The shifting issue content of left–right identification: cohort differences in Western Europe

Abstract Over the last decades, the rising salience of new issues has transformed politics in Western Europe. How has this transformation affected citizens’ left–right identities? This study shows how new issues have become integrated into the meaning of left–right among Western Europeans through generational replacement. Mechanisms of political socialisation and elite cue taking imply that how strongly an issue is associated with left–right identities should reflect this issue’s politicisation during a cohort’s formative years. In line with this, an analysis of the European Social Survey for 12 Western European countries from 2002 to 2018 reveals that environmental protection and immigration attitudes are more strongly associated with left–right positions among those born later. Attitudes towards redistribution tend to be less relevant within more recent cohorts in some countries, though period effects after the European debt crisis point in the opposite direction. This study enhances our understanding of the evolution of political conflict.

The terms 'left' and 'right' continue to be widely used by politicians, journalists and political scientists to describe politics.Likewise, the overwhelming majority of citizens in Western European democracies have a sense of where they stand on the left-right scale and these left-right identifications shape their political outlook and party preferences (Jou and Dalton 2017;Knutsen 1995Knutsen , 1997;;Mair 2008;Otjes and Rekker 2021).The relevance of left-right identification, as one of citizens' most basic orientations towards politics, is reflected in the omnipresence of left-right self-placement as an independent variable in empirical studies of voting behaviour.
What are left-right identifications about?In post-World-War-II Western Europe, 'left' and 'right' have long been primarily understood in economic terms, dividing supporters of redistribution and state intervention in the economy and advocates of a free market.Yet, Western European politics has transformed over the last decades.New issues, such as the environment and immigration, have become important.New parties, chiefly new-left Green and radical right parties, have emerged.Overall, party competition has transformed with new 'cultural' issues increasingly structuring party competition and the social divisions underlying party choices (e.g.Gethin et al. 2021;Kriesi et al. 2008;Oesch and Rennwald 2018).
This article asks how these important developments have affected the meaning of 'left' and 'right' among citizens.Drawing on theories of political socialisation, I argue that the issue content of left-right identification -that is, which issues are associated with left-right self-placement -differs across generational cohorts.I suggest that cohort differences follow from the combination of two mechanisms: a party cueing mechanism and a political socialisation mechanism.The party cueing mechanism holds that issue salience and polarisation (i.e.politicisation) at the party-system level shape how issue positions relate to left-right identification at the citizen level.The political socialisation mechanism builds on findings that political belief systems, especially basic political orientations like left-right identification, are to a large extent formed during adolescence and remain 'sticky' thereafter.Taken together, which issues dominate political conflict between left-wing and right-wing parties during their adolescence is likely to leave a lasting footprint on citizens.In contemporary Western Europe, we should thus see that 'new' issues like immigration or environmental protection are more strongly associated with left-right positions within more recent cohorts, perhaps to some extent crowding out 'old' issues such as redistribution.
My empirical analysis draws on nine waves of the European Social Survey (ESS) for 12 Western European countries from 2002 to 2018.I find that, overall, caring for the environment and immigration attitudes are stronger predictors of left-right positions among those born later.In contrast, attitudes towards redistribution tend to be less relevant within recent cohorts.Analysing countries separately, I find that these overall cohort trends are not uniform, and that variation around those seems to reflect differences in party competition and issue politicisation -in line with the proposed party cueing mechanism.This article's findings add to a large body of work that has been interested in the twin question which issues citizens associate with 'left' and 'right' and how this differs across contexts (e.g.Bauer et al. 2017;Dalton 2018;Dalton et al. 2011;Freire 2008Freire , 2015;;Fuchs and Klingemann 1990;Huber 1989;Inglehart 1990;Inglehart and Klingemann 1976;Kitschelt and Hellemans 1990;Knutsen 1995Knutsen , 1997;;Noël and Thérien 2008;Sani and Sartori 1983).However, past research has paid rather little attention to the dynamic aspect of how new issues are absorbed into the meaning of left-right.There is a handful of mainly cross-sectional studies from the 1990s touching upon the question of how attitudes on new 'post-materialist' cultural issues relate to left-right identification (Fuchs and Klingemann 1990;Inglehart 1990: chapter 9; Kitschelt and Hellemans 1990;Knutsen 1995Knutsen , 1997)).More recently, there have been two single-country studies on the Netherlands in the age of globalisation: de Vries et al. (2013) reveal how left-right positions have become less strongly associated with attitudes towards redistribution and more strongly linked with attitudes towards immigration over time.Rekker (2016) shows that such change is partly driven by generational replacement as attitudes towards immigration matter more in more recent cohorts.However, it has been an open question whether these findings generalise beyond the Dutch case (cf.de Vries et al. 2013: 235;Rekker 2016: 128) -especially since previous studies indicate the issue content of left-right to vary across countries to begin with.
Contributing to this line of research, the present study shows that there are cohort differences in the importance of different issues in Western Europe as a whole.In line with Rekker's (2016) findings for the Dutch case, immigration is more strongly related to left-right identifications among those born later when studying Western European countries jointly and when looking at countries other than the Netherlands individually -among those Belgium, Switzerland or Germany.In addition, I find an increasing relevance of protection of the environment in later cohorts and some evidence of a crowding out of the 'old' redistribution issue.Moreover, this study is able to reveal that, while the main patterns hold in many of the individual countries, there remain strong cross-national differences in the issue content of left-right identifications and, also, in cohort differences within countries.Overall, this study supports a dynamic account of the evolution of left-right identification according to which new issues are absorbed into the issue content of left-right through generational replacement for which previous evidence has been 'still sketchy' (van der Brug and Franklin 2018: 438-9) -thus enhancing our understanding of the evolution of political conflict.
