Education as an Individual-Level Characteristic

Political divides between people with and without higher education have deepened across advanced democracies. Educational attainment has become a robust predictor of support for radical right versus new left parties. Higher-educated individuals also consistently adopt more progressive positions on social issues such as immigration, gender equality, minority rights, or climate protection (Abou-Chadi and Hix Reference Abou-Chadi and Hix2021; Beramendi et al. Reference Beramendi2015; Bornschier Reference Bornschier2010; Bornschier et al. Reference Bornschier2021; Reference Bornschier2024; Iversen and Soskice Reference Iversen and Soskice2019; Kitschelt Reference Kitschelt1994; Kitschelt and Rehm Reference Kitschelt and Rehm2022; Kriesi et al. Reference Kriesi2008; Marks et al. Reference Marks2023). Thus, education – or the related distinction between a lower-educated working and ‘old’ middle class as opposed to a higher-educated ‘new’ middle class – is considered essential for understanding the emergence of a more socio-cultural dimension of political conflict.

Education as an individual-level determinant of political behavior has drawn attention because it plays a central role in electoral realignment, whereby the traditional working class has become overrepresented in the radical right electorate, and highly educated middle class voters increasingly support new left parties (Gingrich and Häusermann Reference Gingrich and Häusermann2015; Kitschelt and Rehm Reference Kitschelt, Rehm, Beramendi, Silja, Kitschelt and Kriesi2015; Reference Kitschelt and Rehm2022; Oesch and Rennwald Reference Oesch and Rennwald2018). As a result, educational divides have become less closely tied to traditional patterns of class conflict. In the mid-twentieth century, education and income were strongly correlated in their effects on political behavior, with higher education associated with voting for right-wing parties and lower education associated with voting for the left (Gethin et al. Reference Gethin, Martínez-Toledano and Piketty2022, 16). Today, this pattern has reversed in almost all developed democracies, driven by educational groups’ ties to new left and radical right party families (Abou-Chadi and Hix Reference Abou-Chadi and Hix2021).

A well-developed literature offers different mechanisms linking education as an individual-level characteristic to political attitudes, which are likely simultaneously at play. From a political economy perspective, educational attainment has become key for life chances in post-industrial economies (Garritzmann et al. Reference Garritzmann, Garritzmann, Häusermann and Palier2022; Iversen and Soskice Reference Iversen and Soskice2019). The share of the population with a tertiary degree has expanded massively as technological change has increased the demand for cognitive, interpersonal, and creative skills. The higher-educated have long been identified as relative ‘winners’ of modernization, while those with lower-to-medium education are considered among the ‘losers’, their skills devalued. Education hence determines who is equipped to benefit from the opportunities offered by the knowledge economy, as well as who feels well-placed to adapt to this fast-changing environment (Beramendi et al. Reference Beramendi2015; Bornschier Reference Bornschier2010; Gidron and Hall Reference Gidron and Hall2017; Kriesi et al. Reference Kriesi2008; Kurer Reference Kurer2020; McNeil and Simon Reference McNeil and Simon2024). From this perspective, there is a rational, self-interested element to educational groups’ varying propensity to support the radical right, which promises to turn back the clock on social change. A vast body of literature links education not just to immediate material outcomes but to concerns about status and identity as important drivers of political behavior (Attewell Reference Attewell2022; Bornschier et al. Reference Bornschier2024; Gidron and Hall Reference Gidron and Hall2017; Reference Gidron, Adams and Horne2020; Häusermann et al. Reference Häusermann, Kurer and Zollinger2023; Kurer and van Staalduinen Reference Kurer and van Staalduinen2022; Mutz Reference Mutz2018).

