Generational (re)alignment: emerging issues and new voters in Western Europe

Abstract

Voting behaviour in Western Europe has become increasingly structured by attitudes towards socio-cultural issues, such as immigration and European unification. This study examines whether generational differences affect this form of realignment. By combining data from the ESS and the CHES, it explores party–voter alignment on specific issues across 17 West European party systems from 2002 to 2020. ‘Age-Period-Cohort’ (APC) analyses demonstrate that younger generations are more strongly aligned than older generations on immigration, LGBTQ+-rights, the environment and left–right placement, but not on the EU and income redistribution. The results also show over-time increases in issue alignment for three of the six issues, but these period effects are more country-specific than the generational differences. Consequently, generational replacement can be expected to increase party–voter alignment on the left–right dimension and (some) new issues.

Keywords:

• Generational replacement
• realignment
• issue alignment
• vote choice
• political socialisation

Realignment in Western Europe is often studied as a process by which party–voter ties are increasingly driven by (socio-cultural) issues like immigration and European integration (e.g. Bornschier 2010; Kriesi 2010; Stubager 2013). These studies typically ignore generational differences, even though there are good reasons to expect that younger generations will be more aligned with the positions of the party they support on more recently politicised issues than older generations (e.g. van der Brug and Rekker 2021). If this is indeed the case, we are only seeing the beginning of a realignment process that will increasingly unfold as newer generations replace older ones.

The expectation that party–voter ties differ between generations on specific issues follows from a large literature in political socialisation, which shows that events occurring around the age of roughly 15–25 (the so called ‘formative years’) will have a relatively large impact on political orientations and behaviours later in life (e.g. Bartels and Jackman 2014; Sears and Valentino 1997). While previous generations were socialised when socio-economic issues were prominent in structuring party conflict, the political understandings of today’s youth are shaped when socio-cultural issues increasingly define party competition (e.g. Rekker 2016; Steiner 2023). Moreover, recent generations are more inclined to support new parties prioritising recently politicised socio-cultural issues (e.g. Dolezal 2010; Lichtin et al. 2023; Mitteregger 2024; Schäfer 2022; cf. Lisi et al. 2021). In this article, we study generational differences in issue alignment, the association between voters’ issue attitudes and the positions of the party they voted for on the same issue. Our guiding hypotheses are that newer generations will be more aligned on socio-cultural issues than previous generations, whereas older generations exhibit stronger issue alignment on ‘traditional’ issues that were more salient during their formative years.

An emerging literature suggests that this could indeed be the case, indicating that the party choice of younger generations is more motivated by ‘newer’ cultural issues like immigration attitudes and Euroscepticism, while older generations hold more steadfast to their traditional political affiliations (e.g. Gougou and Mayer 2012; Wagner et al. 2012; Walczak et al. 2012). However, these studies face a significant limitation as they cannot distinguish between life cycle effects (changes occurring as people grow older, also known as age effects), period effects (events and shocks affecting individuals of all ages and generations), and cohort differences (persisting differences resulting from distinct formative experiences, also referred to as generational effects). As these effects are interdependent, we cannot confidently conclude whether generational replacement will lead to more aligned electorates in Western Europe without disentangling the effects of 1) age, 2) period, and 3) cohort (APC).

To our knowledge, a recent study by van der Brug and Rekker (2021) is the only APC-study of the correlates of party preferences. Examining the case of the Netherlands, they find a pattern in which dealignment over time (e.g. on religion and social class) coexists with realignment over generations (e.g. on education and new cultural issues). However, it remains an empirical question whether the findings of this case study can be extrapolated to other Western European democracies.

Our APC-study contributes to this emerging literature by investigating age, period, and cohort differences in issue alignment (the association between issue attitudes and vote choice). Specifically, we focus on issue alignment between 2002 and 2020 across five generations in 17 West European party systems. To do so, we combine data from the European Social Survey (ESS) and the Chapel Hill Expert Survey (CHES) for six major political divides in Western Europe: income redistribution, social liberty (LGBTQ+-rights), the environment, immigration, European unification, and left–right politics.

We find strong support for our hypothesis that new generations are more aligned than older generations on the newer issues. Yet, contrary to expectations, older generations are not more strongly aligned on the more traditional issues such as income distributions. The overall picture that emerges is that there is increasing issue alignment over the past two decades, which is partially (and in some cases largely) the result of new generations being more aligned to the parties they voted for on some issues (cohort effects). We also delved into the question whether these patterns are the result of changes in the composition of the electorate, in terms of class, education and religion. However, most of the patterns remain, even when controlling for these factors. Finally, despite the obvious differences in the recent histories of the EU-member states (for instance the impact of the financial crisis in Ireland and Southern Europe), these generational patterns are very similar across the 17 countries. By contrast, we observe more country-specific patterns when looking at over-time trends.

Theory and hypotheses

Defining issue (re)alignment and dealignment

The notion of alignment refers to long-lasting ties between parties and voters, usually based on social group membership (e.g. social class or religion) or stable ideological orientations. Dealignment occurs when these stable determinants of vote choice lose their relevance without being replaced by other stable long-term predictors. Realignment means that some long-term determinants of the vote are losing their ability to create stable party–voter ties, while being replaced by other stable factors that connect (groups of) voters to parties.

Research on realignment and dealignment has increasingly shifted its focus from traditional social cleavages (such as class and religion) to the importance of values and issues in connecting voters to parties (and vice versa). Our study builds upon this development in the literature. We do not focus on alignment based on social groups, but on issue alignment: the strength of the relationship between voters’ policy preferences and their vote choice. Issue realignment occurs when parties and voters are increasingly aligned on new issues and less on traditional issues.

Issue alignment is a two-way street. On the one hand, voters can simply choose a party closest to their own political beliefs and preferences. On the other hand, voters may adopt the policy positions of a party they prefer for other reasons. We are not concerned with disentangling whether changes over time or differences between generations in issue alignment are primarily driven by top-down (parties influencing voters) or bottom-up processes (voters choosing the ideologically closest party). Our objective is to assess to what extent patterns of issue (re)alignment exist across all generations and age groups alike (period effects) and to what extent these patterns differ systematically between generations and age groups.