In the next section, I spell out the theoretical argument in more detail.The third section describes the data used.The fourth section presents findings from the main pooled analysis and the fifth section discusses country-specific patterns.The final section summarises the results and discusses limitations as well as implications.

The shifting issue content of left-right identification in Western Europe and the role of generational replacement
Before proceeding with my theoretical argument, a few clarifications on the multifaceted construct of left-right orientations are in order.Previous studies demonstrate that citizens' left-right orientations may carry different meaning.On a basic level, left-right orientations may be about ideological issue positions or group identification.
On the one hand, left-right orientations may reflect a broad ideological outlook on policy issues, an 'issue-based ideology' (Mason 2018).In this predominant account, the left-right scheme functions as a 'super-issue which summarises the programmes of opposing groups' (Inglehart and Klingemann 1976: 244).It bundles positions on a set of more specific policy issues.Citizens may use their left-right orientations to form attitudes on unfamiliar issues; and their attitudes on salient issues may inform their overall left-right orientation (Weber and Saris 2015).Whatever the causal direction, this account emphasises the close connection between left-right orientations and issue positions, i.e. the 'issue component' of left-right orientations (Inglehart and Klingemann 1976).
Yet, what the pertinent issues are varies across space and time (Noël and Thérien 2008;Sani and Sartori 1983).
On the other hand, left-right self-placement may also reflect a social identity, that is, an identification with elite-level actors and social groups associated with the symbolic labels 'left' and 'right' (Conover and Feldman 1981;Mason 2018).When citizens identify with parties and social groups that are associated with the 'right' ('left'), they may identify as 'right' ('left') themselves -independent of any issue content.Due to this 'identity-based ideology' (Mason 2018) left-right orientations may also contain a 'partisan component' , related to party identification and support, and a 'social component' , related to social group membership and identification (Inglehart and Klingemann 1976).Most of the time, citizens' self-placement on the left-right scale is likely informed by both issue and identity considerations.
The present contribution is interested in the 'issue component' of left-right orientations -specifically: in the relative strength of associations between left-right orientations and issue attitudes across generational cohorts in Western Europe.I do not intend to weigh in on the importance of the issue component and social and partisan components vis-àvis each other, as classic contributions on the meaning of left-right orientations have done (see Freire 2006Freire , 2008;;Huber 1989;Inglehart and Klingemann 1976;Knutsen 1997).Yet, the idea that left-right orientations are also informed by how citizens are oriented towards political elites, especially parties, is a useful starting point for the argument that the issue content of left-right orientations is fundamentally shaped by divisions at the elite level.
This 'party cueing' mechanism can be sketched as follows.Essentially, citizens get their cues about which issue positions are 'left' and 'right' from political parties (or other political elites).Those issues that primarily divide political parties considered 'left' and 'right' will come to dominate the understanding of the issue content of left-right at the citizen level.For example, when parties of the 'left' and 'right' primarily differ in their stances on economic policy, the left-right scheme will take on a largely economic meaning.Ultimately, left-right orientations of citizens will be associated with their positions on those issues that are politicised at the party-system level, i.e. issues that are both emphasised by parties (salience) and on which parties take opposing views (polarisation). 1  Both directions of the reciprocal relationship between left-right orientations and issue orientations (Rekker et al. 2017;Weber and Saris 2015) may contribute to this pattern, and these mechanisms may reinforce each other.On the one hand, citizens form left-right orientations based on issue attitudes.They will do so based on issues that are politicised at the party-system level, as it is these that define the public understanding of left-right.On the other hand, citizens' left-right orientations may inform their issue attitudes.For this to happen, issues need to be understood as connected to left-right, which hinges on politicisation at the party-system level.
Previous evidence on the correlates of left-right orientations is in line with the party cueing mechanism.First, on a basic level, studies suggest that the issue component of left-right orientations is overall stronger where parties' positions are more polarised (Freire 2008;Inglehart and Klingemann 1976).Where 'programmatic party competition' (Kitschelt 1995) is weak, left-right orientations are hardly anchored in issue positions (Ruth 2016).Second, cross-country differences in how strongly different issue attitudes are associated with left-right orientations mirror which issues dominate party competition.For example, Knutsen (1995) shows that economic issues are most closely associated with left-right orientations in the Protestant Northern European countries, whereas religious vs. secular attitudes are more relevant in the Catholic countries.He attributes these differences to historical cleavage constellations that either gave rise to an enduring dominance of the economic class cleavage or the state vs. church cleavage in party competition (Lipset and Rokkan 1967;Rovny and Polk 2019).Third, the issue content of left-right orientations seems to integrate new issues once these are taken up by political parties.As de Vries and colleagues (2013) show, once immigration became a salient and polarising issue at the party-system level, Dutch citizens' left-right orientations became increasingly related to their attitudes on immigration.Likewise, Knutsen (1995) shows how postmaterialist value orientations became increasingly related to left-right orientations after the 1970s in economically advanced Western European countries.
