Introduction
A growing body of literature documents a substantial urban-rural divide in electoral behavior, as well as in the underlying attitudes and identities of voters in advanced Western democracies (Armstrong et al. Reference Armstrong, Lucas and Taylor2022; Hegewald and Schraff Reference Hegewald and Schraff2025; Huijsmans Reference Huijsmans2023; Huijsmans and Rodden Reference Huijsmans and Rodden2025; Luca et al. Reference Luca, Terrero-Davila, Stein and Lee2022). As this literature points out, the relevance of urban-rural divides has increased over the past decades, including in Western European multiparty systems (Huijsmans et al. Reference Huijsmans, Harteveld, van der Brug and Bram2021; Huijsmans and Rodden Reference Huijsmans and Rodden2025; Kenny and Luca Reference Kenny and Luca2021). These developments have led to a fundamental transformation of electoral geography, with far-right parties gaining electoral ground in rural areas and left-progressive parties becoming the dominant force in urban centers (Haffert Reference Haffert2022; Huijsmans and Rodden Reference Huijsmans and Rodden2025; Rodríguez-Pose Reference Rodríguez-Pose2018). While these phenomena are well established, there is still disagreement about what drives this ongoing transformation at the level of the electorate. A related question is the role of age and cohorts, since younger voters are arguably more easily affected by their geographic context. More generally, generational replacement has repeatedly been identified as a slow but fundamental driver of social change and a contributor to the ongoing electoral realignment in European party systems (Lichtin et al. Reference Lichtin, van der Brug and Rekker2023; Mitteregger Reference Mitteregger2024; van der Brug Reference van der Brug2010; van der Brug and Rekker Reference van der Brug and Rekker2021; Walczak et al. Reference Walczak, van der Brug and de Vries2012).
Against this background, we ask if cohort effects have fueled the urban-rural divide: are more recently socialized citizens more strongly divided by their place of residence in terms of political attitudes and voting behavior? Haffert and Mitteregger (Reference Haffert and Mitteregger2023) argue that the urban-rural divide mainly divides younger generations, with urban youth holding the most progressive views in society, whereas place-based differences among earlier cohorts are much smaller. However, since they are not using overtime data, they cannot disentangle different components of age and time. Thus, it remains unclear whether the growing urban-rural divide is driven by new cohorts entering the electorate, by life-cycle dynamics, or by a broader realignment affecting all voters regardless of age or socialization context. However, this is an important question if we think about the future development of this conflict. If the urban-rural age gradient is purely due to an age effect, place-based differences will decline as people get older. If it is (also) a cohort effect, by contrast, the conflict will become ever more important, as earlier generations with weak urban-rural differences are replaced by younger, more spatially polarized ones.
To explore this, we focus on the interaction of urban-rural residence and political generations in Switzerland. Switzerland represents a European context in which place has increasingly become relevant for understanding political attitudes and where parties strongly emphasize this cleavage (Audikana and Kaufmann Reference Audikana and Kaufmann2022; Bornschier et al. Reference Bornschier, Häusermann, Zollinger and Colombo2021; Mantegazzi Reference Mantegazzi2021; Maxwell Reference Maxwell2020; Strebel and Müller Reference Strebel and Müller2023; Zollinger Reference Zollinger2024). Swiss voters treat urban-rural characteristics as an important factor for their political identity (Bornschier et al. Reference Bornschier, Häusermann, Zollinger and Colombo2021; Zollinger Reference Zollinger2024; Zumbrunn 2025). Moreover, the far-right Swiss People’s Party (SVP/UDC) – the electorally strongest party in Switzerland – has steadily transformed from a formerly agrarian party into a far-right party (Kriesi et al. Reference Kriesi, Lachat, Selb, Bornschier and Helbling2005). Conversely, left-wing parties (GPS and SP) have since the 1970s increasingly been associated with urban areas and their demands (Strebel and Müller Reference Strebel and Müller2023).
At the same time, most rural regions in Switzerland can hardly be characterized as ‘left behind’ or as being in a ‘development trap’ (Rodríguez-Pose et al. Reference Rodríguez-Pose, Dijkstra and Poelman2024), since the economy is growing in most places and since public infrastructure is of high quality even in rural areas. In many rural places, the population is growing, not declining (Rodríguez-Pose et al. Reference Rodríguez-Pose, Lee and Lipp2021). Politically, the federalist features of the country give rural places disproportionate institutional power. Nevertheless, major differences between urban and rural areas exist, particularly regarding how wealth is generated. Whereas urban centers have become hubs of the tertiary knowledge economy (Zollinger Reference Zollinger2024), many rural areas are still strongholds of an export-focused manufacturing sector. Consequently, rural grievances in Switzerland revolve more around cultural than around economic issues.
To examine whether urban-rural differences in attitudes and voting behavior are stronger for newer cohorts, we combine post-election survey data spanning 28 years with municipality-level population density. This allows us to test whether there are persistent generational differences in the urban-rural divide. We test this in terms of political attitudes on the cultural axis (immigration and environment) and voting behavior for parties representing the poles of this political dimension (progressive-left and far-right).
The empirical analysis shows that urban-rural differences along the cultural dimension are indeed structured by cohort effects: first, cohorts that came of age since the mid-1980s, decades in which the Swiss political landscape underwent a fundamental transformation, exhibit more pronounced urban-rural divides, particularly in attitudes toward immigration, and to a lesser extent, environmental issues. Newer urban cohorts also diverge more clearly from their predecessor generations in their voting behavior: cohort effects on voting for left-wing parties are stronger in more urban municipalities. Conversely, a mirror image emerges for the far-right, with more recently socialized urbanites being less likely to vote for the far-right than their earlier socialized ‘neighbors’.
Altogether, this paper makes three main contributions. Firstly, we replicate the finding that the urban-rural divide is strongly structured by age, so far established for Germany (Haffert and Mitteregger Reference Haffert and Mitteregger2023) and Sweden (Valldor Reference Valldor2025), in a new context, Switzerland. Secondly, we go beyond this finding by analyzing a place-cohort interaction in an APC framework. This allows us to disentangle age from cohort effects and to establish the relevance of cohort effects. In that regard, the study adds to existing research that tries to look at differences both between and within cohorts, for example, when looking at education or gender as a divisive factor within a cohort (Schäfer and Steiner Reference Schäfer and Steiner2025; Shorrocks Reference Shorrocks2018; Steiner Reference Steiner2023). Thirdly, and directly related, we conclude that the political urban-rural divide might become ever more important. Since the APC analysis demonstrates the importance of cohort effects, and since there is ample evidence that people’s understanding of politics remains structured by the cleavages that they internalized during their process of politicization (Steiner Reference Steiner2024), the urban-rural divide is unlikely to fade from people’s minds when they get older. Consequently, cohorts for which this divide did not play a major role will increasingly be replaced by cohorts interpreting politics through the lens of this conflict. This ultimately increases the incentives for parties and other political actors to actively engage in this divide on the supply side of politics.