Issue (re)alignment in a changing political landscape

Dealignment, realignment, and alignment trends are typically studied as period effects in which macro-level social changes affect how party–voter ties are structured. Evidence of dealignment comes mainly from studies showing the diminished relevance of traditional cleavages like religion and class for party choice (e.g. Dalton and Flanagan 1985; Evans and Tilley 2012; Franklin 1992). Research shows that an ideological left–right division was important in structuring party–voter ties since the 1970s in most Western European countries (e.g. Fuchs and Klingemann 1990; van der Eijk et al. 2005). Until the 1980s left–right was primarily interpreted in terms of socio-economic policies. More recently, events and trends like the globalisation of world markets, European integration, climate change, the aftermath of 9/11, and mass migration have placed greater emphasis on cultural issues in structuring party competition (e.g. Hooghe and Marks 2018; Kriesi et al. 2008). In response to these changes at the elite level, voters seemingly prioritise specific policy preferences when deciding on their vote choice (e.g. Costello et al. 2021; Dassonneville 2016; Walczak et al. 2012).

While issue alignment can arise from a shift in voting considerations in response to macro-level changes, it may also reflect that many voters nowadays simply have a larger set of parties to choose from (e.g. Mitteregger 2023). This in turn, could increase the likelihood of supporting a party they agree with on issues they consider important. Although the exact dimensionality of party competition in Western European party systems remains subject to debate (e.g. De Vries 2018; Hooghe and Marks 2018; Kriesi et al. 2008; van der Brug and Van Spanje 2009), there seems to be a consensus that issues like immigration, European unification, and environmental policies increasingly shape party competition and voting behaviour. This phenomenon has also been linked to the electoral fortunes of parties campaigning on ‘new’ socio-cultural issues (e.g. Bornschier 2010; Dennison and Geddes 2019; Hooghe and Marks 2018; Kriesi 2010; Kriesi et al. 2008). These newer parties provide voters options that align more closely with their beliefs without necessitating a complete overhaul of their belief systems. Moreover, even when these new parties do not achieve major electoral breakthroughs, they can indirectly impact voters’ choices by bringing new issues to the political agenda. Within Western European party systems, issues like immigration now receive more attention in party programs (Green-Pedersen and Otjes 2019), election campaigns (Hutter and Kriesi 2022), and in media coverage (Grande et al. 2019). Based on this discussion, we expect to find a general period effect in issue alignment:

H1: Over time, voters of all generations and age groups will exhibit greater alignment than in previous years on recently politicized issues, such as immigration, LGBTQ-rights, European unification, and the environment.

Generations and the formative years

Generational research repeatedly demonstrates that experiences during people’s formative years can have a lasting impression on their attitudes and behaviours later in life (e.g. Carmines and Stimson 1989; Hooghe 2004; Norris and Inglehart 2019; van der Brug and Franklin 2017). Most citizens develop their core orientations and values during their ‘formative years’, roughly between the ages of 15 and 25 (e.g. Bartels and Jackman 2014; Rekker et al. 2019; Schuman and Rodgers 2004). As people grow older, their political outlooks crystallise through partisan loyalties, habits, and ideological ties (e.g. Dinas 2013; Franklin 2004; Krosnick and Alwin 1989; Sears 1983). This reduced attitudinal flexibility facilitates the emergence of political generations: distinct groups with shared political outlooks shaped by their formative experiences (Neundorf and Smets 2017; Stoker 2014).

Broadly speaking, there are two sets of reasons why issue alignment could differ between generations. From a top-down or elite-driven perspective, the viewpoints of political generations can be shaped by the issues made salient by political elites during their formative years (e.g. Steiner 2023). When political elites place new issues on the political agenda, they shape the political socialisation of ‘new’ voters. As ‘the most likely agents of political change’, young voters are not yet bound by deeply engrained party loyalties or past voting habits (Carmines and Stimson 1981: 1989). Instead, their learning experiences are based on the current state of the party system: which political parties compete for power and the issues they emphasise. By contrast, older citizens have already learned to navigate the party system, making them more likely to vote in line with issues that structured politics when they grew up. While older citizens can certainly adapt to and learn about new political issues later in life, they are generally more resistant to change than younger citizens who have not yet fully experienced their formative life phase.

From a bottom-up perspective, issue alignment may differ between generations because of the society that each cohort grew up in. For example, Inglehart (2015[1977]) reasoned that Baby Boomers prioritised new post-materialist issues over older material issues because they experienced their formative years under more favourable economic conditions. Likewise, cohorts that grew up since the rise of globalisation in the 1980s may have formed more outspoken opinions about new issues as a result of being politically socialised amid globalisation and increased immigration (Rekker 2018). Considering both the bottom-up and top-down perspectives, we expect the following trends:

H2a: New generations will be more strongly aligned than older cohorts on recently politicized issues, such as immigration, LGBTQ+-rights, European unification, and the environment.

H2b: Older generations will be more strongly aligned than new cohorts on ‘traditional’ issues (income redistribution).

While our hypotheses focus on specific policy considerations, left–right placement is key to understanding the alignment between parties and voters. As an ideological ‘super-issue’, the left–right dimension encapsulates both traditional socio-economic policies and new socio-cultural issues (see e.g. De Vries et al. 2013). However, different generations may differ in their substantive interpretation of what it means to be ‘left’ or ‘right’ (e.g. Rekker, 2016; Steiner, 2023). Because of these differences, we cannot derive hypotheses as to which generations will be more aligned on left–right. We therefore study party–voter alignment on the left–right dimension as an exploratory part of the analysis.

Compositional changes

In our theoretical discussion so far, we discussed mainly how new issues become politicised due to societal changes and how different generations were socialised in different conditions. However, a third factor worth exploring is the way the electorates are changing due to secularisation and post-industrialisation. These changes have a clear generational component, as younger generations tend to be more secular, higher educated, and work in different professions than older generations. These compositional differences may result in generational differences in issue alignment. For example, newer generations may be more aligned on issues as they tend to be better educated than previous generations (e.g. Puga-Gonzalez et al. 2022). As politics is increasingly structured around multiple issues, choosing a party in close ideological proximity requires a certain level of political sophistication (e.g. Dalton 1984; Dalton et al. 2000). Being higher educated may help newer generations like the Millennials to choose parties in ideological proximity to themselves (e.g. Aldrich et al. 2018; Boonen et al. 2017), and be cognisant of elite cuing in the first place. As a result, educational expansion may reinforce issue alignment of younger generations on recently politicised issues (H2a), while counteracting our expectation that older generations will be more aligned on traditional economic issues (H2b).