However, it is questionable whether all citizens respond equally to such changes in issue politicisation.The 'political socialisation mechanism' expects that the party cueing mechanism is strongest when individuals are in their politically formative years and develop their basic political orientations and identities.This reasoning builds on the 'formative years hypothesis' , which holds that political orientations tend to be formed and consolidated during adolescence and early adulthood and are persistent thereafter.There is ample research supporting the formative years hypothesis, especially when it comes to orientations that are central to individuals' political belief systems (Alwin and Krosnick 1991;Sears and Funk 1999;van der Brug and Franklin 2018).During their adolescence and early adulthood citizens develop left-right orientations in line with their basic issue attitudes -and develop basic issue attitudes in line with their left-right identities (Rekker et al. 2015(Rekker et al. , 2017)).These processes result in a consolidated political belief system that is more stable thereafter.The associations between left-right orientations and issue attitudes that emerge in the formative years of different generations due to these processes -based on patterns of party competition during those years -are then expected to persist over time, leaving lasting footprints.These mechanics should lead to a pattern in which the issue content of left-right orientations of generational cohorts reflects issue politicisation during their adolescence and young adulthood.To the extent that new issues emerge and are integrated into party competition over time, the issue content of left-right orientations should thus differ between cohorts (cf.Rekker 2016).This is the general expectation tested in this article.
To move from this general statement to specific expectations about cohort differences in the issue component of left-right orientations in contemporary Western Europe, we need to consider the changing contours of party competition in Western Europe.Such a discussion comes with the risk of oversimplification as it necessarily glosses over the nuances in such broad patterns that exist across countries.In light of this, the goal is to describe general patterns in the shifting content of party competition that allows general expectations to be derived about differences in the issue component of left-right orientations in Western Europe that I will investigate in the first and main part of the empirical analysis.In an additional, more exploratory, empirical analysis, I will delve into the nuances by analysing and discussing results for individual countries.
I suggest that we can, in line with the conventional wisdom of the pertinent literature, roughly distinguish three phases of party conflict in post-World-War-II Western Europe.The first decades after World-War-II tended to be dominated by a conflict over economic issues -chiefly about state intervention in the economy and redistribution.In cleavage terms, this era was dominated by the class cleavage between capital and workers (Lipset and Rokkan 1967).
Second, beginning in the late 1960s, new 'post-materialist' (Inglehart 1977) cultural issues gained salience.The emerging 'value cleavage' (Kriesi 1998) found expression in the rising importance of a second dimension of party competition setting 'authoritarians' against 'libertarians' (Kitschelt 1994).Questions that pitted conservative adherents of traditional morality against libertarian supporters of cultural diversity and emancipation on issues such as abortion, women's rights, or tolerance towards homosexuality are important for this divide.Related to this 'silent revolution' is the rise of environmental protection to a salient and contested political issue at the party-system level.Though, rather than a revolution this rise of the environmental issue was more gradual in nature -albeit with some ups and downs (Green-Pedersen 2019) -and continues to this day around the issue of global climate change.
Third, issues associated with globalisation gained prominence as a divisive issue for party competition from the 1990s onwards (Kriesi et al. 2008).Central among these is immigration, whose rising salience for parties of both the right and left is documented in Dancygier and Margalit (2020; also see Green-Pedersen 2019).These new globalisation-related issues added to the substance of the second, 'cultural' dimension of party competition, complementing conflicts over traditional morality vs. cultural liberalism (de Vries 2018; Hillen and Steiner 2020; Kriesi et al. 2008;Lefkofridi et al. 2014).
This discussion results in clear expectations on cohort differences regarding the 'new' issues of immigration and environmental protection.According to the political socialisation mechanism, (post-)millennialsi.e.those born from 1980 onward and reaching early adulthood around 2000 and later -should have been most strongly affected by the increasing politicisation of immigration in the late 1990s and 2000s in most Western European countries (Dancygier and Margalit 2020).Attitudes towards immigration should thus be more strongly related to left-right orientations among millennials than among previous cohorts in Western Europe -in line with what Rekker (2016) found for the Netherlands.
In the case of environmental protection, the earlier yet more gradual integration of the environment in party competition should be reflected in a gradually increasing association between environmental attitudes and left-right orientations with each successive generation.When the 'interwar cohort' (born up to 1945) reached early adulthood, i.e. before 1970, environmental protection was hardly a salient and polarising issue at the party-system level.This was increasingly different for the 'baby boomers' (born between 1946 and 1964) and their successors, i.e. 'generation X' (born between 1965 and 1979) and the millennials.
Regarding economic attitudes, a plausible scenario is 'issue crowding out ' (de Vries et al. 2013: 228): the rise of new issues might push the 'old' economic issues that once dominated left-right oppositions at the party-system level into the background.De Vries and colleagues (2013: 228) suggest that issue crowding out follows from the zero-sum character of the public issue agenda: ' As a newly salient policy issue is bundled into the left/right dimension, […] other issues have to at least partially make way' .Despite such a general tendency, there might still be variation: economic left-right conflicts may strongly decline in salience in party competition amidst the rise of other issues in some contexts but less so in others.Research based on party manifesto data indicates that the relative salience of the economy (Green-Pedersen 2019; Hillen 2022) and party polarisation around economic left-right issues (Steiner and Martin 2012) has declined in some Western European countries since the 1960s/1970s, but that the trend is not uniform and that economic issues have tended to resurge in the aftermath of the European debt crisis.Hillen (2022) reveals clear decreases in the relative salience of economic issues in countries such as Germany, the Netherlands and the UK, for example.Yet, economic salience has remained at a relatively high level in Scandinavian countries, and in Southern Europe it has even increased over time.