Argument
Increasing urban-rural divides in Western Europe
Political differences between urban and rural citizens have become more pronounced in recent years. While researchers initially focused mainly on the US (Cramer Reference Cramer2016; Lyons and Utych Reference Lyons and Utych2023; Munis Reference Munis2022; Scala and Johnson Reference Scala and Johnson2017), a growing literature shows similar divides in Western Europe. Divides between urban and rural voters have been analyzed regarding economic (Pinggera Reference Pinggera2023) and cultural (Huijsmans et al. Reference Huijsmans, Harteveld, van der Brug and Bram2021; Luca et al. Reference Luca, Terrero-Davila, Stein and Lee2022) attitudes, political trust (Lipps and Schraff Reference Lipps and Schraff2021; Mitsch et al. Reference Mitsch, Lee and Morrow2021; Zumbrunn Reference Zumbrunn2024), social identities (Zollinger Reference Zollinger2024), affective polarization (Hegewald and Schraff Reference Hegewald and Schraff2025), external efficacy (Del Horno et al. Reference del Horno, Rico and Hernández2023), group appeals (Haffert et al. Reference Haffert, Palmtag and Schraff2024), and party support (Haffert Reference Haffert2022; Huijsmans and Rodden Reference Huijsmans and Rodden2025; Rodríguez-Pose Reference Rodríguez-Pose2018; Rodríguez-Pose et al. Reference Rodríguez-Pose, Terrero-Dávila and Lee2023; Taylor et al. Reference Taylor, Lucas, Armstrong and Bakker2023). These increasing divides likely affect partisan preferences in multiparty contexts, fueling ongoing party system realignment.
While many of these analyses rely on cross-sectional data, there is agreement that these divides have been increasing in the past decades (Huijsmans and Rodden Reference Huijsmans and Rodden2025). For the Netherlands, Huijsmans et al. (Reference Huijsmans, Harteveld, van der Brug and Bram2021) find a growing divergence in cosmopolitan and nationalist attitudes between residents of more and less urbanized municipalities since the 1980s. This is largely driven by urban places becoming more progressive over time. Moreover, urban places are also steadily transforming into strongholds of left-wing, especially green, parties (Haffert Reference Haffert2022; Huijsmans and Rodden Reference Huijsmans and Rodden2025). In contrast, growing economic inequalities coupled with increased cultural resentment have driven support for (populist) far-right parties in rural places (Rodríguez-Pose et al. Reference Rodríguez-Pose, Terrero-Dávila and Lee2023). Furthermore, feelings of place resentment and lack of political representation have been shown to drive anti-immigrant sentiment among rural voters and increasingly lead them to vote far-right (Huijsmans Reference Huijsmans2023; Magalhães and Cancela Reference Magalhães and Cancela2025).
In a related study, Haffert and Mitteregger (Reference Haffert and Mitteregger2023) argue that the processes that are typically analyzed as driving the emergence of urban-rural divides – economic changes such as the transformation to the knowledge economy and cultural changes such as the rise of cosmopolitanism – have affected newer cohorts more strongly than earlier cohorts. Hence, they argue that the urban-rural divide should be stronger for younger individuals. In support of this argument, they demonstrate that younger Germans differ more along the urban-rural divide than their older compatriots. However, they cannot disentangle age effects (in which urban-rural differences fade as people age) from cohort effects (in which these differences persist when people age). Hence, while they speculate that generational replacement will further increase the societal importance of the urban-rural divide, their data does not allow them to test this. In the following, we argue that the effect they observe is indeed likely to be a cohort effect that will not fade when voters age.
Cohorts as drivers of social change
At the center of our argument is the so-called ‘impressionable years theory‘ (de Graaf and Evans Reference de Graaf and Evans1996; Inglehart Reference Inglehart2008; Neundorf and Smets Reference Neundorf and Smets2017). According to this theory, political preferences are formed due to socialization experiences in early adulthood (when citizens are 15–25 years old) and remain stable as individuals age. The early formation of habits and attitudes is then reflected in long-lasting cohort differences. This theory contrasts with approaches that propose either a lifelong openness or an increasing/decreasing persistence of attitudes over an individual’s life course (Alwin and McCammon Reference Alwin, McCammon, Mortimer and Shanahan2003; Peterson et al. Reference Peterson, Smith and Hibbing2020; Sears and Funk Reference Sears and Funk1999).
Whereas typical applications of the ‘impressionable years theory’ focus on differences between cohorts, the theory does not propose that a generation uniformly adapts to changing political circumstances and emerging conflicts. In fact, birth cohorts consist of different groups (so-called ‘generational units’, see Mannheim Reference Mannheim1928) that are differently exposed to events, traumas, historical developments, and political debates. Individuals can thus, even if they belong to the same birth cohort, have divergent attitudes depending on their sociodemographic or political background, because not all members of a cohort may be similarly exposed to certain socialization effects (Mannheim Reference Mannheim1928; Ryder Reference Ryder1965; Shorrocks Reference Shorrocks2018; Steiner Reference Steiner2024).
Other authors have shown that different gender and educational groups belonging to the same birth cohort not only differ in their behavior from previous cohorts but also within their cohort (Schäfer and Steiner Reference Schäfer and Steiner2025, Shorrocks and Grasso Reference Shorrocks and Grasso2020; Steiner Reference Steiner2023). One might expect such generational units to have formed in recent decades as well with regard to the role of place: across Western Europe, urban dwellers have been exposed to a booming economy in city centers coupled with a growing ‘creative industry’. By contrast, cohorts living in rural areas have been coming of age in places that have been characterized either by the persistence or by the decline of a more manufacturing-based economy (Rodríguez-Pose Reference Rodríguez-Pose2018; Rodríguez-Pose et al. Reference Rodríguez-Pose, Terrero-Dávila and Lee2023). This suggests that it is not a person’s age but rather socialization experiences that drive attitudinal differences between urban and rural places over time and could increasingly structure political attitudes and behavior.
Attitudinal differences between cohorts along the urban-rural divide
A major debate in the literature on urban-rural divides is whether these divides are mainly a compositional phenomenon, driven by self-selection into urban or rural contexts, and to what extent these contexts themselves influence their residents (Brown and Mettler Reference Brown and Mettler2024; Fitzgerald Reference Fitzgerald2018; Gallego et al. Reference Gallego, Buscha, Sturgis and Oberski2016; Maxwell Reference Maxwell2019). Here, we follow the ‘impressionable years’ theory and argue that differences in place-based context are most important, and context effects are strongest, during the period of socialization. Indeed, the literature indicates that the place of residence during socialization has long-lasting consequences: focusing on the role of local decline in the UK, McNeil et al. (Reference McNeil, Lee and Luca2023) demonstrate that socialization experiences in such an environment have a long-lasting effect on individual attitudes. Similarly, Palmtag (Reference Palmtag2023) demonstrates that local inequality in youth has a long-term effect on attitudes toward redistribution.