Other theoretical perspectives link higher education to values underpinning greater acceptance of progressive societal change. Some argue that education itself socializes students into a set of social and political perspectives, either directly via classroom instruction (Cavaillé and Marshall Reference Cavaillé and Marshall2019; Gelepithis and Giani Reference Gelepithis and Giani2022; Stubager Reference Stubager2009, 329–32) or through interactions with classmates (Mendelberg et al. Reference Mendelberg, McCabe and Thal2017). The idea that educational experiences have a ‘liberalizing effect’ has figured prominently in accounts of electoral realignment (Kitschelt and Rehm Reference Kitschelt and Rehm2022, 5, 38–9; Kriesi et al. Reference Kriesi2008, 7). Relatedly, direct exposure to different social realities through interpersonal professions later in life (as a teacher, psychologist, social worker, etc.) may foster other-oriented values and perceptions of marginalized groups as deserving (Kitschelt Reference Kitschelt1994, 199; Oesch Reference Oesch2006).

From another perspective, education effects on political attitudes about immigration (Lancee and Sarrasin Reference Lancee and Sarrasin2015), European integration (Kuhn et al. Reference Kuhn, Lancee and Sarrasin2021; Kunst et al. Reference Kunst, Kuhn and van de Werfhorst2020), or gender attitudes (Simon Reference Simon2022) are due simply to selection into education. These panel data analyses show individuals do not become more cosmopolitan while in education; they instead self-select into or out of education based on parental background. Those who pursue higher education tend to have educated parents and are socialized into a more cosmopolitan, progressive value structure, and vice versa for particularist lower-educated groups. Taking this into account, recent studies adopt a life course perspective, suggesting that while self-selection based on parental background plays an important role, value divides between educational groups widen during and after education (Ares and van Ditmars Reference Ares and van Ditmars2024; Hooghe et al. Reference Hooghe, Marks and Kamphorst2025).

Increasingly, this research pays attention not just to level but also field of education, both in terms of socialization and self-selection effects. A degree in business (or having parents with one) exposes individuals to a different normative environment to a degree in sociology (Damhuis Reference Damhuis2020; Hooghe et al. Reference Hooghe, Marks and Kamphorst2025; Ladd and Lipset Reference Ladd and Lipset1975; van de Werfhorst and Graaf Reference van de Werfhorst and de Graaf2004; van de Werfhorst and Kraaykamp Reference van de Werfhorst and Kraaykamp2001). Using a Bourdieusian framework, Damhuis (Reference Damhuis2020, 209) distinguishes degrees cultivating general/theoretical cultural capital (for example humanities, science/mathematics/etc., social sciences) from degrees focused on applied cultural capital (for example technical and engineering, economics and business, public order and safety, law). Educational fields associated with general/theoretical cultural capital are more closely linked to voting for new left parties, while educational fields associated with applied cultural capital are linked to voting for parties of the (radical) right. Relatedly, others focus on fields’ so-called Cultural, Economic, Communicative, Technical (CECT) content – whether they reward economic/business or cultural/artistic expertise, and whether they are communicative or technical – showing that people with cultural–communicative education tend to be to the left of and more progressive than those with more economic–technocratic education (Bornschier et al. Reference Bornschier2024; Hooghe et al. Reference Hooghe, Marks and Kamphorst2025; van de Werfhorst and Graaf Reference van de Werfhorst and de Graaf2004).

While this literature tends to treat education as an individual-level characteristic – whether as a labor market asset, a direct source of progressive values, or an indication of predispositions formed early on in life – each of these mechanisms could imply some degree of educationally distinct social networks. Our interest is hence not in disentangling these mechanisms, but in taking this literature as our starting point to study education as a feature of individuals’ social circles.

A relational perspective makes sense if we take claims about an educational divide solidifying into a fully fledged new ‘cleavage’ seriously, since this implies a deeply political–sociological perspective. Although much of the above-discussed literature is framed in precisely these terms (see, for example, Bornschier Reference Bornschier2010; Bornschier et al. Reference Bornschier2021; Bornschier et al. Reference Bornschier2024; Kriesi et al. Reference Kriesi2008; Marks et al. Reference Marks2023; Stubager Reference Stubager2010), the literature on educational divides has yet to fully consider the role networks might play in contemporary cleavage formation. While some of the theoretical arguments above clearly go beyond conceptualizations of individuals as atomistic entities (for example, by considering socialization experiences or parental background), they typically do not make networks themselves a theoretical and empirical object of study. As we argue next, doing so can help address a major puzzle about how a new cleavage achieves its structuring power, compared to the major twentieth-century cleavages of religion or class.