In line with evidence of a decreasing salience of economic issues at the party-system level in the Netherlands, both de Vries et al. (2013) and Rekker (2016) find attitudes towards redistribution to be less strongly associated with left-right orientations in more recent years (up until 2006) in the Dutch case.Yet Rekker (2016) cannot confirm the hypothesised cohort differences.Overall, this discussion suggests a general tendency for attitudes towards redistribution to be less strongly linked to left-right orientations among later cohorts.Yet this is likely not true for countries in which classic economic conflicts have remained dominant.
In case of traditional morality vs. cultural liberalism -which in this study will be operationalised via tolerance for homosexuality -there are competing considerations.On the one hand, there are grounds to expect these to matter most among baby boomers.These cultural liberalism issues entered party politics through the 'silent revolution' that Inglehart (1977) traced back to changing value orientations among this cohort.
When issues associated with globalisation added to the second dimension later, this came to some extent at the expense of the 'older' cultural liberalism issues (de Vries 2018; Kriesi et al. 2008).On the other hand, cultural liberalism issues, and rights for homosexuals specifically, remained politically controversial for a long time.For example, same-sex marriage had not been allowed in any EU country before 2001 but was then introduced in 14 of them by 2015 (Abou-Chadi and Finnigan 2019).What is more, the varying historical importance of the religious cleavage has brought about tremendous variation in the historical importance of party contestation over cultural liberalism issues (Rovny and Polk 2019).While in Catholic Southern European countries these proved highly divisive, this was much less the case in the Protestant Northern European countries.Thus, we would expect attitudes towards traditional morality vs. cultural liberalism to be much more closely related to left-right positions in, say, Spain rather than Sweden, particularly among older cohorts.By extension, the relative differences between cohorts may also differ across countries.
In sum, the theoretical argument in combination with the discussion of shifts in party competition in Western Europe leads to clear expectations on cohort differences in the issue content of left-right identities regarding two 'new' issues: immigration and the environment should be more closely associated with left-right orientations among more recent cohorts -immigration foremost among millennials and environmental protection more gradually.In contrast, the 'old' redistribution issue is likely overall less closely associated with left-right orientations among more recent cohorts, though with variation across countries.Regarding traditional morality vs. cultural liberalism, there is no clear theoretical expectation in either direction and a heterogenous picture across countries seems likely.

Data and methods
My empirical analysis includes data from those twelve Western European countries that participated in all nine waves of the ESS from 2002 to 2018.These are: Belgium, Finland, France, Germany, Ireland, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom.
As previewed above, I distinguish between four generational cohorts: those born before 1945 ('interwar cohort'), those born between 1946 and 1964 ('baby boomers'), those born between 1965 and 1979 ('generation X') and those born from 1980 onwards ('millennials').This parsimonious scheme is widely applied in public opinion research (see, e.g.Norris and Inglehart 2019).In the cumulated ESS data, it leads to roughly similar-sized cohorts with 47,524 individuals in the interwar cohort, 68,794 baby boomers, 52,982 in the generation X, and 43,094 (post-)millennials.
The main dependent variable is a standard measure of self-placement on a left-right scale ranging from 0 to 10. 2 I consider four issue items as predictors of an individual's left-right orientation.As a standard measure of attitudes on classic economic left-right issues, I include agreement with the statement that governments should reduce differences in income levels on a five-point scale ('redistribution').To capture attitudes on cultural liberalism vs. traditional morality, I utilise agreement with the statement that gays and lesbians should be free to live life as they wish on a five-point scale ('homosexuality').Attitudes towards protection of the environment are proxied for by an item from the human values module.On a six-point scale, respondents were asked to indicate how much a person described in the following way is like them: 'she/ he strongly believes that people should care for nature.Looking after the environment is important to her/him' ('environment').Attitudes towards immigration are captured by a question on whether 'many' , 'some' , 'a few' or no immigrants 'from the poorer countries outside Europe' should be allowed ('immigration'). 3To homogenise the issue scales, I recoded them to range from zero ('right-wing' position) to one ('left-wing' position), and to align the left-right scale with them, I reversed the coding of the left-right scale such that it ranges from 'right' (0) to 'left' (10).Attitudes on these issues should thus positively relate to the leftright scale, with the strength of the association differing between cohorts.
After presenting bivariate associations, I will present results from regressions that regress left-right positions on the four issue attitudes simultaneously -thereby treating the left-right scheme as an ideological 'super issue' position that is related to several issue attitudes at once.The regressions model the left-right position, lr i , of an individual i as follows: where i refers to individuals, j to the four issues, k to the twelve countries and t to the nine ESS waves.The pooled ESS data set has a multilevel structure with individuals (level 1) being nested in contexts defined by country k and wave t (level 2; n = 108, reflecting 12 countries times 9 waves), i.e. country-waves.Accordingly, I run linear multilevel models with random intercepts, α kt , for country-waves, in addition to the individual-level error term,  i .The model also contains a set of fixed effects, µ k , at the country level k, which control for differences in left-right positions across countries.Gender and age groups, distinguishing between young (up to 35), middle (36 to 59) and old (60 and older), are included as simple control variables, X i .
Of prime interest in Eq. ( 1) is the interaction between each of the four j issue variables and the cohort an individual belongs to.In interpreting the results, I will focus on the conditional marginal effects of the issue variables for different values of cohort.