On the other hand, if older generations are more likely to be manual workers, they may be more likely to experience some of the negative consequences of transnationalism (e.g. Kriesi et al. 2008). This could make the newer issues of immigration and European unification more salient to them, potentially leading to more alignment on these issues. So, compositional changes might lead to more alignment on some issues and less alignment on others. Since we cannot really predict how this plays out, we decided not to formulate hypotheses regarding the consequences of changes in the electorate in terms of class, education, and religion. So, we will first test our hypotheses in models without control variables to estimate the APC-components. We will then add the control variables to assess to what extent these estimated age, period and cohort effects are the result of changes in the composition of the electorate.

Method

Data

In order to answer to what extent age, period, and cohort differences drive issue alignment, we combine data from the European Social Survey (ESS) and the Chapel Hill Expert Survey (CHES). The ESS provides information on how respondents (stated to have) voted during the last national election, their issue attitudes, and their issue demographic characteristics. We use all ESS rounds released at the time of writing (1–10), covering vote choice at biennial intervals from 2002 until 2020 in 17 West European party systems.¹ The analysis includes all respondents who reported voting for a party covered in the corresponding CHES round and provided response to the selected ESS questions. We omit all other respondents through list-wise deletion.

We use the CHES to construct the outcome variable: the issue position of the political party the respondent voted for during the last national election. To estimate issue alignment, we predict this dependent variable by the respondent’s own position on this issue. For example, when measuring issue alignment on immigration, we estimate the degree to which respondents’ own attitudes towards immigration predict the position on this issue of the party they voted for. The dependent variable, the issue position of a party, represents the average of the perceived position of a party by all experts answering the CHES questionnaire. The CHES measures party positions on a range of specific issues, allowing us to construct our outcome variables without making strong assumptions about the ideological identity of parties. We determine issue alignment by predicting the position of the party a respondent voted for in the last national election (a position measured by CHES) by the respondent’s own position on that same issue.²

Table 3 in Online Appendix B lists the selected CHES surveys used to establish our outcome variables. Since the ESS and CHES are not fielded simultaneously, we rely on older expert surveys to impute party positions. If a party is missing in one of these CHES rounds but is present in another survey, we use the earliest data point available to determine a party’s position. We thus assume that party positions are relatively stable.³

There are sometimes substantial time differences between the ESS-surveys and national elections, making the recollection of a respondent’s previous voting behaviour prone to biases (e.g. Van Elsas et al. 2014). Therefore, we reproduced our study using the political party the respondent ‘feels closest to’ when filling out the ESS-survey to construct our outcome variables. As we show in Online Appendix D (Table 5), these models yield slightly stronger estimates of cohort effects.

Variable selection

We study issue alignment on the following six political issues:⁴

1. Redistributive policies are a ‘traditional’ issue in West European party politics. The ESS asks respondents whether the government should do more to reduce income levels. In contrast, the CHES asks experts to position parties on wealth redistribution from the rich to the poor.

2. Social liberty concerns another important cleavage in many West European democracies. The ESS asks respondents about their views on whether homosexual men and lesbians should have the freedom to live as they wish. Comparatively, the CHES covers a broader range of social liberty issues, including both same-sex rights and gender equality.

3. Environmental policies emerged as a salient issue in West European politics from the 1970s and have continued to grow in importance. Unfortunately, the ESS lacks a longstanding measure of environmental policy attitudes, instead assessing how important respondents consider looking after the nature/environment as part of their personal values.⁵ This is a relatively weak indicator of environmental attitudes since it taps into personal values rather than policy preferences (cf. Steiner 2023), so the measure may underestimate the level of issue alignment. However, since we use the same measure for all years and generations, we have no reason to think that it is not suitable to observe period effects or generational differences. The CHES assesses the trade-off between economic growth and environmental protection.

4. European integration has long been considered a secondary issue in national elections, but recent research demonstrates more EU-issue voting (e.g. De Vries 2010), which has a generational component to it (van der Brug and Rekker 2021). The ESS inquires attitudes vis-a-vis European unification, while the CHES examines party positions on European integration.

5. Immigration attitudes have attracted immense attention in both scholarly and public debates over the past few decades. To model (anti-)immigrant attitudes, we compute a scale by averaging responses on three questions in the ESS: whether immigrants undermine or enrich a country’s cultural life, the extent to they make the country a worse or a better place to live, and whether immigration is good or bad for the country’s economy (Cronbach’s α = .86). The CHES assesses liberal to strict party positions on immigration.

6. Left–right positions are asked similarly in the ESS and CHES. Respondents place themselves on a left–right scale, while experts position parties on the same scale, ranging from ‘(extreme) left’ to ‘(extreme) right’.

These six issues encompass a wide range of political orientations, varying in salience across time and space.⁶ A complete list of the selected variables and their question formulations in the ESS and CHES is provided in Online Appendix A. Although the same variables in the CHES and ESS have been compared to study party–voter and government–voter issue incongruence (e.g. Devine and Ibenskas 2021; Rosset and Stecker 2019; Stecker and Tausendpfund 2016), one potential pitfall is that the wordings of the items are sometimes slightly different in the two data sets. This would be problematic in a study of issue voting, which employs distances between parties and voters to predict party choice. However, in our design, we predict the position of the party one voted for by a respondent’s own issue attitudes. The fact that these are measured on different scales is not a problem, as long as the two are functionally equivalent. To the extent that (slight) differences in wordings reduces the strength of the estimated degree of issue alignment, it will do so in a similar way at different moments in time and for different generations and age groups. So, there are no reasons to think that this undermines the validity of our estimated APC-effects. However, when interpreting the results, we should be aware of the fact that we cannot directly compare the strength of the issue alignment between different issues.

To ensure the robustness of issue alignment on immigration, we present APC models using alternative independent and outcome variables focusing on multiculturalism and the assimilation of immigrants in Table 8 (Online Appendix D).

Demographic variables

Although age, period, and cohort are purely exogeneous variables, controlling for demographic shifts offers valuable insight into whether compositional differences between generations and changes over time contribute to patterns of issue alignment. New generations are on average higher educated and more secular than older generations (e.g. Lindskog and Oskarson 2023; Puga-Gonzalez et al. 2022). This may help them choose a party in close proximity to their own positions (e.g. Aldrich et al. 2018; Inglehart 2015[1977]), whereas older generations may be more likely to choose a party solely based on their religious affiliation or social class. To assess whether generational differences stem from socialising experiences or are the result of compositional differences, we also present models that control for religion, education, and class (a linear 5-point scheme of occupational status) differences (Oesch 2006).