Education as a Feature of Social Networks

Claims about educational divides representing a full-blown new ‘cleavage’ predict an element of stability in political conflict (Bornschier et al. Reference Bornschier2021; Reference Bornschier2024; Hooghe and Marks Reference Hooghe and Marks2018, 126–7; Westheuser Reference Westheuser2021; Westheuser and Zollinger Reference Westheuser and Zollinger2024). Work in the cleavage tradition draws a comparison to the classic class or religious cleavages that reportedly ‘froze’ into place for decades, at least in terms of their party-political representation (Bornschier Reference Bornschier2009). The dominant twentieth-century cleavages only gradually started to lose their structuring power in the face of fundamental socio-economic transformations ushered in by the postindustrial era.

If the educational cleavage today shares key qualities with these older cleavages, we would expect it to be similarly ’sticky’ and slow to fade. In many countries, new left and radical right parties have built consolidated, socially structured electorates (Abou-Chadi and Hix Reference Abou-Chadi and Hix2021; Bornschier and Kriesi Reference Bornschier, Kriesi and Rydgren2013; Dolezal Reference Dolezal2010; Marks et al. Reference Marks2023; Oesch and Rennwald Reference Oesch and Rennwald2018). Universalist–particularist conflict has become a staple of political competition, initially centered on socio-cultural issues but spilling over into distributive conflicts (Attewell Reference Attewell2021; Bornschier et al. Reference Bornschier2021; Häusermann and Kriesi Reference Häusermann, Kriesi and Beramendi2015). It persists across elections and shapes how newly arising issues become politicized. The structuring power of this new cleavage seems comparable to that of past cleavages (Hooghe and Marks Reference Hooghe and Marks2025).

Yet the comparison to older cleavages focalizes a puzzle about the sources of this long-term structuring power in a twenty-first-century context. Historical accounts of cleavage formation and stability emphasize the role of unions and religious organizations in mobilizing social groups and forging group consciousness (Amat et al. Reference Amat2020; Bartolini Reference Bartolini2000; Bellucci and Heath Reference Bellucci and Heath2012; Korpi Reference Korpi1983; Kalyvas Reference Kalyvas1998; Lorwin Reference Lorwin1971; Rokkan Reference Rokkan1977). Such organizations also ‘encapsulated’ voters in a more mundane sense, institutionalizing their embeddedness in social milieus. Full societal pillarization, with ‘competing networks of schools, communications media, interest groups, leisure time associations, and political parties along segmented lines’ (Lorwin Reference Lorwin1971, 142), represents an ideal-typical example. Even short of this extreme level of social closure, the central notion is that individuals’ everyday interactions, shared activities, and conversations used to be structured by formal organizations.Footnote ³ One of the most established observations of the realignment literature is of a decline – even collapse – of these mass organizations that shaped social and political life in the twentieth century.