In order to obtain reliable measures of these cohort-specific issue effects, I include control variables Z i that are also interacted with the j issue variables.In the baseline model, two variables are included in Z i .The first is a categorical variable for education distinguishing between less than lower secondary, completed lower secondary, completed upper secondary and completed tertiary education. 4The second is a categorical variable for ESS waves that capture period effects, i.e. changing associations between issue attitudes and left-right attitudes over time among all cohorts.Like cohort differences, such period effects may also arise from party cueing (cf.de Vries et al. 2013) and/or be a reaction to external events that drive up the salience of specific issues.In this article, I focus on the cohort dimension, not least because the short time period covered (2002 to 2018) allows for only limited insight into longer-term period effects.Nonetheless, accounting for period effects is not only important to isolate genuine cohort effects but also enables interesting substantive insights, as will be discussed below.In a second model, Z i includes the categorical age group variable.This is done to ensure that initial differences between cohorts reflect cohort rather than age effects, following standard practice of age-period-cohort analysis of repeated cross-sectional data (Neundorf and Niemi 2014;Yang and Land 2013). 5

Cohort differences in the issue content of left-right identification in Western Europe
Figure 1 provides a first descriptive look at the bivariate associations of interest.It shows a set of scatterplots relating the four issue attitudes to left-right positions, separately for the four generational cohorts.In line with expectations, attitudes towards immigration and the environment are more strongly correlated with left-right orientations among millennials as compared to previous cohorts.In contrast, attitudes towards redistribution as well as tolerance of homosexuality correlate similarly with left-right self-placement within all four cohorts. 6These bivariate results provide initial evidence that immigration and environmental protection matter more for left-right identification among those born later.Next, these relations are studied in a regression framework that allows the inclusion of the four issue attitudes simultaneously, alongside potential confounders.
Figure 2 shows the conditional effects of the four issue attitudes across cohorts based on the baseline regression model described above.A full table with all regression coefficients can be found in the online appendices (see model 1 of Table B1).The cohort-issue interaction terms in the regression table indicate whether the issue effects are statistically different from those in the baseline (interwar) cohort.In Figure 2, statistical significance of differences (with p < 0.05) is approximated for by the non-overlap of the thick 85% confidence intervals.The estimates in Figure 2 show how the predicted left-right positions change when moving from a 'right-wing' to a 'left-wing' position on the issue, with cohort set to different values.
Figure 2 shows that the 'new' issues of immigration and environmental protection are more closely associated with left-right positions within more recent cohorts.In case of immigration, the difference is mainly between millennials and the three previous cohorts.This pattern corresponds to the theoretical argument: immigration became much more salient for party competition at the turn of the millennium.Millennials' political belief systems are most affected by this development, in accordance with the suggested party cueing and political socialisation mechanisms.
Importance attached to caring for the environment is gradually more associated with left-right positions with each successive cohort.This is in line with the environment's steadier rise to a salient and divisive issue at the party-system level.
Attitudes towards redistribution are less strongly associated with leftright orientations within generation X as well as among millennials compared to the interwar cohort and the boomers.The decreasing relevance of redistribution corresponds to the idea of 'issue crowding out' note: conditional effects with 95% and 85% confidence intervals of issue attitudes on left-right positions based on multilevel regressions with individuals (n = 180,142) nested in country-waves (n = 108).the model also controls for interactions between issue attitudes and ess waves, and issue attitudes and education.see the full regression results in table B1 of the online appendices (model 1).
(de Vries et al. 2013): the increased associations of immigration and environmental protection with left-right orientations among generation X and especially millennials go along with a decreased association with the 'old' redistribution issue.The result is also consistent with the observation of lower party polarisation around and salience of economic issues in a subset of Western European countries in the 1990s and 2000s compared to the 1960s and 1970s (Hillen 2022;Steiner and Martin 2012).
For attitudes towards homosexuals, the results do not indicate any difference in their relation to left-right orientations across cohorts.
The general pattern regarding the relevance of the four issues vis-à-vis each other is that attitudes towards redistribution are most strongly associated with left-right orientations, followed by immigration, homosexuality, and the environment.However, it is striking how different these patterns are for different cohorts.Within the interwar cohort, redistribution dominates clearly.Among millennials, redistribution and immigration attitudes are on a par.While redistribution no longer dominates within the most recent generational cohort, it is also important to keep in perspective that redistribution remains rather strongly associated with left-right positions even among millennials and that its conditional effect is only modestly smaller than for previous cohorts.That redistribution and immigration matter almost equally among millennials is much more a result of the increasing relevance of immigration.
Recall that the conditional effects in Figure 2 are from models that also include interactions between the four issue items with ESS wave and with education.In the next step, I present results from a model that additionally includes interactions between the four issue items and age groups.To fully evaluate the model, Figure 3 presents conditional effects by cohort, by age group and by ESS wave. 7 The cohort effects look similar to the results in Figure 2, especially the point estimates, suggesting that the variation across cohorts observed in the baseline model is not driven by life-cycle effects.The main difference is that the conditional effects are less precisely estimated, reflecting the collinearity between cohorts and age groups.Importantly, the differences of interest remain statistically significant -albeit for the environment the hypothesis of equal effects for the interwar cohort vs. millennials can be rejected only at the p < 0.10 level.
Looking at effects by age group next, the results show little evidence for the association between the issue positions and left-right identification varying with age.The only tentative difference is a marginally stronger association between redistribution attitudes and left-right orientation among the old.These results help to rule out that life-cycle effects play a major role for the issue content of left-right.As a result, while the predicted values by cohort become less precise when age is included as an interactive confounder, they are qualitatively similar whether we control for life-cycle effects or not.