Age, period, cohort, and the identification problem

An important concern for generational research is that age, period, and cohort effects are perfectly multicollinear, as: $$\mathit{\text{Cohort}}=\mathit{\text{Period}}-\mathit{\text{Age}}$$

In order to circumvent this ‘identification problem’, we place theoretically informed constraints on two of these components following Kritzer (1983), since constraining only one APC component requires heroic assumptions about the nature of these effects (Bell 2020). The validity of these constraints is vital to provide meaningful estimates of the three APC effects (Glenn 1976; Thijs et al. 2020). As recommended by Thijs et al. (2020: 17–18), we assess the robustness of our findings by comparing models with varying constraints for period and cohort which we discuss below.

We control for period effects in two ways. First, we introduce a dummy variable for every country-period combination (dyads). This approach fully controls for all period effects, all country differences, and all country-specific period effects, ensuring that the estimated effects of age and generation cannot be contaminated by election specific idiosyncrasies. However, the high number of dummies make it challenging to interpret pooled issue alignment trends across all 17 party systems. To be able to interpret the period effects, we implement a second set of models with a series of dummy variables for each ESS round and another set of dummies to control for country differences. Since the cohort and age estimates of both model specifications produce the same patterns and conclusions, we present the results from the second set of models as they allow us to infer issue alignment trends (period effects). For the sake of comparison, age and cohort effects derived from the country-period dyad models are presented in Table 6 in Online Appendix D. Both model specifications cluster the robust standard errors by country-period dyads.

Our theoretically informed specification of generations in Table 1 follows Grasso’s study on the political participation of political generations in Western Europe, which roughly coincides with both top-down and bottom-up factors structuring the political socialisation experiences of distinct political generations (Grasso 2014). Although idiosyncratic events may shape the formative years within a domestic context, the emergence of these generations roughly coincides with the onset of broader transnational societal trends like the foundation of the welfare state and globalisation (e.g. Hooghe and Marks 2018; Mitteregger 2024). For example, while the post-WWII generation grew up during an era of mass parties and frozen cleavages in times reconstruction following the war, the 80s generation was politically socialised amidst a period of the Chernobyl disaster when green parties started gaining electoral success in many Western European party systems. These generations also exhibit distinct voting patterns across Western Europe (Mitteregger 2024).

Table 1. Cohort classification and their political socialisation experiences, based on Grasso (2014).

	Post-WWII	60–70s	80s	90s	Millennials
Formative years	1946–1965	1966–1977	1978–1988	1989–2001	2002–
Year of birth	1926–1945	1946–1957	1958–1968	1969–1980	1981–1996
Political socialisation context	Rise of the welfare state, Cold War, Berlin Wall, frozen cleavages, and party systems	Protest movements, Vietnam War, decline of cleavage voting, secularisation	Chernobyl disaster, electoral breakthrough Green parties	Emergence of the EU, globalisation, fall of the Berlin Wall, first successes far right	9/11 aftermath, EU enlargement, European refugee and Eurozone crises, BREXIT, electoral breakthrough radical right-wing populism

We adjust Grasso’s cohort classification in two ways. First, we exclude respondents socialised before World War II due to data limitations. Second, we add the Millennial generation, which includes all respondents born from 1981 until 1996 (Dimock 2019).⁷ To ensure the robustness of our findings, we replicate our study after constraining cohorts by birth decennials (e.g. 1930–39, 1940–49, etc.). These models yield similar generational patterns as shown in Table 7 in Online Appendix D.

Although the literature does not provide a strong enough basis to develop hypotheses regarding how age informs issue alignment (e.g. O’Grady 2022), it is essential to account for life cycle effects to meaningfully estimate period and cohort effects. Age may play a role in shaping individuals’ issue priorities as they undergo psychological development and political learning, often coinciding with changes in social standing such as entering the job market, completing education, or starting a family. Moreover, people may simply continue to learn about politics at a later age. Drawing on developmental psychological research, we incorporate theoretical constraints supported by studies indicating that the most significant life cycle effect occurs during adolescence (ages 16–25). As individuals progress through early adulthood (26–35), middle adulthood (36–65), and late adulthood (65+), their attitudes tend to stabilise (e.g. Bartels and Jackman 2014; Dassonneville 2016; Rekker et al. 2015; Steinberg 2010). By coding age groups accordingly, we remove much of the multicollinearity from our models.

Model specification

We focus on period and cohort differences in issue alignment and estimate separate models for all six outcome variables. Each of these models predicts the position of the party a respondent voted for on an issue. The main effect provides a baseline of party–voter alignment on the issue. By interacting the respondents’ attitudes on that same issue with their age, period, and cohort membership, we can detect life cycle effects, period effects and generational differences in issue alignment. Since the main effect can be expected to be positive, a positive and significant interaction effect indicates that the specific group is more strongly aligned on this issue compared to the baseline group. Next, we add interactions between the demographic controls (religion, education, and social class) and a respondents’ attitude to assess the extent to which the APC-effects can be attributed to changes in the electorate’s composition.

We start with a series of pooled analyses, in which we control for country-level differences by including country-dummy variables and their interactions with attitudes. In a final set of analyses, we compare the patterns between countries. Before turning to the results, it is important to note that it is not feasible to incorporate multiple issues in a single model. Doing so would make it impossible to establish an outcome variable, namely the position of a party a voter voted for on a specific issue. Consequently, we only compare models with caution since we cannot directly compare the independent effects of different issue attitudes.

Results

Period effects

Our first hypothesis posits that over time, voters of all ages and generations will exhibit greater alignment on recently politicised issues such as immigration, the EU, social liberty (LGBTQ+-rights), and the environment. Our APC-models displayed in Table 2, show some support this idea. Figure 1 illustrates the interaction effects between attitudes and period while holding country-differences, age, and cohort at their mean effects.

Figure 1. Period effects of issue alignment. The figure shows the interaction between a voter’s attitudes and the period in which they filled out the survey on their vote choice during the last national election. The results are presented with and without demographic covariates (education and religion) and are based on pooled fixed-effects regression models presented in Table 2. Data: ESS (voter attitudes and vote choice) and CHES (mean expert party placement on an issue).

Table 2. Results of pooled regression models.