Today, it is unclear what exactly holds the new cleavage together. It has an identifiable structural basis (see, for example, Abou-Chadi and Hix Reference Abou-Chadi and Hix2021; Marks et al. Reference Marks2023; Oesch and Rennwald Reference Oesch and Rennwald2018) and identity element (Bornschier et al. Reference Bornschier2021; Reference Bornschier2024; Steiner et al. Reference Steiner, Mader and Schoen2024; Stubager Reference Stubager2009; Zollinger Reference Zollinger2024a; Reference Zollinger2024b), but there are no powerful member-organizations articulating the competing political interests of, for example, the vocationally educated or university graduates. The parties mobilizing an emerging conflict lack obvious institutionalized links anchoring them in voters’ everyday lives, particularly given the decline of party membership (van Biezen et al. Reference Van Biezen, Mair and Poguntke2012; Scarrow Reference Scarrow2000). Although some may point to an individualization of modern societies, we can point to clubs or leisure activities associated with typically radical right or new left lifestyles (Flemmen et al. Reference Flemmen, Jarness, Rosenlund and Blasius2019; Florida Reference Florida2012; Savage et al. Reference Savage2013; Westheuser Reference Westheuser and Zollinger2024),Footnote ⁴ or with social movements and activist groups linked to these political camps (see, for example, Hunger and Hutter Reference Hunger and Hutter2021; Kitschelt Reference Kitschelt1994; Kriesi Reference Kriesi and Marks1999, Mierke-Zatwarnicki et al. Reference Mierke-Zatwarnicki, Borbáth and Hutter2025). Some might also point to the role of social media ‘echo chambers’ (Sunstein Reference Sunstein2018; Terren and Borge-Bravo Reference Terren and Borge-Bravo2021). Still, there are no obvious substitutes for the class- or religion-based organizations of the past, if we consider these institutions’ encompassing nature, their penetration into social life, and their formalized links to party politics (Dassonneville Reference Dassonneville2023; Westheuser and Zollinger Reference Westheuser and Zollinger2024).

From the perspective of this puzzle, we think it worthwhile to (re)consider less formalized (physical) social networks as alternative forms of cleavage ‘glue’. Pioneering cleavage scholars laid out two broad forms of social closure: one based on organization-building by parties, and one rooted more ‘naturally’ in social structure.Footnote ⁵ On the one hand, Lipset (Reference Lipset1959, 86–7) highlights partisans’ active endeavors to intensify and formalize intragroup interactions, with Catholics and Socialists attempting ‘to increase intra-religious or intra-class communications by creating a network of social and economic organizations within which their followers could live their entire lives’. On the other, he argues social structure may itself ‘isolate naturally individuals or groups with the same political outlook from contact with those who hold different view’ (1959, 86–7). Political cohesion can thus emerge from social structures fostering intragroup communication while hindering intergroup communication (1959, 248–9).Footnote ^{6 ,} Footnote ⁷

Renewing our focus on the role of social networks can help us to address the puzzle of cleavage formation and stability. Many have observed ‘centripetal’ forces at work in post-industrial knowledge economies fostering highly segregated social networks (Ford and Jennings Reference Ford and Jennings2020; Gingrich and Ansell Reference Gingrich and Ansell2014; Iversen and Soskice Reference Iversen and Soskice2019; Maxwell Reference Maxwell2019; Mijs and Roe Reference Mijs and Roe2021; Mijs Reference Mijs2023). The literature on deepening geographical inequalities in advanced democracies has proliferated in recent years (Hobolt and Tilley Reference Hobolt and Tilley2016; Huijsmans et al. Reference Huijsmans2021; Patana Reference Patana2022; Rodríguez-Pose Reference Rodríguez-Pose2018). The clustering of highly educated knowledge workers in urban innovation hubs appears to be an important factor driving residential segregation and inequality, pricing especially people without tertiary education out of the most attractive cities and neighborhoods (Adler and Ansell Reference Adler and Ansell2020; Moretti Reference Moretti2012). Workplaces, too, have become less integrated across skill-levels, as technological change has reduced complementarities between workers with high and middling skill-levels (Card et al. Reference Card, Heining and Kline2013; Consiglio and van Staalduinen Reference Consiglio and van Staalduinen2024; Iversen and Soskice Reference Iversen and Soskice2019; Mijs and Roe Reference Mijs and Roe2021). Lastly, education itself has become an ever more important part of early adult life, and hence for the formation of social networks – one that tends to cement pre-existing differences in parental background (Bornschier et al. Reference Bornschier2024, chap. 3; Garritzmann et al. Reference Garritzmann, Garritzmann, Häusermann and Palier2022; Hooghe et al. Reference Hooghe, Marks and Kamphorst2025; de Jong and Kamphorst Reference de Jong and Kamphorst2024,Stevens et al. Reference Stevens and Armstrong2008).