We argue that urban and rural context effects during their socialization period were relatively similar for members of earlier cohorts but differed more strongly for newer cohorts. Crucial for this argument is the urban transformation to the knowledge economy, which caused an increasing divergence of the type of jobs that are dominant in urban and rural areas, as well as in the lifestyles and cultural attitudes associated with these jobs. During the postwar period, the economic experience of urban and rural places became more similar, as the importance of agriculture declined and rural areas industrialized, as road-based transport of individuals and goods made the location of production sites much more flexible (Rosés and Wolf Reference Rosés and Wolf2021). Hence, in 1970, most urban and rural places in Western Europe were part of an industrial economy.
With the transition of industrial societies to a knowledge-based economy, however, urban places started to transform their economic models. In some countries, particularly in liberal market economies, they pulled ahead economically, leaving rural areas behind (Moretti Reference Moretti2012; Rodríguez-Pose Reference Rodríguez-Pose2018; Rodríguez-Pose et al. Reference Rodríguez-Pose, Terrero-Dávila and Lee2023). In other countries, mostly in Coordinated Market Economies (CMEs) that kept a stronger manufacturing base, wealth differences remained smaller, but the structure of the economy diverged: while cities expanded the knowledge-intensive service sector, rural areas kept a much stronger focus on industrial sectors.
Switzerland belongs to the CME group here. Over the last 50 years, the country has experienced a strong transformation to the knowledge economy. At the same time, rural regions remained economically successful and embedded in strong infrastructure networks. The transformation of the Swiss economy was thus more characterized by a sectoral divergence between cities and the countryside than by growing inequalities between these places. In 1964, i.e., exactly when baby boomers started to enter the labor market, the share of the Swiss labor force employed in the secondary sector peaked at 48.8%, whereas 39.6% of the labor force was employed in the tertiary sector. Over the next decades, the Swiss economy deindustrialized and tertiarized in a process that started slowly and then massively accelerated after the oil crisis. In 1972, employment in the tertiary sector for the first time surpassed employment in the secondary sector, but 45.6% of the employed were still working in the secondary sector. By the year 2000, this number had almost halved, to 24%, whereas 71.9% were now employed in the tertiary sector. Thus, the generation that entered the labor market in the 1990s entered a completely different labor market than their parents. Since the year 2000, this trend has continued, albeit at a much slower pace: in 2024, the secondary sector employed 19.9% of the Swiss labor force.
This transformation to a tertiary knowledge economy was faster in urban centers than in mid-sized towns. In the former industrial stronghold of Winterthur (population 1970: 92,722), for example, the share of people working in the secondary sector declined from 55.1% to 16.4% between 1970 and 2020. In Altdorf (population 1970: 8,647), by contrast, it declined from 58% in 1970 to still 27.8% in 2020. While both places had a similar share of industrial jobs in 1970, the share was almost twice as high in Altdorf in 2020. Figure 1 demonstrates the systematic pattern behind this example: while Switzerland’s biggest cities were never dominated by manufacturing in the 20th century, they had an above-average secondary sector share in 1970. By 2020, they were as deindustrialized as the most remote mountain villages.
Figure 1. Swiss municipality size and secondary sector employment, 1970 and 2020.
Thus, while rural areas in Switzerland are not ‘left behind’, the structure of their economies and the experiences of younger cohorts increasingly diverge. This can be illustrated by a comparison of the city of Zurich with the rest of the canton of Zurich, the biggest canton in Switzerland. There are almost no differences in the educational profile of those aged 70–79: 21% of city-dwellers in this age group had a tertiary education in 2016, while the respective number was 20% for those in the rest of the canton. Among the 30–39-year-olds, by contrast, there is a strong urban-rural gradient: in the city, 65% of this age group had a tertiary education, compared to only 44% in the rest of the canton (BFS 2023a). We thus expect that the transformation to the knowledge economy and creative industry in urban areas has had a much stronger impact on newer cohorts.
At the same time, the fact that rural areas can hardly be characterized as ‘left behind’ in economic terms suggests that political conflict may be less about economic than about cultural issues. Indeed, while progressive urban voters may put a strong emphasis on politics of ‘social investment’ (Pinggera Reference Pinggera2023), they still share support for the traditional welfare state with many rural voters. However, the fact that recent urban and rural cohorts have very different educational experiences and work in very different occupations suggests that they may hold increasingly different views on cultural issues. Indeed, both educational and occupational sites have been analyzed as places of preference formation (de Jong and Kamphorst Reference de Jong and Kamphorst2024, Kitschelt and Rehm Reference Kitschelt and Rehm2023). These economic changes thus are the sociostructural basis for cultural changes, since urban centers offer abundant employment opportunities for tertiary-educated individuals, especially those in sociocultural professions, and the so-called ‘creative class’ (Florida Reference Florida, Mellander, Adler and Stolarick2005). These groups exhibit distinctly liberal cultural attitudes compared to citizens belonging to other classes (Lindskog and Oskarson Reference Lindskog and Oskarson2023; Oesch and Rennwald Reference Oesch and Rennwald2018).
Hence, while newer cohorts are generally expected to hold more liberal and progressive attitudes on cultural issues, such as Europeanization, gender equality, or same-sex marriage, (Caughey et al. Reference Caughey, O’Grady and Warshaw2019; Lauterbach and de Vries Reference Lauterbach and de Vries2020; O’Grady Reference O’Grady2022; van der Brug and Rekker Reference van der Brug and Rekker2021) and may attribute a greater importance to these issues (Jocker et al. Reference Jocker, van der Brug and Rekker2024b; Rekker and van der Brug Reference Rekker and van der Brug2023), we argue that this trend towards social liberalism has likely not unfolded at the same speed everywhere, with urban places again being at the forefront.
As a result, citizens living in urban areas have adopted increasingly liberal stances on cultural issues (Huijsmans et al. Reference Huijsmans, Harteveld, van der Brug and Bram2021). By contrast, cultural experiences and attitudes are less geographically stratified for earlier cohorts. This is most true for cohorts who came of age before the mid-1960s, but even in the early 1970s, Western European cities in general, and Swiss cities in particular, had not yet become hubs of cosmopolitanism (Huijsmans et al. Reference Huijsmans, Harteveld, van der Brug and Bram2021).
Thus, the socialization experiences of newer cohorts are more geographically stratified than those of earlier cohorts. This should particularly be the case for cultural issues, like immigration, the environment, and gender equality. We, therefore, argue that it is not a person’s age, but rather socialization experiences, or a combination of the two, that drives attitudinal differences between urban and rural places over time. Accordingly, we hypothesize:
H1: Urbanity is more strongly associated with progressive attitudes on cultural issues among newer than among earlier cohorts.
