Factor analyses (not shown) confirm that identities relate to each other in ways that correspond theoretically to the universalism-particularism divide. In particular, closeness to university-educated, urban, cosmopolitan, and cultured people load clearly onto a factor that seems to capture closeness to a liberal cosmopolitan elite, while national identity, rural identity and closeness to people with an apprenticeship are defining features of a second factor that might be interpreted as capturing attachment to a national “homeland.” Similar patterns emerge from analysis of responses to open-ended questions about identity that were asked in the same survey (analyzed separately, see Zollinger, 2020). Furthermore, additional closed-ended questions in our survey indicate that voters themselves are aware of how some of these categorizations overlap and reinforce each other. ⁹ All of these different analyses suggest that we might think of these specific group identities as jointly forming the boundaries of more overarching antagonistic collective identities that are taking shape in the conflict over universalism versus particularism. Here, we continue our analysis using the more specific identities, in order to avoid losing valuable information: in particular, the more fine-grained break-down may give us an indication of which group categories are (becoming) the most defining features of overarching collective identities.

Linking Socio-Structural Attributes, Identities and Political Preferences

Having presented results first from a bottom-up and then from a politicization perspective, we now go one step further and link structure, identities, and party preferences empirically. We do this by calculating logit regression models using partisan preferences as dependent variables and regressing them on both socio-demographic categories, as well as on the social groups which partisans report feeling closest or least close to. We calculate the models separately for the new left and the far right. In a first model, we include only our main structural variables: education, class, and urban-rural residence (along with age and gender as controls). In a second model, we add relevant in-groups and out-groups for the respective electorates.

In order to determine the most relevant in-groups and out-groups for every electorate (to be included in the models), we use the average closeness ratings each party electorate gives to each group. For every social group, we calculate the distance of partisans’ average closeness from the sample mean (as shown in Figure 3). This allows us to determine which social groups partisans identify with or demarcate themselves from most distinctively. In the models we include the three groups with the largest positive (+), and the largest negative (−) difference between partisan mean and sample mean. Hence, we define as out-groups those to which a certain partisan electorate feels comparatively least close.

Left partisans most positively and distinctively (compared to the entire sample) identify with cosmopolitans, culturally interested people, and urban residents, while they feel distinctively less close than average to rural residents, Swiss people, and the wealthy. For voters of the far right, Swiss citizens, people in rural areas and people who hold a production job most clearly represent in-groups, while cosmopolitans, culturally interested people and people with migration background are primary out-groups. The results of the estimation are presented in Table 3. ¹⁰ The table shows logit coefficients and standard deviations. Overall, the results demonstrate that there is indeed added analytical value of including subjective understandings of group belonging in our analyses of electoral preferences. Model fit increases substantively when social identities are included, both for new left and far right party choice. Coefficients indicate whether socio-demographic criteria and social identities that resonate positively or negatively with a particular electorate indeed predict vote choice in the entire sample.

Table 3.

Logistic Regressions Explaining Party Preference with Structural Categories and Subjective Social Identities–Restricted Sample (Only Partisans).