	Redistribution		Social liberty		Environment		EU		Immigration		Left–right
Intercept	0.37*** (0.01)	0.35*** (0.02)	0.35*** (0.01)	0.34*** (0.01)	0.50*** (0.01)	0.50*** (0.01)	0.19*** (0.02)	0.21*** (0.02)	0.44*** (0.03)	0.43*** (0.03)	0.31*** (0.02)	0.31*** (0.02)
Variable	0.13*** (0.02)	0.12*** (0.02)	0.16*** (0.03)	0.12*** (0.03)	0.09*** (0.03)	0.08** (0.03)	0.02 (0.04)	0.03 (0.04)	0.18*** (0.05)	0.17*** (0.05)	0.40*** (0.03)	0.39*** (0.03)
Cohort
*Post-WWII	REF		REF		REF		REF		REF		REF
*60–70s	0.01 (0.01)	−0.01 (0.01)	0.05*** (0.01)	0.04*** (0.01)	0.02** (0.01)	0.01 (0.01)	0.02 (0.01)	0.02 (0.01)	0.07*** (0.01)	0.06*** (0.01)	0.03*** (0.01)	0.02* (0.01)
*80s	0.01 (0.01)	−0.01 (0.01)	0.06*** (0.01)	0.05*** (0.01)	0.07*** (0.01)	0.05*** (0.01)	0.04** (0.01)	0.04** (0.01)	0.09*** (0.01)	0.08*** (0.01)	0.05*** (0.01)	0.04*** (0.01)
*90s	0.01 (0.01)	−0.02 (0.01)	0.07*** (0.01)	0.04** (0.01)	0.08*** (0.01)	0.06*** (0.01)	0.03 (0.02)	0.03 (0.02)	0.13*** (0.02)	0.11*** (0.02)	0.07*** (0.01)	0.04** (0.01)
*Millennials	0.03 (0.01)	0.00 (0.01)	0.09*** (0.02)	0.07** (0.02)	0.10*** (0.02)	0.07*** (0.02)	0.06** (0.02)	0.06* (0.02)	0.16*** (0.03)	0.14*** (0.02)	0.09*** (0.02)	0.06** (0.02)
Period
* Round 1	−0.01 (0.02)	−0.01 (0.02)	−0.03 (0.02)	−0.02 (0.02)	0.00 (0.02)	0.00 (0.02)			−0.03 (0.04)	−0.04 (0.04)	0.00 (0.03)	0.00 (0.03)
* Round 2	REF		REF		REF		REF		REF		REF
* Round 3	0.01 (0.02)	0.01 (0.02)	−0.02 (0.02)	−0.02 (0.02)	0.01 (0.02)	0.01 (0.02)	−0.01 (0.03)	−0.01 (0.03)	−0.03 (0.04)	−0.03 (0.04)	0.05 (0.03)	0.05 (0.03)
* Round 4	0.01 (0.02)	0.00 (0.02)	−0.05* (0.02)	−0.04 (0.02)	0.01 (0.02)	0.01 (0.02)	0.00 (0.03)	0.00 (0.03)	−0.06 (0.04)	−0.03 (0.05)	0.05 (0.04)	0.06 (0.04)
* Round 5	0.02 (0.02)	0.03 (0.02)	−0.04 (0.02)	−0.04 (0.02)	0.03 (0.02)	0.03 (0.02)			0.04 (0.04)	0.03 (0.04)	0.09*** (0.03)	0.08*** (0.02)
* Round 6	0.03 (0.02)	0.03 (0.02)	−0.04* (0.02)	−0.04 (0.02)	0.01 (0.02)	0.02 (0.02)	0.01 (0.03)	0.01 (0.03)	0.01 (0.04)	0.01 (0.04)	0.03 (0.03)	0.04 (0.03)
* Round 7	0.04* (0.02)	0.04* (0.02)	−0.01 (0.02)	−0.01 (0.02)	0.01 (0.02)	0.01 (0.02)	0.07** (0.02)	0.07** (0.02)	0.09 (0.05)	0.09 (0.05)	0.04 (0.03)	0.04 (0.03)
* Round 8	0.04* (0.02)	0.04* (0.02)	−0.05 (0.03)	−0.05 (0.03)	0.02 (0.02)	0.02 (0.02)	0.06* (0.03)	0.06* (0.03)	0.06 (0.05)	0.06 (0.05)	0.02 (0.03)	0.02 (0.03)
* Round 9	0.06*** (0.02)	0.05** (0.02)	0.01 (0.03)	0.01 (0.03)	0.04 (0.02)	0.03 (0.02)	0.08** (0.03)	0.08** (0.03)	0.13** (0.05)	0.12* (0.05)	0.06* (0.03)	0.06* (0.03)
* Round 10	0.10*** (0.02)	0.09*** (0.02)	0.06* (0.03)	0.05 (0.03)	0.07** (0.03)	0.06* (0.03)	0.13*** (0.03)	0.13*** (0.03)	0.22*** (0.04)	0.21*** (0.04)	0.06 (0.03)	0.04 (0.03)
Age
*16–25	REF		REF		REF		REF		REF		REF
*26–35	0.01 (0.01)	−0.02 (0.01)	0.01 (0.02)	−0.02 (0.02)	−0.01 (0.01)	−0.04** (0.01)	0.01 (0.02)	0.01 (0.02)	−0.03 (0.02)	−0.05** (0.02)	0.02 (0.01)	−0.02 (0.01)
*36–64	0.01 (0.01)	−0.01 (0.01)	−0.01 (0.02)	−0.04 (0.02)	−0.01 (0.02)	−0.04* (0.02)	0.00 (0.03)	0.00 (0.03)	−0.04 (0.02)	−0.06* (0.02)	0.02 (0.02)	−0.01 (0.02)
*65+	0.01 (0.02)	−0.01 (0.02)	0.00 (0.02)	−0.02 (0.02)	−0.03 (0.02)	−0.05* (0.02)	0.01 (0.03)	0.01 (0.03)	−0.04 (0.03)	−0.06* (0.03)	−0.01 (0.02)	−0.04* (0.02)
Covariates
*Education		0.07*** (0.01)		0.06*** (0.01)		0.06*** (0.01)		−0.02 (0.01)		0.05*** (0.01)		0.08*** (0.01)
*Religion		−0.08*** (0.01)		−0.01 (0.01)		−0.05*** (0.01)		−0.03** (0.01)		−0.03*** (0.01)		−0.09*** (0.01)
*Social class		0.06*** (0.01)		0.09*** (0.01)		0.05*** (0.01)		−0.01 (0.01)		0.06*** (0.01)		0.08*** (0.01)
Num.Obs.	160,480	160,480	160,206	160,206	148,762	148,762	124,794	124,794	156,023	156,023	157,002	157,002
R^2	0.124	0.139	0.138	0.158	0.082	0.097	0.205	0.211	0.190	0.197	0.379	0.385
R^2 Adj.	0.124	0.138	0.138	0.157	0.082	0.096	0.204	0.210	0.190	0.196	0.379	0.385