Electoral differences between cohorts along the urban-rural divide
Following this increased cultural attitudinal urban-rural gap, we expect these differences to be reflected in electoral behavior as well. This should also be the result of the growing salience of party competition along the urban-rural divide. In the 1970s and 1980s, parties competed much more about issues of vertical economic distribution or – increasingly – about value change as such. In recent decades, by contrast, parties have started to explicitly frame their politics around matters of place. When political parties begin to politicize cultural issues coupled with the urban-rural divide, they might employ group appeals around it (Haffert et al. Reference Haffert, Palmtag and Schraff2024; Zollinger Reference Zollinger2024; Zumbrunn Reference Zumbrunn2025). In turn, this might have a greater impact on cohorts who grew up in such a context of higher salience.
For newer cohorts, where they live should thus more strongly shape voting behavior. Yet, there is little research that empirically tests this: using a categorical urban-rural measure, Walczak et al. (Reference Walczak, van der Brug and de Vries2012) find no evidence that voting in European elections is more strongly associated with the urban-rural divide among younger generations. However, more recent findings call these results into question. Focusing on the German context and relying on a fine-grained continuous measure of urbanity, Haffert and Mitteregger (Reference Haffert and Mitteregger2023) show that the urban-rural divide is more pronounced among younger groups of voters: young voters in urban areas are more likely to vote for the Greens than both older urbanites and people in rural areas, while the far-right is significantly more popular among younger voters in rural areas. However, both studies look at only one specific point in time, preventing any conclusions about longitudinal (and thus cohort) developments.
Turning to the context of Switzerland, the country’s party system can be characterized as ‘tripolar’ with three equally sized political blocs, resembling other European party systems (Hutter and Kriesi Reference Hutter and Kriesi2020, Oesch and Rennwald Reference Oesch and Rennwald2018). Regarding the urban-rural and cultural divide in the Swiss party system, two of these poles appear to be relevant. On the one hand, the parties most strongly representing a cultural-liberal stance and urban spaces are the two left-wing parties, the Social Democratic Party (SP) and the green-left Green Party of Switzerland (GPS), as well as minor left-alternative parties (Häusermann et al. Reference Häusermann, Abou-Chadi, Bürgisser, Enggist, Mitteregger, Mosimann and Zollinger2022; Petitpas and Sciarini Reference Petitpas and Sciarini2022). Since the mid-1990s, these parties have gotten a stable 30% of the vote. Looking at their ideological positions, the SP and the GPS are very closely positioned (Rovny et al. Reference Rovny, Bakker, Hooghe, Jolly, Marks, Polk, Steenbergen and Vachudova2024). They also compete for very similar sociodemographic electorates, with voters of both parties belonging to the highly educated new middle class (Häusermann et al. Reference Häusermann, Abou-Chadi, Bürgisser, Enggist, Mitteregger, Mosimann and Zollinger2022). On the other hand, the far-right, consisting of the SVP and smaller far-right parties (SD, EDU, MCG, LEGA), represents the opposite pole, with anti-immigrant and eurosceptic positions coupled with economic right-wing positions and a rural attachment (Bornschier Reference Bornschier2015; Lauener Reference Lauener2022). Generally, the rise of the far-right in Switzerland has generated much interest (e.g., Bernhard and Hänggli Reference Bernhard and Hänggli2018, Mazzoleni and Ivaldi Reference Mazzoleni and Ivaldi2022). Similar to the progressive-left pole, the cumulative vote share of this party family has reached slightly more than 30% since the mid-1990s.
Moreover, in this general context of a polarization between the new left and the far-right, the urban-rural divide has increasingly become the subject of direct democracy and electoral campaigns, with, for example, campaigns led by the Farmers’ Union trying to mobilize their rural electoral base, the far-right pejoratively depicting urbanites as ‘lazy’, or tenants’ associations trying to rally urban voters around proposals demanding lower rents in cities (Linder and Mueller Reference Linder and Mueller2021, Zumbrunn Reference Zumbrunn2025). Indeed, urban-rural divides in direct democratic voting have consistently increased since the 1990s (Mantegazzi Reference Mantegazzi2021; Zollinger Reference Zollinger2024).
For voters coming of age in the last three decades, the urban-rural divide thus played an important (and still increasing) role in their political socialization, with clear party options representing urban and rural poles, a strong electoral realignment, diverging educational opportunities, and a persistently high salience of this divide, also driven by direct democracy. For cohorts coming of age before the 1980s, by contrast, this divide played a smaller role.
We, therefore, formulate the following hypotheses:
H2a: Urbanity is more strongly associated with left-wing voting among newer than among older cohorts.
H2b: Rurality is more strongly associated with far-right voting among newer than among older cohorts.
Data and method
Data and operationalization
Next to theoretical considerations, Switzerland is ideally suited to empirically analyze the outlined questions due to its rich (macro and micro) data availability on the municipality level. To analyze the individual level, we rely on data from the Swiss Post-Election Survey, SELECTS. SELECTS is a high-quality, weighted, representative survey conducted after each national election from 1995 until the most recent elections in 2023 (Tresch et al. Reference Tresch, Lauener, Bernhard, Lutz and Scaperrotta2020). SELECTS contains rich information on voting behavior, sociodemographics, attitudes, and the place of residence (ZIP code) for around 5,000 respondents in each election year. Several questions on attitudes and voting behavior have been consistently asked, which makes an APC analysis feasible. The sampling method of SELECTS also allows for an analysis of regional differences (Tresch et al. Reference Tresch, Lauener, Bernhard, Lutz and Scaperrotta2020). Still, this dataset (like all cross-sectional survey data) has some limitations. It does not allow us to observe all generations at all periods of their life course since newer cohorts only enter the dataset in the more recent survey rounds, whereas the earliest cohorts only enter when they are 50 or older. This should be kept in mind when interpreting the results.
Measurement of urbanity
The literature has presented various ways of measuring urbanity with different strengths and weaknesses (Nemerever and Rogers, Reference Nemerever and Rogers2021). For example, some measurements use discrete urban and rural categories, often stemming from classifications provided by statistical agencies (Huijsmans et al. Reference Huijsmans, Harteveld, van der Brug and Bram2021). Other researchers rely on respondents’ subjective categorizations, also mostly resulting in categorizations (Nemerever and Rogers Reference Nemerever and Rogers2021). A second approach treats urbanity as a continuum, e.g., by using the number of inhabitants in a region of interest (Zumbrunn and Freitag Reference Zumbrunn and Freitag2023) or by relying on population density (Nemerever and Rogers Reference Nemerever and Rogers2021).