		Left	Left	Far right	Far right
Class	Below secondary	0.197 (0.42)	0.101 (0.20)	0.096 (0.27)	0.134 (0.35)
	Secondary	r	r	r	r
	Tertiary	0.269 (0.94)	0.062 (0.19)	–0.813*** (−3.32)	–0.814** (−3.13)
	Small bus owners	–1.143 (−1.10)	–2.106 (−1.73)	–0.057 (−0.13)	0.160 (0.32)
	Technical profs.	0.469 (0.76)	0.140 (0.21)	–1.349** (−2.70)	–1.350* (−2.56)
	Prod workers	1.011 (1.93)	0.646 (1.12)	–0.515 (−1.28)	–0.666 (−1.50)
	Managers	0.340 (0.68)	–0.019 (−0.03)	–0.498 (−1.34)	–0.113 (−0.28)
	Office clerks	0.487 (1.06)	0.138 (0.28)	–0.621 (−1.86)	–0.416 (−1.16)
	Socio–cult profs	1.665*** (3.42)	1.414** (2.64)	–1.636*** (−3.53)	–1.329** (−2.71)
	Service workers	r	r	r	r
Residence	Urban	r	r	r	r
	Rural	–0.974** (−3.12)	–0.564 (−1.64)	0.536* (−2.48)	0.393 (−1.63)
Age	<30	r	r	r	r
	30–50	0.056 (−0.17)	–0.062 (−0.17)	–0.075 (−0.27)	0.005 (−0.02)
	>50	–0.107 (−0.29)	–0.228 (−0.56)	–0.452 (−1.50)	–0.298 (−0.90)
Sex	Female	r	r	r	r
	Male	–0.733** (−2.82)	–0.457 (−1.53)	0.475* (−2.14)	0.441 (−1.84)
Subjective group IDs	Cosmopolitans		0.353* (−2.1)		–0.354** (−2.70)
	Cult. interested		0.703*** (−3.79)		–0.465** (−3.22)
	Urban		0.404 (−1.92)
	Rural		–0.454* (−2.19)		–0.039 (−0.24)
	Swiss		–0.399* (−2.00)		0.456** (−2.7)
	Wealthy		–0.752*** (−3.79)
	Production job				0.209 (−1.45)
	Migration background				–0.267 (−1.79)
	Intercept	–0.992 (−1.68)	–0.981 (−0.83)	–0.247 (−0.51)	0.733 (−0.74)
	r2_p	0.118	0.247	0.094	0.168
	N	520	514	520	511

Logit regression models, cells show coefficients and std dev.

^*p < .05. **p < .01. ***p < .001.

Looking first at the models including only objective socio-demographic categories, we observe that highly-educated respondents are significantly less likely to support the far right than people with secondary education (without controlling for class, respondents with higher education are also significantly more likely to vote new left than voters with mid-level education). The same holds for sociocultural professionals compared to the reference category, service workers (as well as production workers). By contrast, the odds of sociocultural professionals supporting new left parties are significantly higher than for service workers. Lastly, people in rural areas are significantly less likely to vote for the new left and more likely to vote for the far right than people in urban areas. All these findings correspond to expectations derived from the literature.

Turning to the models that include social identities, the results in Table 3 support the idea that subjective group perceptions contribute to explaining electoral choices, especially for the new left (less so for center parties, see Appendix 4). Feeling close to cosmopolitan and culturally interested (but not urban) people is positively correlated with a preference for new left parties, while closeness to Swiss, rural and wealthy people is negatively associated with leftism. This finding is interesting, as it indicates that demarcation from the rich continues to complement primarily cultural identities among new left voters who are objectively not poor on average. For far right parties, we find the expected strong and significant positive effect of closeness to Swiss nationals, but no effect of identification with production workers or rural people (holding all else constant). Nor do we find a significant negative effect of closeness to people with migrant background. Hence, feeling close or distant from people with migrant background does not predict vote choice for the far right, which is surprising at first glance. Despite migration being a salient topic, anti-immigrant attitudes don’t seem to reflect people’s distance to migrants themselves, but rather how they think about societal diversity more generally. This resonates with the fact that we indeed find significant negative effects of feeling close to cosmopolitans and culturally interested people on the probability of far right preference. Note that, while such nuance is lost when we combine our measures of identity using factor analysis, the “liberal cosmopolitan” and “national homeland” factors discussed above predict new left and far right voting in the expected ways: when jointly included in a multivariate framework identical to that in Table 3 (not shown), a “liberal cosmopolitan” identity is positively and significantly related to new left voting and negatively related to far right voting. The signs are reversed for identification with “national homeland.” We note that the predictive power of “liberal cosmopolitan” as both a negative and a positive identity is stronger, indicating that far right voting might to a large extent express negative identification with urban liberal elites.

To demonstrate the substantive significance of the results in Table 3, Figures 4 and 5 show predicted probabilities of left and far right party preference on the basis of the models including both social structural determinants and identities. We calculate probabilities for different levels of the most relevant positively and negatively correlated subjective group identities. As a reference for effect sizes, we also calculate predicted probabilities for the most important socio-demographic attribute.