In contrast, there is clear evidence of associations between left-right orientations and issue attitudes to change over time, i.e. of period effects.Two patterns stand out.First, redistribution has gained in relevance beginning with the 2012 wave, in parallel to the European debt crisis and the rising salience of economic issues in party communication it triggered in many countries (Green-Pedersen 2019; Hillen 2022) -as speculated in de Vries et al. (2013: 228).Second, immigration has jumped in importance in the 2016 wave, plausibly in response to its rising salience in the aftermath of the European refugee crisis of 2015.These results suggest that the changing issue content of left-right orientations is both a cohort and a period phenomenon.Like the cohort differences, period differences seem linked to the salience of issues in party competition (cf.de Vries et al. 2013), as triggered here by external events, in line with the party cueing argument.
It is interesting that period and cohort effects pull in different directions in case of redistribution.This is not a direct contradiction, however, as the cohort differences are linked to longer-term developments and the period effects to short-term developments.Still, the countervailing period effects seem to indicate that redistribution's relevance for left-right identification could potentially rebound.
Returning to the cohort effects, it is crucial to consider how they hold up in alternative model specifications.In section C of the online note: conditional effects with 95% and 85% confidence intervals of issue attitudes on left-right positions based on multilevel regressions with individuals (n = 180,142) nested in country-waves (n = 108).in addition, to the three interactions displayed, the model also controls for interactions between issue attitudes and education.see the full regression results in table B1 of the online appendices (model 2).
appendices, I present results from such robustness checks.First, I relied on alternative, more fine-grained cohort classifications, using either tenor five-year intervals.Second, I used alternative codings of age groups.These models lead to similar conclusions.Third, I estimated models with the issue items included in separate models.The main models build on the 'super-issue' understanding of left-right according to which left-right orientations are simultaneously related to several issues.Yet, the conclusions do not hinge on this assumption -the findings are similar with separate models per issue.The main difference is that the coefficients of the issue variables are all a bit bigger, while the differences between cohorts are largely the same.
Fourth, I estimated models excluding all those who indicated to have voted for either a Green or a radical right party in the last national election.I thereby address a rivalling interpretation of cohort differences: rather than reflecting socialisation effects it could be that left-right orientations differ across cohorts because there are different parties that appeal to those cohorts.For example, self-categorized left-wing voters from older cohorts may overwhelmingly vote for social-democratic parties, whereas self-categorized left-wing voters from more recent cohorts may increasingly turn to 'new left' Green parties.Because these parties differ in the issues that they stress, voters get different cues about what it means to be left from their preferred parties.Yet it is also plausible that self-categorized left-wing voters from different cohorts may vote for different parties of the left that stress different issues because their left-wing political identification is tied to different issues (cf.van der Brug and Rekker 2021).It is thus plausible that these are reinforcing processes.In any case, if we exclude Green and radical-right voters, cohort effects remain similar.Thus, the cohort differences are not just driven by those who support parties of these two new party families for whom environmental protection and immigration, respectively, are especially salient.The results are also similar when we include only voters of the 'old' division between social-democratic parties, on the one hand, and conservative as well as Christian-democratic parties, on the other -with one exception: the evidence for a dwindling association with redistribution becomes notably weaker.
Overall, the results confirm the two expectations on the increasing relevance of new issues for left-right identification among those born later: immigration attitudes are more strongly associated with left-right orientations among millennials and environmental attitudes are gradually more linked to left-right orientations with each cohort succeeding the interwar generation.In contrast, attitudes towards redistribution are less strongly associated with left-right positions among members of generation X and millennials.These three results are in line with the general idea that cohort differences in the relevance of different issues for left-right identification reflect broad historical shifts in the salience of issues for party competition in Western Europe.Attitudes towards homosexuals are similarly relevant across cohorts, at least when studying Western Europe as a whole.

Cohort differences in the issue content of left-right orientations across countries
Whereas the previous section focussed on the big picture for Western Europe collectively, this section discusses results from a more exploratory analysis of the patterns in individual countries.From previous research, we know that the issue content of left-right orientations strongly varies across countries (Knutsen 1995), and while there are broad trends in issue politicisation in Western Europe, these are certainly not uniform across countries.We would thus not only expect that the issue content of left-right identification differs across countries in general, but also that relative differences across cohorts are not the same everywhere.
Figure 4 presents empirical evidence on cross-country variability in the form of conditional effects from regressions by country.The specification is similar to Eq. ( 1), interacting issue positions with cohorts, ESS waves and education, though random intercepts for country-waves as well as country fixed effects are obviously not included.I present results from models without the additional age-issue interaction in the main text and results from models with the age-issue interaction in the online appendices (Figure D2).The results are, again, similar, though confidence intervals get larger with the age effect included.
Figure 4 reveals strong cross-country differences in how relevant the different issues are for left-right identification overall.These results are in line with previous research (Knutsen 1995) and reflect cross-national variation in historical cleavage constellations (Lipset and Rokkan 1967;Rovny and Polk 2019).Mirroring the historical and enduring dominance of economic issues in the Protestant Nordic countries, the three Scandinavian countries stand out by their dominance of attitudes towards redistribution.In Sweden, for example, the predicted difference between someone who 'agrees strongly' and someone who 'disagrees strongly' that governments should equalise differences in income levels amounts to roughly a third of the full left-right scale for all cohorts.In contrast, in the Southern-Catholic countries (Spain and Portugal) -where conflicts over traditional morality vs. cultural liberalism were historically important for party competition -attitudes towards homosexuals tend to be overall at least as relevant as attitudes towards redistribution.