In this paper, we measure urbanity using population density on the municipality level.Footnote 1 Swiss municipalities are rather small, especially in a comparative context, which makes them suitable units of analysis to measure urbanity and to use population density as a measurement for their degree of urbanity (Strebel Reference Strebel2025). This means that each municipality has a relatively homogeneous population structure. For example, population density in the 12 districts of Zurich ranges from 2,574 to 10,008 inhabitants per km2 because Zurich has not incorporated any suburbs since 1934. In the slightly bigger German city of Düsseldorf, by contrast, population density in the 10 districts ranges from 636 to 7,828 inhabitants per km2. Population density also has the advantage that it is consistently available over a long time for all municipalities.
In the following, we present results with a continuous and with a categorical measure. In regression tables, we display results using a continuous measure of population density in a logarithmic form (log10) for both theoretical and methodological reasons. Theoretically, it is not very plausible that a difference of 100 inhabitants per km2 has the same effects at low and high levels of population density. Methodologically, this allows us to account for the skewed distribution of the variable (only a few municipalities have very high levels of population density; see Figure A1 and Figure A2 in online Appendix A). To facilitate interpretation, we visualize the results of analyses with a categorical variable for which we break the population density variable into quintiles. Ultimately, both approaches lead to very similar results.
We use the Swiss Municipal Data Merger Tool (SMDMT) to assign respondents to a municipality that has merged with another municipality over the observed period (see Knechtl and Stutzer (Reference Knechtl and Stutzer2021). Even though large-scale territorial reforms have been rare in Switzerland, small mergers have been increasingly happening since the year 2000 (Strebel Reference Strebel2025). We also re-run the analysis by relying on other spatial indicators from the Federal Office of Statistics, which groups all Swiss municipalities into different groups. Here, we rely on two different categorizations, one grouping municipalities according to their size (resulting in four categories) and another one grouping them according to their economic, administrative, and cultural centrality, further disaggregating rural, urban, and suburban municipalities (resulting in five categories).Footnote 2
Dependent variables
To analyze differences in attitudes, we focus on two questions in the SELECTS survey that have been asked consistently over the observation period and can be seen as related to the cultural dimensionFootnote 3 : first, for attitudes toward immigration, the question of whether Swiss citizens should have more opportunities than foreigners.Footnote 4 Second, we look at the question of whether the environment should be prioritized over the economy or vice versa.Footnote 5 ,Footnote 6 The two questions were measured on a scale from ‘strongly disagree’ (1) to ‘strongly agree’ (5). To facilitate comparison, we dichotomize both variables by contrasting those who strongly or somewhat agree (1) with all others (0). An analysis with these variables in their original form can be found in online Appendix C (see Table C9). We also include non-voters in the analysis of attitudinal differences.
For voting behavior, we use vote recall in the national elections. To test H2a, we group Social Democrats, Greens, and other small green-left-alternative parties (AL, EàG, Solidarités, PdA) into one left-wing bloc (vs. all other parties). The two main parties of the left (SP and GPS) are both ideologically very similarly positioned and appeal to similar sociodemographic and ideological groups of voters (see Bochsler and Sciarini (Reference Bochsler and Sciarini2010); Häusermann et al. (Reference Häusermann, Abou-Chadi, Bürgisser, Enggist, Mitteregger, Mosimann and Zollinger2022); Petitpas and Sciarini (Reference Petitpas and Sciarini2022). Still, we also run regressions for the two main left-wing parties separately (see Table B3). For H2b, we use the far-right vote (SVP and small far-right parties (EDU, LEGA, MCG, NA, SD) vs. all other Swiss parties) recall. In line with other research on electoral cohort differences (Lichtin et al. Reference Lichtin, van der Brug and Rekker2023; Mitteregger Reference Mitteregger2024; Steiner Reference Steiner2023b), we exclude non-voters from the analysis of voting behavior.Footnote 7
Methodological approach
The APC-identification problem and political generations in Switzerland
Every study trying to disentangle age, period, and cohort (APC) effects is confronted with the fundamental issue of these three components’ perfect multicollinearity (Bell Reference Bell2020; Bell and Jones Reference Bell, Jones, Burton-Jeangros, Cullati, Sacker and Blane2015; Glenn Reference Glenn2005; Tilley and Evans Reference Tilley and Evans2014). An important precondition to deal with this problem is the usage of panel or cross-sectional data spanning a time as long as possible. While panel data are better suited to detect life-cycle effects, cohort effects are harder to detect with this type of data (Bell Reference Bell2020; Bell and Jones Reference Bell, Jones, Burton-Jeangros, Cullati, Sacker and Blane2015). Because our main interest is to examine cohort differences, we prefer over-time cross-sectional data. However, this does not allow us to track individuals within a specific context when they age and entails a limitation for our analysis.
While many solutions to overcome the APC identification problem have been suggested, no method has become a universally accepted benchmark (Bell Reference Bell2020; Glenn Reference Glenn2005; Neundorf and Niemi Reference Neundorf and Niemi2014). In this paper, we adopt an approach grounded in a theoretical specification of the relevant APC components that has become increasingly applied in social science APC research (Jocker et al. Reference Jocker, van der Brug and Rekker2024a, Mitteregger Reference Mitteregger2024, van der Brug and Rekker Reference van der Brug and Rekker2021). This ‘solution’ proposes to model at least two of the three components as theoretically grounded and non-linear effects. This approach removes the assumption of a perfect linear relationship between age, period, and cohort and improves the interpretability of the estimates. While this cannot entirely overcome the APC-identification problem, it is preferable to ‘mechanical’ approaches, such as hAPC models, that have been criticized for being a ‘methodological black box’Footnote 8 (Bell Reference Bell2020; Bell and Jones Reference Bell, Jones, Burton-Jeangros, Cullati, Sacker and Blane2015; Thijs et al. Reference Thijs, te Grotenhuis and Scheepers2020). Compared to these approaches, constructing categorical APC variables comes with the advantage of being able to theoretically contextualize the estimates while at the same time overcoming the problem of multicollinearity (Bell Reference Bell2020, Jocker et al. Reference Jocker, van der Brug and Rekker2024a).
Regarding age groups, we follow categorizations that stem from psychological developmental research (Rekker et al. Reference Rekker, Keijsers, Branje and Meeus2015; Wink and Dillon Reference Wink and Dillon2003) and alter them slightly.Footnote 9 Those age groups represent theoretically meaningful life phases: ‘Late adolescence’ (18–23) when most individuals still live at home and finish education, ‘early adulthood’ (24–34) when individuals have finished their education and have moved out of parental home, ‘middle adulthood’ (35–64) when most individuals are part of the labor market and attitudes mostly stabilize, and ‘late adulthood’ (65+) when individuals retire.Footnote 10 To account for period effects, we add survey years as a discrete variable in all regressions (Grasso et al. Reference Grasso, Farrall, Gray, Hay and Jennings2019). Robustness tests control for the inclusion of a continuous period variable (see Table C4). Following the approach of Shorrocks (Reference Shorrocks2018) and Shorrocks and Grasso (Reference Shorrocks and Grasso2020), we only add an interaction term for cohort.Footnote 11
To account for cohort effects, we construct a theoretically rooted generational classification to constrain the effect of birth years and to be able to meaningfully interpret the obtained estimates (Bell Reference Bell2020). In that regard, studying political generations in a single country has the advantage of overcoming imprecise cross-national generational categorizations (Grasso Reference Grasso2014; Mitteregger Reference Mitteregger2024). Since Switzerland has rarely been the focus of APC electoral research (see Hug and Kriesi (Reference Hug and Kriesi2010) as an exception), we need to come up with a novel cohort categorization that fits this context. Our scheme divides respondents into cohorts based on distinct periods of political and historical socialization, which are outlined below.