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. The entries represent OLS estimates, with robust standard errors reported in parentheses, clustered by country-period dyad. The entries are interaction effects between individual voter attitudes on an issue in the ESS and other variables on the average party position they voted for during the last national election on that same issue (CHES). Main APC-effects and country*attitude interaction effects have been omitted from the table. Note that the main effect for EU-issue is not statistically significant, which depends on the selection of the reference category (here: Germany).

When examining the significance of interactions between issue attitudes and rounds, we find increased issue alignment in the cases of income redistribution, the EU-issue, and immigration when taking age, cohort, and other demographic covariates into account. In contrast, while the social liberty and environment models do show some statistically significant increases in the most recent ESS-round, it is hard to speak of a real trend. Although the remaining period coefficients are generally positive throughout the remaining periods, most fail to reach statistical significance as reflected by the wide confidence intervals in Figure 1.

To further validate these findings, we re-estimated these models by substituting the period dummies with a linear (trend) variable. While the linear period coefficient in these models is consistently positive, it reaffirms that issue alignment trends are statistically insignificant in the left–right models. Conversely, these models reveal a statistically significant increases in alignment on the environment and social liberty at p < 0.01 and p < 0.05, respectively, although these effects are weak. In line with the models presented in Table 2, the trend towards stronger alignment over time remains statistically significant (p < 0.001) for the issues of income redistribution, the EU, and immigration. The increase in issue alignment on a ‘traditional issue’ (income redistribution) is noteworthy, since we did not hypothesise this effect. So, in sum, we find mixed evidence for hypothesis 1.

Generational effects

Based on the formative years model, we expect stronger issue alignment on emerging issues among younger generations compared to older generations (H2a). Vice versa, we expect older generations to exhibit higher levels of issue alignment than younger generations on traditional issues, in this case, income redistribution (H2b). Figure 2 plots the associated cohort effects.

Figure 2. Generational effects of issue alignment. The figure shows the interaction between a voter’s attitudes and the cohort which they are part of on their vote choice across six issues. The results are presented with and without demographic covariates (education and religion) and are based on pooled fixed-effects regression models presented in Table 2. Data: ESS (voter attitudes and vote choice) and CHES (mean expert party placement on an issue).

These models support hypothesis 2a, revealing significant generational increases in issue alignment across all four models that focus on new issues (social liberty, the environment, the EU, and immigration).⁸

Notably, the immigration model yields especially strong cohort effects, seeing each successive generation significantly more aligned than the previous. This pattern holds across model specifications employing alternative measures of voter and party positions on immigration, as shown in Table 8 presented in Online Appendix D. The social liberty, environment, and EU models show similar patterns, albeit slightly less pronounced. As the social liberty model encompasses a broader moral cleavage that continues to divide contemporary party lines, it makes sense that each generation enters the electorate more aligned on this issue than the preceding cohort. Similarly, the generational increase in alignment on the environment seems to have started with the onset of the 80s generation, who grew up when the first green parties started to gain electoral support. However, this effect is weaker once we control for demographic shifts in the electorate, suggesting that generational alignment on this issue is partially the result of educational expansion and secularisation. Additionally, we observe the emergence of generational issue alignment on the EU with the entry of the 80s and 90s generations. Growing up during a period when European cooperation gained momentum, these generations exhibit increased generational alignment in line with our theoretical expectation, although to a lesser extent than initially expected. We will explore this further in our country-comparison.

We also predicted older cohorts to exhibit stronger issue alignment on traditional issues, like income redistribution, compared to younger generations (H2b). Our models reject this hypothesis. All cohorts align on redistributive policy considerations to a similar degree. If anything, Millennials are slightly more aligned on income redistribution than older generations in some robustness checks, but only when we do not control for education, religion, and social class (see Tables 5 and 9 in Online Appendix D). Nevertheless, the emergence of cultural issues has not prevented new generations from voting for parties close to themselves to a similar extent as earlier generations on this ‘traditional’ economic issue.

We also examine whether there are generational differences in alignment on the left–right dimension. We find that younger generations are more aligned than older generations on left–right. We interpret left–right as an ideological dimension (also called a ‘super-issue’) that encompasses both traditional and newly emerging issues. Our results show that younger voters are more aligned on recently politicised issues than older generations (H2a supported), yet they are just as strongly aligned as older generations on the more traditional issues (H2b rejected). So, if left–right does indeed encompass all issues, old and new, and if younger generations are more aligned on some of these issues, it should not be surprising that they show significantly more alignment on the broader left–right spectrum. This finding is robust under all model specifications, including our robustness checks. Since left–right placement represents more stable value orientations, the generational increase in alignment on this dimension indicates that generational issue alignment extends beyond specific policies.

Life cycle effects

Finally, although we did not formulate hypotheses about ageing, we need to control for possible life cycle effects affecting issue alignment. Table 2 reveals minimal, if any, uniform age effects. While some of the individual coefficients achieve statistical significance, we only observe consistent ageing patterns when including demographic covariates. Specifically, voters significantly de-align on the environment and immigration as they grow older when controlling for education, religion, and social class.⁹

All other models do not reveal any consistent ageing effects. While we cannot offer explanations for the differences between these models, they underline the need to account for life cycle effects to accurately estimate period and cohort effects.

Country comparison

Although our pooled analysis offers insights into generational differences in issue alignment, it also begs the question whether the period trends and generational patterns are specific to particular party systems or regions. To explore this possibility, we estimated our models with three-way interactions between attitude, country, and cohort or period. These models show several statistically significant three-way interactions, even when applying Bonferroni corrections to account for Type-1 errors. To examine these differences in more detail, we conducted our main analysis for each party system while clustering the robust standard errors by period (ESS round), resulting in a total of 102 regression models.¹⁰ However, as our pooled models include dummies for all ESS rounds to estimate age and cohort effects, running the same models by country would result in too many period coefficients to compare systematically across all 17 party systems. To maintain the interpretability of issue alignment trends by country, we employed a linear period specification while controlling for age and generations. Figure 3 presents the results.