Cleavage Stabilization through Educational Networks

From a social closure perspective, we can think of education as having direct consequences for social networks, as perpetuating the effects of family ties and early childhood socialization, and as laying out a path for social network formation in adult life. While some network studies focus on empirically isolating specific causal effects of social influence (see Jackson Reference Jackson2009, 496–8), we theoretically build on sociological work that shows self-selection into educational (and occupational) paths and socialization effects to be mutually reinforcing. Recent evidence of progressively deepening attitudinal divides over the life course is compatible with viewing self-selection into socializing environments (jobs, friendship networks, etc.) at one point in life as affected by socialization earlier in life, reaching as far back as childhood predispositions (Hooghe et al. Reference Hooghe, Marks and Kamphorst2025, Ares and Van Ditmars Reference Ares and van Ditmars2024, see also Lindh et al. 2001, 698–9).

In our account, this implies an amplification of previously discussed individual-level mechanisms linking education to political attitudes (for level and field, see Hooghe et al. Reference Hooghe, Marks and Kamphorst2025, 4–5). In a knowledge-based economy, people will tend to make friends with, be in relationships with, and work with others who share their skill endowments and broad set of life chances. Their social circles will also, on average, be made up of people who developed shared values due to similar socializing experiences on their educational trajectory – because they went to school or university together, because co-workers pursued similar degrees, or because friends of friends might have similar educational (and subsequently occupational) backgrounds. By shaping adult social networks, education may also contribute to stabilizing attitudinal predispositions that form earlier in life, as we continuously interact with others who might have self-selected into a similar educational path.

We can think of homophily with regard to all of these education-related traits as largely driven by structure and institutions (the workings of the knowledge economy, stratifying and segregating effects of educational and labor market institutions, etc.) but intensified by individuals choosing to interact, make friends, work, and start romantic relationships with others who resemble them – sometimes opting out of ties with those who do not (McPherson et al. Reference McPherson, Smith-Lovin and Cook2001, Smith et al. Reference Smith, McPherson and Smith-Lovin2014). While social closure research emphasizes how such patterns may reflect efforts to draw symbolic boundaries and protect status (Bourdieu Reference Bourdieu1984, Reference Bourdieu1987, 5–6, Lamont and Molnár Reference Lamont and Molnár2002, Mijs and Roe Reference Mijs and Roe2021, 3; also see Lindh et al. Reference Lindh and Andersson2024, 3), shared information, ease of communication, or common cultural tastes are additional self-selection mechanisms plausibly at work (McPherson et al. Reference McPherson, Smith-Lovin and Cook2001: 435–6, Smith et al. Reference Smith, McPherson and Smith-Lovin2014, 434–7, Mijs Reference Mijs2023, 3, Christakis and Fowler Reference Christakis and Fowler2009, Jackson Reference Jackson2021).

All this leads us to expect a high degree of segregation in people’s close social networks, not just in terms of educational level but also in terms of field of study.

We view close social ties as a proxy for the ‘milieu’ that has been the focus of cleavage research. While, for example, university-educated individuals are likely to count lower-educated contacts among their weak social ties,Footnote ⁸ we are most interested in the stronger ties that are particularly likely to influence individuals with respect to identities, attitudes, and vote choice. Especially if we consider that these forms of influence in social networks – over attitudes, behaviors, etc. – tend to spread across connections up to three degrees (friends’ friends’ friends’) (Christakis and Fowler Reference Christakis and Fowler2009, 27–30), we are approximating the notion of a milieu.