Figure 4.

Predicted probability of left party preference for different levels of the most important socio-demographic attribute and the two most relevant positively and negatively correlated subjective group identities (95% confidence intervals).

Figure 5.

Predicted probability of far right party preference for different levels of the most important socio-demographic attribute and the two most relevant positively and negatively correlated subjective group identities (95% confidence intervals).

We observe very large effect sizes of subjective identities, similar or larger than those of the most important socio-demographic categories. For the far right, the difference between respondents who feel close to Swiss nationals and those who do not is over 35 percentage points. Equally massive is the effect of demarcation from culturally interested people, especially when compared with the difference of around 20 percentage points in the predicted probability of support between respondents with below-secondary education and those with a university degree (keep in mind that education is recognized as a key predictor of far right voting in the literature).

This massive identity-based polarization is mirrored on the left. Those who feel very close to “culturally interested” people vote for left parties with a probability of over 30 percent, while that probability is below 10% for those who do not feel close to this social group at all. Meanwhile, demarcation from wealthy people increases the likelihood of voting for the left by around 15 percentage points. We compare these effect sizes to the most important socio-demographic marker for left voting in our models, occupational class. The probability of small business owners in Switzerland voting for the new left is vanishingly small, while sociocultural professionals are estimated to do so with a probability of over 30% in our model. Taken together, Figures 4 and 5 illustrate how important social identities are in explaining partisan preferences in Switzerland. Importantly, effect sizes are not all that different from those of working-class identification on the vote for the left in the 1970s, when the traditional class cleavage was highly salient. Supplementary analyses reported in Appendix 5 based on the comparative Political Action Survey dataset ¹¹ show that identifying as working class made the probability of voting for the left rise from 25% to 52% in the mid-1970s across a sample of seven countries. This suggests that the social identities we are looking at in our own survey tap into something similarly “real” and sizeable as the identities related to the classical cleavages before electoral realignment took place.

We have now looked at voters’ social structural attributes and social identities side by side. However, the idea underlying our theoretical model and that of classical cleavage theory suggests that social structure translates into voting behavior at least in part through identities. The structure of our data does not allow for an actual mediation analysis, but comparing the different models gives some indication that the effect of socio-structural variables on party preference runs at least partially through identities. In particular, the effect of higher education (as opposed to secondary education) and socio-cultural professionals (as opposed to workers) on the vote weakens once we introduce the relevant identities into the model. Without controlling for class, a strongly significant effect of higher education on the left vote even becomes altogether insignificant once we control for identities. What is more, our full models indicate that subjective identities are predictive of vote choice above and beyond socio-structural group belonging, entirely in line with cleavage theory. ¹² While clearly anchored in social structure (see the first step in our empirical analysis), as social constructs, collective identities may acquire meaning beyond their socio-structural basis, especially when harnessed by political actors to forge electoral coalitions. This partial detachment from social structure is precisely what lends collective identities their stabilizing potential: even if a cleavage’s structural basis is not immutable, its expression in an identity antagonism may persist.

Conclusion

Polarization between parties of the new left and far right is often interpreted as evidence of an emerging or fully mobilized electoral cleavage, given clear socio-structural underpinning of these electoral preferences. We present new evidence on how the electorates situated on the two sides of this cleavage antagonistically relate to each other in terms of their social identities. Together with the substantive nature of these social identities—centering on cosmopolitanism, nationality, and cultural lifestyles—this suggests that the growth of the far right and the resilience of the new left are related to the same universalism-particularism cleavage: The supporters of these two party families are mirror images not only in terms of their socio-structural location and ideological outlooks, but also with regard to their group identifications. Our results thus support theories of electoral realignment. However, we also go a step further by showing that realignment has led to the crystallization of distinctive collective identities. Beyond representing a conflict over new issues, the universalism-particularism divide has the potential to structure how people think about who they are and where they stand in an emerging group conflict that meshes economics and culture. Importantly, we show that even culturally connoted identities are structurally rooted, which suggests that debates over the primacy of “objective” economic and culturalist explanations of electoral preferences are misleading.