Figure 3. Period effects in issue alignment and dealignment based on 102 individual APC-models. Entries are regression terms capturing linear trends in issue alignment for each model and country with 90% confidence intervals. A positive estimate indicates that voters of all ages and generations align their vote choice more strongly with their issue attitudes, and vice versa.

In the cases of income redistribution and the EU, there seems to be a rather consistent pattern of increasing issue alignment with no signs of dealignment. The social liberty and environment models display similar trends, albeit less pronounced. We also see one case of significant dealignment on the social liberty issue (Wallonia). By comparison, larger country differences emerge when analysing immigration and the left–right dimension. Countries like Austria and Belgium, where immigration was already politicised in the twentieth century, show no significant or meaningful issue alignment trends. In contrast, several other countries where immigration became a salient issue much more recently, like in Sweden, exhibit stronger signs of alignment. Conversely, Norway and Switzerland demonstrate significant dealignment on immigration. When focusing on left–right, we see clear indications of growing alignment in some party systems, while others experienced dealignment.

Nevertheless, overall, across six issues and 17 party systems (102 parameter estimates), we observe only seven cases of significant dealignment and 40 cases of increasing alignment. So, the general trend clearly points towards greater issue alignment. However, in contrast to our expectations (formulated in H1), we do not find strong evidence of increased issue alignment on recently politicised socio-cultural issues across the board, i.e. across countries, generations, and age groups, except on immigration and the EU. Our country-comparison also reveals a relatively widespread increase in alignment on the ‘traditional’ economic redistribution issue, which was not part of our hypothesis.

We now turn our attention to country-level differences in generational differences to determine whether the general patterns shown in our pooled analysis also hold in a country comparison. Figure 4 shows differences in issue alignment between the youngest (the Millennials) and the oldest (Post-WWII) generation in each model.¹¹ We find support for H2a in some countries, where the Millennials are more aligned on social liberty, the environment, the EU, and immigration. In no country are Millennials significantly less aligned than the oldest generation on recently politicised issues, but in many cases the differences are not statistically significant. Consistent with our pooled results, generational differences are most pronounced on immigration (10 out of 17 party systems). By contrast, we find less evidence of generational issue alignment on the EU-issue (3 party systems). van der Brug and Rekker (2021) found that the most recent generation is most strongly aligned on European unification in the Netherlands, which we replicate in our analysis at p < 0.05. Even though we are hesitant to make claims about issue alignment within a particular party system because we work with limited data by country, the fact that we find a significant generational pattern in only three cases on this issue (Wallonia, Finland, and the Netherlands) indicates that this pattern may not exist across Western Europe at large.

Figure 4. Generational differences in issue alignment and dealignment based on 102 individual APC-models. Entries are regression terms capturing the difference between Millennials and the Post-WWII generation in issue alignment (the interaction between an individuals’ issue attitudes and the position of the party they voted for during the last national election on said issue) with 90% confidence intervals. A positive estimate indicates that Millennials align their vote choice more strongly with their issue attitudes than the Post-WWII generation, and vice versa.

Overall, we find little evidence generational issue dealignment, as only the German post-WWII generation votes more in line with their attitudes on income redistribution than the Millennials. Given evidence of significant generational alignment on income redistribution in five other party systems, we still firmly reject the hypothesis that older generations align more strongly on income redistribution (H2b). The remaining models only show two cases of significant generational dealignment, namely, in Sweden on the EU-issue, and in Italy on left–right. Beyond the Swedish outlier on the EU, however, all significant generational differences follow the hypothesised issue alignment patterns (H2a).

In sum, while these models do not allow us to make firm statements about cohort effects within countries because of data limitations, we find the overall generational patterns to be highly similar across these 17 political systems.

Conclusion

Do we observe patterns of issue (re)alignment or dealignment across Western Europe, and, if so, does generational replacement contribute to these processes? Our APC analysis of issue alignment provides the following takeaways that underline the need to consider age, period, and cohort effects for understanding the evolution of party–voter ties in Western Europe.

Starting with period effects, Western European electorates have become more strongly aligned on several key political issues over the past two decades. Specifically, this increase is significant on two recently politicised socio-cultural issues (immigration and the EU), and a traditional issue (income redistribution). However, we did not observe strong evidence for a consistent alignment trend on the environment, social liberty, and the left–right dimension. Our country comparison reveals that period effects matter but are not uniform. Particularly in the case of left–right, we find patterns of increasing alignment in some countries and dealignment in others. Similarly, in the context of immigration we identify substantial country differences, even though increasing alignment appears to be the trend in most cases. Such variations may be explained by differences in the timing of issue politicisation across Western Europe.

Generational replacement significantly contributes to issue alignment across most issues we examine. While the rise in generational issue alignment can be partially attributed to secularisation, class differences and educational expansion, socialisation experiences leave a lasting mark on vote choice later in life. Compared to period effects, the generational differences in issue alignment appear more similar across Western Europe. We find virtually no evidence of generational issue dealignment. This relative uniformity in generational issue alignment suggests a commonality in bottom-up socialisation processes across countries, like growing up during times of globalisation. In contrast, the variability in top-down processes such as the substance of party competition, may explain why significant generational issue alignment does not consistently hold in a country-comparison. Particularly the EU-issue has been politicised at different times across different political systems and in some countries, it was hardly politicised in most of the years that we study. This could well explain the lack of generational effects in these countries.

Whereas period effects and generational differences all point in the direction of increasing issue alignment, the age differences provide some (but very limited) evidence of issue dealignment as people grow older. The most notable life cycle effect concerns the slight weakening of party–voter ties on the environment and immigration with age once we control for religion, education, and social class. Even though age plays a secondary role in issue alignment compared to period and cohort effects, this life cycle pattern deserves further investigation.