Existing literature points not only towards overt political discussion, but also a more subtle set of micro-mechanisms through which people imitate each other; sanction or endorse norms of behavior; share information, aspirations, narratives, or emotions; consolidate worldviews; shape one another’s opportunities; cultivate notions of (dis)similarity and group belonging; and help crystallize views on current affairs (Bourdieu Reference Bourdieu1987, 5–6, Lindh et al. Reference Lindh, Andersson and Volker2021, 968, Mijs Reference Mijs2023, 3, McPherson et al. Reference McPherson, Smith-Lovin and Cook2001, 435–6, Hooghe et al. Reference Hooghe, Marks and Kamphorst2025, 4, Jackson Reference Jackson2009, 503–6, Jackson Reference Jackson2021,Christakis and Fowler Reference Christakis and Fowler2009, for example 20–4, ch6, Hall and Lamont Reference Hall and Lamont2013, 7–10) – all against the background of earlier socialization experiences (Jennings and Niemi Reference Jennings and Niemi1968, Kuhn et al. Reference Kuhn, Lancee and Sarrasin2021, Lancee and Sarrasin Reference Lancee and Sarrasin2015).Footnote ⁹ In short, even if early life experiences are responsible for setting people on a path associated with radical right/new left voting, these mechanisms hold the theoretical key to the puzzle of contemporary cleavage stabilization: the question of how a new cleavage might consolidate, durably anchoring individuals and structuring political competition in predictable ways.

If, for an emerging universalism–particularism cleavage, these networks indeed fulfill some of the stabilizing functions that unions or church-related organizations once did, their composition should condition individuals’ anchoring within political divides. Highly educated individuals in homogeneously higher-educated networks should be particularly likely to support new left parties, and vice versa for the lower-educated and radical right parties. The well-known individual-level associations between educational attainment and voting should be reinforced in homogeneous networks, while voters in more heterogeneous networks are exposed to countervailing, less consistent social influence (Mutz Reference Mutz2002, Hobolt et al. Reference Hobolt, Lawall and Tilley2024, see also Dassonneville Reference Dassonneville2023, 2–3, Roccas and Bewer Reference Roccas and Brewer2002, Christakis and Fowler Reference Christakis and Fowler2009, 117). Again, this is plausibly due first to structural conditions and self-selection into more homogenous or mixed networks, but then perpetuated by exposure to those networks. Similarly, the association between educational degrees that are cultural/artistic/communicative in content with the new left on the one hand, and that between technical/economic/business degrees and the radical right on the other should be amplified in the presence of homogeneous networks.

Beyond voting, networks should be related to attitudes and identity formation along the new cleavage. This follows from our discussion of micro-mechanisms, which we envision as more encompassing than individuals simply following their close ties’ votes in any given election. It is also implied by a cleavage perspective, which points towards a deeper permeation of the universalist–particularist logic into voters’ lives, identities, and worldviews. Our social closure argument relies on individuals developing notions of ‘us’ versus ‘them’. This group consciousness, once linked to a particular political camp, is what ‘narrows mobilization markets’ (see Rokkan Reference Rokkan1999; think of a working class identity historically linked to social democracy). Voters with strong social identities and distinct political attitudes become less available for alternative forms of mobilization.

For our puzzle about cleavage stability, networks help explain how core new left and radical right electorates come to form well-articulated group identities such as nationalist, rural, locally rooted as opposed to cosmopolitan, urban, feminist etc. (Stubager Reference Stubager2009; Bornschier et al. Reference Bornschier2021; Reference Bornschier2024; Zollinger Reference Zollinger2024a). Network dynamics (shared norms, imitation, information flows, idea-sharing, etc.) may continually crystallize, say, what it means to lead a cosmopolitan urban life, how this manifests in perceptions of self and of others, and how individuals come to make sense of new political issues in consistently patterned ways – even amid economic, social, and political transformations. Homogeneous networks reinforce predispositions in terms of identity and outlook, supplying the templates for, and entrenching the expression of, those identities and the translation of worldviews into concrete opinions.

The important flipside to this expectation is that political divides will be muted among individuals whose networks do not mirror their own educational characteristics. Even with strong segregating forces in the knowledge society, there is still room for cross-pressures (see Bornschier et al. Reference Bornschier2021; Rokkan Reference Rokkan1999). People move to go to university or for jobs but might retain childhood/youth contacts; they might pursue degrees allowing for a broad variety of occupational choices; or their occupational biographies might unfold in a non-linear manner. This is precisely the variation that interests us in assessing the anchoring role of social closure in a new cleavage.