First, the ‘Postwar Generation’ (born: 1920–1945) spent its formative years amid the Second World War (during which Switzerland formally maintained its neutrality) and experienced the creation of the ‘magic formula’ that would cement the Swiss governmental composition for decades, as well as the establishment of the pension system in the 1950s.
Second, the ‘Boom Generation’ (born 1946–1961) came of age during a period of economic prosperity and political stability (the ‘golden age’ of the Swiss welfare state; see Obinger (Reference Obinger1998)), with full employment and almost no political changes in national elections. In addition, this generation was born in the years of the highest fertility rates in 20th-century Switzerland (van Bavel and Reher Reference van Bavel and Reher2013), and also experienced the introduction of female suffrage in 1971.
Third, the ‘Movement Generation’ (born 1962–1973) grew up in a decade in which the relatively calm decades after World War II were shaken by youth movements in large cities and the emergence of new parties on the political left (the radical left POCH and the Green Party) and the right (‘Car Party’, National Action). Especially after the Chernobyl disaster in 1986, environmental parties and associations became stronger. The growing salience of ‘green demands’ led to the acceptance of two popular initiatives demanding more environmental protection and the end of nuclear energy (Lorenzini et al. Reference Lorenzini, Monsch and Rosset2021).
Fourth, we identify a ‘Post-EEA Generation’ (born 1974–1985) that came of age in the 1990s, when the decline of the industrial sector and the growth of tertiary employment accelerated. Furthermore, the Swiss party system started to fundamentally transform. Beginning with the direct-democratic rejection of accession to the European Economic Area (EEA) in 1992 (Lauener Reference Lauener2022), the following decade was characterized by the rise of the far-right SVP and increasing polarization (Bornschier Reference Bornschier2010, Reference Bornschier2015; Kriesi et al. Reference Kriesi, Lachat, Selb, Bornschier and Helbling2005). The SVP became the strongest party in 1999, which later led to the first change of government composition after almost 50 years (Häusermann et al. Reference Häusermann, Abou-Chadi, Bürgisser, Enggist, Mitteregger, Mosimann and Zollinger2022).
Fifth, the ‘Polarization Generation’ (born 1986–2001) came of age between the mid-2000s and the end of the 2010s and was socialized in an increasingly polarized political landscape with the far-right and the green-left block both gaining electoral strength (Hänggli and Häusermann Reference Hänggli and Häusermann2015; Häusermann et al. Reference Häusermann, Abou-Chadi, Bürgisser, Enggist, Mitteregger, Mosimann and Zollinger2022). This generation experienced ongoing debates about immigration, with several anti-immigration direct-democratic initiatives succeeding among the electorate. In addition, political parties tried to politicize the urban-rural divide, with the far-right negatively portraying urbanites as the most visible example (Audikana and Kaufmann Reference Audikana and Kaufmann2022, Zumbrunn Reference Zumbrunn2025). Altogether, the three newest generations most strongly experienced the rising salience of cultural issues, often tied to debates about urban and rural spaces, as well as the electoral rise of parties linked to these issues.
Still, while we base our generational classification on theoretical considerations, we cannot fully solve the issue of partly arbitrary generational boundaries. Therefore, generalized additive models (GAMs) are applied as a robustness check.Footnote 12 GAMs have the advantage of smoothing one of the APC parameters of interest while still constraining the others. In these models, the categorical ‘cohort’ term is replaced by a smoothed ‘year of birth’ term, which allows us to plot smoothed splines (Grasso Reference Grasso2014, Reference Grasso, Farrall, Gray, Hay and Jennings2019; Mitteregger Reference Mitteregger2024). We also run additional robustness checks with different cohort categorizations to test if the cohort effect holds.Footnote 13
Models, controls, and weights
Age, period, and cohort are purely exogenous variables, and including sociodemographic covariates can, therefore, mainly inform us about compositional effects of the APC variables (Lichtin et al. Reference Lichtin, van der Brug and Rekker2023; van der Brug and Rekker Reference van der Brug and Rekker2021). Still, because controlling for sociodemographic variables can help understand how compositional differences between generations contribute to urban-rural differences, full models control for several sociodemographic variables, namely the respondent’s gender, highest level of education (low, medium, high), class (eight ‘Oesch classes’), trade union membership, income (quintiles adjusted for household size), and language region (German, French, Italian). These variables are chosen because they may have a long-lasting effect on party preferences and attitudes that differ between cohorts in different places of residence, and controlling for them partly isolates the contextual effect of being socialized in a particular political surrounding.
An important debate regarding the urban-rural divide in political attitudes concerns the question how much individuals self-select into certain neighborhoods and whether moving patterns explain the political divides between different places. Unfortunately, we do not have data on where respondents spent their adolescence. To still partly address this, we conduct a robustness test that excludes all respondents from the analysis who indicate having moved to a different canton in the past five years (see Table C8 in online Appendix C). In Switzerland, moving is relatively rare, with less than 3% of the population moving out of their municipality in a year. Thus, most moving occurs within the same municipality, and only a small share of movers (12%) move further away than 30 kilometers (see BfS 2024). We are thus confident that this robustness test can largely rule out the effect of moving patterns.
Because all our dependent variables are dichotomous, we run APC-logistic regressions, including all three APC components and an interaction term for urbanity and cohort. All regressions contain survey weights (‘total weight’ for canton, turnout, and party choice) controlling for regional oversampling and selection bias in SELECTS (Tresch et al. Reference Tresch, Rennwald, Lauener, Lutz, Alkoç, Benvenuti and Mazzoleni2024). Because log-odds ratios are hard to interpret, we visualize differences between cohorts with predicted probabilities. They also allow us to compare cohorts besides the reference category. To facilitate interpretation, these plots are based on regressions in which we disaggregate population density into a categorical variable consisting of equal-sized quintiles of this variable. Based on these estimates, we predict probabilities for the place and generation interaction.Footnote 14
Results and discussion
Attitudes
Looking at APC effects on attitudes first, Table 2 presents the results from the APC logistic regressions on attitudinal differences. Regressions 1 and 4 display the APC effects, as well as the direct effect of urbanity on cultural attitudes. Regressions 2 and 5 then add the interaction between age and urbanity. Finally, Regressions 3 and 6 add sociodemographic controls. To facilitate readability, Table 2 only displays the effect of age and urbanity and the interaction effect between cohort and place. Table B1 in online Appendix B displays all covariates included in Regressions 3 and 6.