Turning to the cohort differences next, Figure 4 reveals a nuanced picture with the three main patterns from above being pronounced in some of the countries, while not holding in others.At least part of this variation can be linked to differences in party competition and issue politicisation, in line with the party cueing argument.note: conditional effects with 95% and 85% confidence intervals of issue attitudes on left-right positions based on regressions by country.the models also control for interactions between issue attitudes and ess waves, and issue attitudes and education.conditional effects by ess wave are shown in Figure D1 of the online appendices.
First, attitudes towards immigration are more relevant among millennials than their predecessors in five of the twelve countries -Netherlands, Belgium, Germany, Switzerland and Sweden.Thus, the increased relevance of immigration among younger cohorts found in Rekker's (2016) study of the Netherlands extends to these other countries, especially to Germany, Switzerland and Belgium where the cohort patterns look very similar.Except for Germany, these are cases of early realignment with a longer-term experience of electorally strong radical right parties politicising the immigration issue (see, e.g.Kriesi et al. 2008).There is no country with an opposite pattern, i.e. immigration being significantly less relevant among millennials.The countries without cohort differences in the expected direction include those countries in which electorally relevant radical right 'anti-immigrant' parties have not emerged yet (Ireland) or did so only after 2018 (Spain, Portugal).Thus overall, the expected increasing association of immigration attitudes with left-right identification among more recent cohorts seems more likely to emerge where changes in party competition and issue politicisation have provided the corresponding cues -and it is these countries which significantly drive the overall trend.
Second, environmental protection is positively associated with left-wing identification for the generation X and, especially, millennials in most countries.In contrast, there seldom is a statistically significant association between caring for the environment and left-right self-placement among members of the interwar cohort.While cohort differences in the relevance of environmental protection are often modest, and the conditional effects are often not precisely estimated, the tendency for more positive associations among more recent cohorts is visible in most countries.In line with the party cueing perspective, the tendency seems clearer in countries with a longer history of electorally relevant Green parties and a significant and steady party-level salience of the environment from the 1980s onwards (e.g.Belgium, Germany, Sweden, Finland) (see Green-Pedersen 2019: 119) than in countries where Greens have played a more marginal role (e.g.Portugal, United Kingdom, Norway) (see Grant and Tilley 2019).
Third, in at least four countries attitudes towards redistribution are less closely associated with left-right identification among more recent cohorts: Netherlands, Germany, United Kingdom and Norway.In the UK, for example, a full-scale increase in attitudes towards redistribution goes along with a 1.8 increase in the predicted left-right position among the interwar cohort but only 1.0 among millennials.Similar numbers hold for Germany (1.7 vs. 1.0).In line with the party cueing perspective, the UK and Germany are paradigmatic cases of decreasing polarisation around economic policy issues between the mainstream left and the mainstream right.Around the turn of the millennium, Labour and the SPD moderated their economic policy positions in the context of the 'Third Way' (Karreth et al. 2013).Like the Netherlands as well, the UK and Germany are also among those countries in which the relative salience of economic issues in party manifestos has notably decreased over time (Hillen 2022: 12).However, in a majority of countries the effect of redistribution attitudes on left-right is more or less stable across cohorts.Interestingly, this includes countries in which the association is equally strong across cohorts (e.g.Sweden) and countries where it is equally weak across cohorts (e.g.Portugal).Overall, there appears to be no uniform crowding out of the 'old' redistribution issue.Rather it seems that it matters whether conflicts around redistribution remain salient and polarised in the party system amidst the rise of the new issues or not.
Fourth, regarding attitudes towards homosexuality, the modal pattern remains one of stability.However, there are a few countries in which tolerance for homosexuality is more closely associated with left-right positions among those born earlier -in particular within the interwar cohort.This mainly includes the three predominantly Catholic countries of France, Spain and Portugal, in which cultural liberalism vs. traditional (religious) morality issues have been historically strongly polarising (Lipset and Rokkan 1967;Rovny and Polk 2019).This historical background is still visible in the relatively close association between tolerance for homosexuality and left-right positions we see in these countries (see above).As this applies more to the interwar cohort than to recent cohorts, generational differences emerge in these countries.
The comparative approach adopted here allows us to place the previous single-country studies on the Netherlands (de Vries et al. 2013;Rekker 2016) into a cross-national context.Doing so reveals that the Netherlands constitutes a model case.Cohort differences regarding the relevance of immigration, environmental attitudes, and redistribution for left-right identification in the Netherlands all resemble the patterns for Western Europe as a whole.The same can be said of Germany and, a bit less so, Belgium.In these three countries, there is a remarkable shift in which issues dominate left-right positions: within the interwar cohort, redistribution dominates, whereas among millennials left-right orientations are more strongly related to attitudes towards immigration than to attitudes towards redistribution.