We proceeded in three steps. We first looked at how potentially relevant social identities are rooted in socio-structural categories. We find that educational, occupational and place-based groups are meaningful not only in the eyes of scholars but also resonate with individual voters’ self-perceptions. Moreover, we find that culturally connoted identities are equally clearly anchored in socio-demographic groups.

In a second step, we looked at which of these identities are indeed politicized. While we find striking differences in the self-perceptions of new left and far right voters, we also find that not all social identities with roots in social structure relate to electoral preferences. This finding highlights the role of agency and political actors in mobilizing structural potentials. Generally, we find strong evidence for a politicization of divides that are structurally rooted but culturally connoted: identities such as cosmopolitanism, nationality and cultural lifestyles are by far most distinctive between far right and new left voters. The only politicized identity which clearly reflects the corresponding sociodemographic divide is urban-rural residence. Education and occupational class groups come up somewhat more marginally. These observations suggest that even “economically” connoted social identities need to be culturally politicized to unfold their structuring potential.

In a final step, we brought together the societal and politicization-perspectives by including the most distinctive in-groups and out-groups in multivariate models of partisan preference. Including in- and out-groups in models of partisan preference substantially increases model fit, both for predicting new left and far right party choice. Effect sizes of several identities are on a par or even stronger compared to the effects of the most important socio-structural predictors of vote choice.

We want to highlight that, while our analysis of specific identities provides insights on the focal points of collective identity-formation, other, more aggregating ways of looking at our data (e.g., multiple identities’ roots in similar structural categories, factor analysis of group identities, or voters’ perceptions of overlap between groups) suggest that we may be witnessing the emergence of two overarching antagonistic identities (similar to class identities before realignment, which also bundled several group belongings). We might label these overarching identities “liberal cosmopolitan” and “national homeland,” and think of their boundaries as jointly defined by several of the categories we look at here. Our approach may thereby also contribute to our understanding of growing affective polarization that scholars have documented in the advanced democracies of North America and Europe (e.g., Gidron et al., 2020; Mason, 2018; Westwood et al., 2018).

There is, of course, an element of speculation to these thoughts about new, superordinate collective identities coming to structure politics across advanced democracies, in the pervasive way that traditional class identities did for decades. Extending this study beyond Switzerland, refining our measure of “collective identity” to better capture perceived “shared fate,” or integrating survey-based work more systematically with ethnographic approaches are all avenues for future research that might bolster our conclusions from this study. Yet, based on the Swiss case, we suggest that electoral realignment along a universalism-particularism divide is being stabilized by collective identities, which in turn are rooted in social structure. As Switzerland represents a paradigmatic case of a realigned party system, distinctive collective identities may also exist in other countries where realignment progressed early and forcefully, such as France or Denmark. Further, Switzerland could be at least a harbinger for cases of more recent realignment, such as the UK or Germany. In fact, given recent electoral developments in those countries, we might expect that collective identities over universalism versus particularism are taking shape across most advanced Western democracies.

Author Biographies

Simon Bornschier directs the Research Area Political Sociology at the Institute for Political Science at the University of Zurich. His research focuses on the formation and transformation of cleavages and party systems in Western Europe and Latin America. Personal homepage: www.simon-bornschier.eu.

Silja Häusermann is a Professor at the Department of Political Science, University of Zurich. Her current work focuses on electoral and party system change and their impact on distributive politics and policies, as well as on the transformation of welfare state and labor market politics in European democracies. Personal homepage: www.siljahaeusermann.org

Delia Zollinger is a PhD student at the Department of Political Science, University of Zurich. Her current work focuses on long-term transformations of political conflict structures in knowledge economies, specifically on the formation of new collective identities. Personal homepage: deliazollinger.ch

Céline Colombo is a research associate at the Department for Justice and Home Affairs of the canton of Zurich, Switzerland. She is responsible for citizen participation and democratic innovation projects.