If education indeed plays a particular role for a newer universalism–particularism divide, we should find that the association of educational networks with political outcomes is stronger when we look at this cleavage than when we consider the traditional state–market divide still associated with established mainstream parties. As discussed above, education has become decoupled from traditional class conflict, even as it has emerged as a key factor underpinning the successes of new left and radical right parties. In particular, the emergence of a new segment of highly educated middle class professionals who care about social justice, support redistribution, and demand social investment (Abou-Chadi and Hix Reference Abou-Chadi and Hix2021; Attewell Reference Attewell2022; Oesch and Rennwald Reference Oesch and Rennwald2018) should blur the association between education and traditional forms of economic-redistributive conflict, including as a feature of networks. This should be particularly pronounced for network composition by educational attainment, since it is at graduate level (more than within fields of study) that higher-educated voters have become heterogeneous in terms of their economic-redistributive stances.

Case Selection, Data, and Methods

We fielded an online survey in Germany, Switzerland, and BritainFootnote ¹⁰ between 2022 and 2023. The motivation for our case selection is twofold. We cover a range of different political-economic and skill regimes – liberal in Britain, coordinated in Germany and Switzerland (Busemeyer Reference Busemeyer2015). This allows us to test whether there is evidence of network-reinforced cleavage formation across knowledge societies with very different political–economic institutions. Notably, vocational training systems with strong employer involvement in Germany and Switzerland are known to have directly stratifying effects, since early tracking reproduces inequalities in parental background. Meanwhile, the liberal education system (and broader liberal market economy) of Britain stratifies and segregates more indirectly, through market logics evident, for example, in the importance of private education spending (Iversen and Stephens Reference Iversen and Huber2008, Busemeyer Reference Busemeyer2015, Thelen Reference Thelen2014).

Second, we include in our sample countries where new left versus radical right opposition varies in terms of strength and consolidation. Switzerland is an example of an entrenched, socially structured, polarized divide between the new left and radical right (Bornschier et al. Reference Bornschier2021, Zollinger Reference Zollinger2024a). The consolidation of a new cleavage at the level of party politics is comparatively less advanced but nevertheless visible in Germany and Britain (Bornschier et al. Reference Bornschier2024).

We recruited 5,100 respondents via the social research company Bilendi/Rispondi. Our sample includes respondents between 18 and 65, since elderly respondents would have been socialized in contexts predating mass educational expansion and the emergence of the educational divide.Footnote ¹¹ We employed crossed gender, education, and age quotas to recruit a sample matching the 18–65-year-old population of each country, based on Eurostat data.

Modeling and Controls

We estimate OLS regression models controlling for self-reported household income insecurity (an ordinal measure ranging from ‘very comfortable to live on current income’ to ‘very difficult to make ends meet’), rural/urban location (rural, suburbs or small town, or urban), gender, age, number of close ties reported (three or four), network size (a count of how many close friends respondents report having in total), father’s education level or CECT-tercile (to partially account for selection into networks), and country fixed effects. The general form of our models is:

$$\eqalign{ Y = \,& {\beta _0} + {\beta _1}\left( {respondent\_education} \right) + {\beta _2}\left( {network\_homogeneity} \right) \cr & \!+ {\beta _3}(respondent\_education*network\_homogeneity) + \beta \left( {Controls} \right) + \varepsilon \cr} $$

Where for different models:

• Y is either vote preference (PTV), identities, or political attitudes.

• Respondent_education is either our binary education level or CECT (field) tercile.

• Network homogeneity is the share of close ties with the same level or field of education.

• β₃ is our key parameter of interest.

In keeping with Clark and Golder’s recommendations for best practices in interpreting interaction models (Clark and Golder Reference Clark and Golder2023, 103), we do not focus on regression coefficients. Instead, we present predicted values of the dependent variable for university degree v. non-degree holders or low v. high CECT respondents, across the range of network homogeneity. We include full regression tables in Appendix B.