While our analysis provides key insights to understand the evolution of party–voter ties across Western Europe, it also has its limitations. First, although we find evidence of issue alignment trends, it is important to keep in mind that our analysis covers data from the previous two decades when recently politicised socio-cultural issues have, to some extent, already shaped West European politics (e.g. Dennison and Geddes 2019; Kriesi et al. 2008). Second, the cross-national differences presented here deserve scholarly attention, but discussing them in detail would be beyond the scope of this article. Future studies should examine whether the salience and the timing of the politicisation of an issue during the formative years explains how important these issues are for a generation later in life. Other contextual variables, like the entry of certain types of parties, the legacies of undemocratic regimes, economic performance, and the (in)stability of party systems may also shape the socialisation experience of future voters. Third, our analysis cannot jointly examine the independent effects of multiple issues. Nor can we directly measure personal issue salience because of data limitations in the ESS, which is a promising avenue for future research, especially when focusing on cross-pressured voters. Finally, our analysis cannot disentangle whether new generations of voters have changed their political orientations to align more closely with their party preferences (a supply-side explanation), or, whether (new) parties move towards pre-existing beliefs in the electorate (a demand-side explanation).

Bearing these limitations in mind, the increase in issue alignment holds important implications for the future of Western European politics. From a demographic representation perspective, our study paints an optimistic picture as the party preferences of new generations align more strongly with their substantive policy preferences than those of older generations. In other words, younger generations are more likely than older ones to agree on important issues with the party they support. It remains an open question whether issue alignment will lead to more stable party systems. Conceptually, alignment implies stability between parties and voters, whereas the increase in issue alignment we observe coincides with increasing levels of electoral volatility throughout Western Europe (e.g. Emanuele et al. 2020). These combinations of findings are consistent with the idea that many voters nowadays are not so much loyal to specific political parties but rather to the set of ideas that they represent (e.g. Dalton et al. 2000: 60). In every election, issue-aligned voters can re-evaluate their vote choice in terms of what issues are most salient (Costello et al. 2021). To the extent that issue alignment provides electoral stability, this may, therefore, be a stability at the level of ideological blocks within which voters can more easily switch between ideologically proximate parties (e.g. Rekker and Rosema 2019; van der Meer et al. 2015).

Taken together, our findings clearly reveal that people’s issue preferences increasingly align with the policies of the parties they vote for in Western European democracies, regardless of whether issues are new or old. Generational replacement plays a key role in this process, as new generations tend to vote more in line with their attitudes on new issues and to a similar extent on traditional issues. This pattern cannot solely be attributed to demographic changes but reflects their distinct formative experiences. Future research should focus on uncovering the mechanisms driving these patterns and the role of contextual variables in shaping the political socialisation of voters, as they will shape the political landscape in Western Europe for generations to come.

Acknowledgements

Earlier versions of this paper were presented at the Political Behaviour Colloquium at the European University Institute (11 April 2023), the ECPR Joint Sessions in Toulouse (25–28 April 2023), and at the CORE lectures at Bergen University (28 September 2023). We would like to thank all the participants during those meetings, and Sylvia Kritzinger, Sarah de Lange, Tom van der Meer, Reto Mitteregger, and Chiao Li in particular for their valuable feedback. In addition, we thank the anonymous reviewers at West European Politics for their helpful suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Nederlandse Organisatie voor Wetenschappelijk Onderzoek under Grant [406.18.RB.025] and Veni Grant Number [VI.Veni.191R.039].

Notes on contributors

Thomas Jocker

Thomas Jocker is a PhD candidate at the University of Amsterdam. His dissertation focuses on generational differences in determinants of party choice. [[email protected]]

Wouter van der Brug

Wouter van der Brug is Professor of Political Science, University of Amsterdam. His research interests focus on collective political behaviour, in particular electoral behaviour, right-wing populism and political parties. [[email protected]]

Roderik Rekker

Roderik Rekker is an Assistant Professor of Political Science at Radboud University, Nijmegen. His research focuses on stability and change in political attitudes across time, generations, and the lifespan, as well as on political polarisation over facts and science. [[email protected]]

Notes

1 Namely: Austria, Belgium (Flanders and Wallonia separate), Denmark, Finland, France, Germany, Greece, Ireland, The Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and The United Kingdom. We do not include respondents from Eastern Europe due to their substantially different formative experiences compared to those growing up in Western Europe. We also omit respondents from Northern Ireland, Cyprus, and Iceland because of data limitations in the CHES. See Table 4 in Online Appendix B for an overview of the countries included in each ESS round of the analysis.

2 We use the Party Facts database (Döring and Regel 2019) to link vote choice in the ESS with the associated party positions in the CHES. We utilize Sophie Hill’s updated codification of parties in the ESS, which we further update to include the newest ESS Round (2020, see: https://github.com/sophieehill/ess-partyfacts-crosswalk).

3 A cursory glance at a party’s current and subsequent issue position yields correlations from .91 on social liberty to .97 on the left–right dimension, indicating high stability. We also explored the absolute change in positions over time (see also Dalton and McAllister 2014). On the six issue scales, the aggregate median changes are small, ranging between 0.3 and 0.6 on a 0–10 point scale. As an additional robustness check to account for the fact that most questions related to issue positions were introduced in the CHES in 2006, we ran our models excluding observations from the first two ESS rounds. These models are presented in Online Appendix D (Table 9) and produce similar age and cohort estimates.

4 For linguistic reasons we employ the terms issues and issue alignment, even though the policy dimensions to which they pertain are often broader than a single issue. This is especially the case for left–right, even though it has been referred to as a ‘super-issue’.

5 Unfortunately, this question was not asked in the (self-completion) surveys conducted during the tenth ESS round in Austria, Germany, Sweden, and Spain.

6 We do not study other issues (e.g. GAL-TAN and gender) because of data limitations in the ESS.

7 We lack the respondents to include a separate Gen-Z cohort (those born after 1996). Expanding the Millennial category to include these respondents did not yield substantively different cohort effects.

8 Although not the focus of our study, the effects of these variables deserve mention. As expected, higher education levels and social class generally predict higher issue alignment. Interestingly, these effects are negative, although not statistically significant, for the EU-issue. Religion has a significant negative effect across all models. These factors partially account for generational differences, given that younger generations tend to be higher educated and less religious, whereas older generations can be expected to vote more in line with their class membership. However, political socialization experiences also independently influence issue alignment, with their effect sizes comparable to education in most of our models.

9 Our robustness check incorporating alternative measures for immigration attitudes does not provide evidence of age-related issue dealignment, as shown in Table 8 in Online Appendix D.

10 These models do not include demographic covariates to avoid overfitting.

11 These models control for period effects by employing dummies for every ESS round.