Table 1. Cohort categorization: Swiss political generations in the 20th/21st centuries
Table 2. APC logistic regressions: attitudes
Note: ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
First, there is a similar pattern for both attitudes: citizens who live in more densely populated (i.e., urban) municipalities are more likely to hold progressive positions. Second, looking at the direct cohort effect, results are mixed for both attitudes: while newer cohorts are more likely to favor environmental protection and to have progressive immigration attitudes, these effects are not significant for the youngest cohort. Before discussing the interaction of cohort and place, the other APC components deserve some attention. Looking at the full models, age is a relevant predictor of attitudinal differences, with middle-aged respondents being less progressive than the youngest citizens. Looking at the survey year variable, we find significant period effects on environmental attitudes until 2019 (with all individuals becoming more supportive), while Swiss citizens have not uniformly become more or less progressive on immigration over the observed period, but there has been a trend of more immigration-skeptic stances in recent years.
Turning to the interaction between cohort and urbanity, more recently socialized citizens are more likely to hold progressive attitudes when they live in urban rather than rural municipalities. However, this differs between issues: while this is the case for all cohorts socialized after the 1960s when looking at immigration attitudes, only three of the four newer cohorts display an urban-rural difference in their attitudes towards the environment when compared to the oldest generation (when not accounting for controls). Still, the results partly lend support to our expectations put forward in H1.
Figures 2 and 3 visualize predicted probabilities. As noted above, we plot these probabilities relying on a categorical population density variable. Figure 2 shows that newer cohorts in urban contexts have become more progressive on immigration, whereas very rural cohorts hardly differ in their immigration stances. Generally, the main difference lies between the two oldest generations and the three newest cohorts, with the movement and the post-EEA generation displaying the strongest urban-rural difference, whereas urbanites from the very newest cohort tend to hold slightly more conservative attitudes than earlier urban cohorts.
Figure 2. Predicted probabilities: immigration attitudes.
Note: Predicted values are calculated from logistic APC regression estimates, with population density quintiles. The plots show 95% confidence intervals.
Figure 3. Predicted probabilities: environment attitudes.
Note: Predicted values are calculated from logistic APC regression estimates, with population density quintiles. The plots show 95% confidence intervals.
Figure 4. Predicted probabilities: voting for a left-wing party.
Note: Predicted values are calculated from logistic APC regression estimates, with population density quintiles. The plots show 95% confidence intervals.
Figure 5. Predicted probabilities: voting for a far-right party.
Note: Predicted values are calculated from logistic APC regression estimates, with population density quintiles. The plots show 95% confidence intervals.
Figure 3 presents the predicted probabilities for environmental attitudes. The plot shows that the main difference regarding environmental issues lies between the Postwar Generation (where urban-rural differences are negligible) and the Boom, Movement, and Post-EEA Generations, with these cohorts showing similar tendencies of more progressive urban individuals. However, there is no significant urban-rural difference for the newest generation. Generally, the differences regarding the environmental question are rather negligible.
Voting behavior
Left-wing parties
Do we see similar urban-rural cohort differences in voting behavior? To answer this question, Table 3 shows the results for left-wing parties. As in the results presented above, only the relevant age and urbanity variables are displayed (see Table B1 in online Appendix B for all covariates). Model 1 presents the direct effects of age, period, cohort, and urbanity on the likelihood of voting for a left-wing party, whereas Model 2 adds the interaction of urbanity and cohort. In Model 3, we account for covariates. We find a non-significant negative period effect, while life-cycle effects exist for 18–23-year-old voters (who are more likely to vote for a left-wing party when controls are added) compared with middle-aged voters. Looking at the cohort variable, all newer cohorts are more likely to vote for a left-wing party than the oldest cohort. Furthermore, population density has a significant positive effect on the likelihood of voting left, confirming earlier findings on the macro-level (Huijsmans and Rodden Reference Huijsmans and Rodden2025; Strebel and Müller Reference Strebel and Müller2023).
Table 3. APC Logistic Regressions: Vote Choice (Left-Wing Parties)
Note: ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
Turning to the interaction of cohort and urbanity in Models 2 and 3, we find a significant interaction effect of population density for most cohorts when compared with the Postwar Generation: cohorts socialized from the 1980s onward are more likely to vote for left-wing parties when they live in an urban context. All effects remain significant when controlling for sociodemographic variables. This supports H2a and shows that while newer cohorts have become more likely to vote for left-wing parties in Switzerland, this is even more strongly the case in urban areas. We demonstrate that this effect is most pronounced for the Greens, especially among newer cohorts: there is no significant interaction effect of urbanity for the very youngest cohort when solely looking at the SP, but the effect for the Green party remains significant (see Table B3 in online Appendix B).
Far-right parties
In a final step, we look at the other pole of the cultural dimension, the far-right. Again, Table 4 only displays the age variables, as well as the effect of urbanity (see Table B1 in online Appendix B for all covariates). Like the earlier regressions, Model 1 only displays the direct effects of age, period, cohort, and urbanity on the likelihood of voting for a far-right party, whereas Model 2 shows the interaction of urbanity and cohort. In Model 3, we add covariates. Model 1 reveals a significant and strong positive period effect (with all voters becoming more likely to vote for the far-right over the observed period), whereas age effects are less relevant to explain far-right voting (except for pensioners). Looking at the (direct) cohort effect, the far-right is more popular among the earliest cohort. Moreover, the direct effect of urbanity on voting far-right is significant and negative.
Table 4. APC Logistic Regressions: Vote Choice (Far-Right Party)
Note: ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001.
Turning to the interaction effect in Models 2 and 3, the effects are inverse to those for left-wing parties: for the Boom Generation, there is no interaction effect of urbanity on the likelihood of voting for the far-right. However, the effect of urbanity is significant and strong for those generations socialized from the mid-1980s onwards and persists when controlling for sociodemographics. Robustness tests (see Figure B6 in online Appendix B) provide more detailed evidence that voters born between 1970 and 1999, who were socialized when the SVP broke through electorally, show the strongest effect here. Generally, this lends support to H2b. To visualize how other cohorts differ, we again calculate and plot the predicted probabilities by breaking population density into quintiles.Footnote 15 Figure 5 shows that among the two oldest cohorts, around 20% of urbanites vote far-right, whereas only around 12% of newer generations in urban places vote far-right. However, there is no tendency for more recently socialized rural people to be less likely to vote far-right: if anything, they are even slightly more likely to vote far-right than less recently socialized rural voters. In rural municipalities, around 40% of the newest generations vote far-right. This shows that, in line with attitudes towards immigration, it is not only young left-wing urbanites who seem to drive the rural-urban divide.
Conclusion
Is generational replacement fueling Western Europe’s urban-rural divide, and are newer cohorts increasingly divided along this line? This paper studies these questions by focusing on Switzerland, a country that has experienced both a strong polarization of urban-rural differences and an increased salience of cultural issues that has led to electoral realignment. This makes the country a well-suited case for studying urban-rural cleavages over time with a perspective on generational differences.
Our analysis demonstrates an interaction effect for cohort and place, which is robust for different operationalizations of cohort, age, and urbanity and consistent with earlier findings from Germany (Haffert and Mitteregger Reference Haffert and Mitteregger2023) and Sweden (Valldor Reference Valldor2025). First, newer generations show the strongest urban-rural divides on the issue of immigration and weaker interaction effects on environmental issues. Second, newer cohorts in urban areas are more likely to vote left-wing, whereas the effect of urbanity is less strong for individuals born before the mid-1960s. Third, support for the far-right shows an inverse relationship: newer cohorts are less likely to vote far-right if they live in urban areas, which is not the case in rural areas. If anything, there is even a tendency for newer rural cohorts to be more likely to vote for the far-right than earlier rural cohorts. Notably, the two earliest cohorts stand out as being distinct from other generations. This is particularly relevant, as these two cohorts still made up a third of the eligible voting population in the Swiss elections of 2023 and even 40.5 percent of the actual voters in that same election (Tresch et al. Reference Tresch, Rennwald, Lauener, Lutz, Alkoç, Benvenuti and Mazzoleni2024). The urban-rural divide is thus likely to deepen in future elections as generational replacement continues. Overall, the findings point to widening within-cohort gaps in attitudes and voting behavior that underpin the urban-rural divide, which increasingly structures Swiss politics.
Related to this, it might not be surprising that Swiss parties have increasingly tried to politicize this divide, with direct democratic proposals specifically targeting urban or rural citizens (Audikana and Kaufmann Reference Audikana and Kaufmann2022, Zumbrunn Reference Zumbrunn2025). This politicization is also particularly crucial in the highly federalist Swiss political context. First, rural regions are strongly overrepresented in the bicameral parliament due to the federalist character of the upper chamber. Second, each direct democratic proposal to change the constitution also needs to reach a ‘cantonal’ majority. Thus, with increasing urban-rural divides, smaller cantons could be further able to veto policy change. These institutional particularities make it attractive for Swiss parties to appeal to voters in rural regions.
Given that parties increasingly appeal to this divide, and given that people already tend to identify with their urban or rural place of residence, emerging cohorts may develop a political identity that is structured around this divide. As Tilley and Hobolt (Reference Tilley and Hobolt2023) argue for the case of Brexit, intense political campaigns can create non-partisan political identities that structure specific policy preferences and can be politically mobilized. If this happens for the urban-rural divide, this would further entrench it among newer cohorts and would make it even more likely that urban-rural conflict will structure Swiss politics in the foreseeable future.
Having said this, the results have broader implications beyond the Swiss context: cohorts are not becoming uniformly more progressive, as newer cohorts in rural areas do not seem to be more progressive than their predecessors. If anything, they tend to be slightly more conservative, while more recently socialized citizens in urban areas are driving value change in these increasingly cosmopolitan places. This confirms earlier findings that could not be tested directly with a cohort analysis (Haffert and Mitteregger Reference Haffert and Mitteregger2023; Huijsmans et al. Reference Huijsmans, Harteveld, van der Brug and Bram2021; Valldor Reference Valldor2025). As our results point in a similar direction as other research, we are confident that the findings are likely to not be a specific Swiss story. This is especially the case as Switzerland lacks rural regions that are structurally and economically ‘left-behind’ (Zollinger Reference Zollinger2024) and because distances between rural places and urban centers are smaller than in comparable European countries. As the results still reveal growing cohort differences along the urban-rural divide, we might find even bigger cohort divides in other European countries when looking at the role of place for political behavior and attitudes. That said, because rural regions in other countries are often less prosperous than in Switzerland, it may be the case that the divide in some of these countries has a stronger economic dimension.
However, our paper has several limitations: first, because it does not use panel data, it cannot properly test for life-cycle differences. Second, it shows descriptive rather than causal effects and is not able to test a specific mechanism (although the problem of reverse causality is unlikely to exist for cohort effects). Third and relatedly, the effect of moving (and related sorting processes) cannot be properly captured (see Maxwell Reference Maxwell2019, Reference Maxwell2020), making it impossible to disentangle contextual and compositional effects. Nevertheless, to the best of our knowledge, this paper is the first to attempt to disentangle APC effects in interaction with urbanity to explain attitudes and voting behavior in a multiparty context. While such an analysis remains methodologically and theoretically challenging, the results suggest that disentangling these effects can provide a more nuanced picture.
Overall, the paper underlines the relevance of understanding cohorts not as homogeneous blocs, but rather as groups consisting of ‘generational units’. In line with findings on within-cohort differences for gender and education (Harsgor Reference Harsgor2018; Schäfer and Steiner Reference Schäfer and Steiner2025; Shorrocks Reference Shorrocks2018), it is important to stress that differences can exist between and within generations. Regarding our findings, generational replacement may make urban-rural differences a persistent feature of European political systems, with diverging place-based socialization experiences causing geographical stratification within and not (only) between generations.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1475676525100200.
Data availability statement
The data that supports our findings is publicly available via SWISSUBase: Selects 1995–2019 Post-Election Survey: https://doi.org/10.48573/q99z-aa77, reference number 20738. Selects 2023 Post-Election Survey: https://doi.org/10.23662/FORS-DS-495-2, reference number 495. And via the Swiss Federal Office of Statistics (BfS): https://www.bfs.admin.ch/bfs/de/home/statistiken/regionalstatistik/atlanten/statistischer-atlas-schweiz.html.
Acknowledgments
Previous versions of this paper were presented at the ‘Territorial Cleavages’ Workshop in Bergen in 2023, the ‘Challenges to Democratic Representation’ Seminar at the University of Amsterdam in 2023, the Workshop on ‘Regional inequalities in labor markets and politics’ at the University of Lausanne in 2023, the InEquality Conference 2024 in Konstanz, and the SPSA 2024 in St. Gallen. We wish to thank the participants on these occasions for their helpful comments. We are especially grateful for valuable and insightful input from Kiran Auerbach, Matthias Enggist, Enrique Hernández, Twan Huijsmans, Thomas Jocker, Daniel Oesch, Wouter van der Brug, and Malte Wehr on earlier versions of this article.
Financial support
The research for this article has benefited from support from the Swiss National Science Foundation (snsf Doc.CH Grant No. 203868).
Competing interests
The authors report no conflicting financial or non-financial interests.
Open access