In sum, the results of the by-country analysis show that the patterns identified for Western Europe as a whole do not apply to each individual country.Yet, in line with the party cueing and political socialisation mechanisms put forth in this article, variation in cohort patterns seems to reflect historical differences in party competition and issue politicisation across countries.Immigration is more relevant among millennials especially in countries in which electorally strong radical right parties have enduringly politicised the issue.While the environment tends to gain in importance for recent cohorts almost everywhere, the trend appears particularly pronounced in countries with a longer history of strong Green parties and a high salience of the environment.Evidence of crowding out of redistribution is limited to about a third of the countries and these countries -which seem to drive the overall trend -tend to have experienced episodes of a decline in politicisation of economic left-right issues in party competition.While a more detailed discussion of patterns in all individual countries is beyond the scope of the present contribution, the exploratory by-country analysis reveals cross-country variation that accords with this article's core claim that party-level issue politicisation and political socialisation together shape the issue content of left-right identification.

Conclusion
Political competition in Western Europe has changed a lot over the last decades.'New' issues, such as immigration and environmental protection, have gained in salience and are important for today's political divisions.In some cases, the rise of new issues seems to have pushed 'old' issues that once dominated political competition -most notably economic conflicts around redistribution -to the back.However, the tradition to describe politics via the left-right semantic has persisted.Against this background, this article has studied how attitudes on old and new issues are associated with left-right identifications among Western European citizens.I have suggested that the issue content of left-right orientations differs across cohorts due to a combination of a party cueing mechanism and a political socialisation mechanism.
In an empirical analysis of ESS data for 12 Western European countries from 2002 to 2018, I found that environmental and immigration attitudes are stronger predictors of left-right positions among more recent cohorts, immigration chiefly among millennials.In contrast, attitudes towards redistribution are, overall, less relevant within recent cohorts, suggesting a moderate crowding out of old issues.A by-country analysis reveals that these trends do not apply equally to all countries but seem to vary in correspondence with patterns in party competition and issue politicisation.The trend towards the environment and immigration becoming more important for more recent cohorts tends to be more pronounced in countries with a longer history of strong Green and, respectively, radical right parties; and the trend towards redistribution becoming less important appears concentrated in countries with decreases in the salience of party competition over economic issues.Overall, the empirical results provide ample support for the expectations on cohort differences derived from the party cueing and political socialisation mechanisms.
Yet, this study is not without limitations.An important limitation is the rather short observation period of just 16 years which makes disentangling cohort and life-cycle effects challenging.This adds uncertainty to the findings, though there is little a priori reason to expect strong life-cycle effects on the issue content of left-right, and the empirical analysis also suggests that these play a subordinate role at best.Still, future research may draw on a longer time series to reduce this uncertainty.With a longer time series, it would also prove possible to incorporate additional complexities and to relax the simplifying assumptions associated with modelling left-right orientations as linear additive functions of issue attitudes.For example, previous research suggests that individual issue attitudes might relate to left-right orientations in non-linear ways (Lachat 2018) and that issue attitudes might be interactively related to left-right orientations (Gidron 2022).Finally, a longer time series may also provide more insights into period effects.While these have not been the focus of this article, the analysis has provided two important findings: immigration has become more important in the aftermath of the European refugee crisis and redistribution in the aftermath of the European debt crisis.Only future research will be able to tell whether these trends persist, and, especially, whether the latter finding foreshadows a reversal in the generational shift towards a decreasing relevance of redistribution identified in this study for some countries.
Future research may also attempt to clarify the role of external, structural drivers of salience versus the issue politicisation by parties.For example: is immigration (the environment) increasingly relevant for more recent cohorts simply because these have been socialised into a world with much higher levels of immigration (where environmental issues have become more urgent)?From the perspective of this article, such structural drivers are important because they provide the potential for issue politicisation by parties, but it should ultimately be the issue politicisation by parties that matters for the issue content of left-right orientations.
Limitations aside, the present study has several important implications.First, the article's findings add to evidence demonstrating the dynamic nature of the issue content of left-right and bring new evidence on how this dynamic plays out.The findings support what Fuchs and Klingemann (1990: 234) expected some thirty years ago: 'the left-right schema will be retained even if the conflict structures of advanced industrial societies are changing.This means that the new dimensions of conflict have to be integrated into the left-right schema' .This study -adding to Rekker's (2016) study of the Netherlands and Knutsen's (1995) older study of Western Europe -suggests that this transformation of left-right works to a significant extent through generational replacement.This is an important result because it suggests that issue content does not only change over time, but that different understandings among cohorts coexist at any single point in time.
Second, this coexistence of different understandings of what it primarily means to be 'left' (or 'right') may cause tension and rifts between different generations within the 'left' (and 'right').These rifts, in turn, create challenges and trade-offs for parties who aim to appeal to voters identifying as 'left' (or 'right') from different generations whose left-wing (or right-wing) identities are differentially connected to issues.
Third, for political scientists, these cohort differences are also important on the grounds that they illuminate how issue competition has evolved and give indications on what to expect for the future.As generational replacement is guaranteed by nature, left-right understandings of more recent cohorts will supersede those of those previous cohorts leading to a new balance of left-right oppositions at the public level.
Fourth, with the shifting issue content of left-right the relation between socio-economic characteristics and left-right identification might change as well.When left-right is mainly about redistribution, we can expect those from socio-economically disadvantaged strata to tend to identify as left-wing.When left-right is mainly about immigration and the environment for new cohorts this might be no longer the case.This is worth investigating in future research.
Fifth, this study's findings are yet another reason to use cautiously left-right positions as a measure of issue positions.On the one hand, the findings indicate that citizens' understanding of the left-right dimension integrates new conflictual issues -thereby speaking for the continued relevance of left-right as an orientation scheme.On the other hand, the fact that the issue content of left-right varies not only across countries but also within countries across cohorts points out that it is problematical to assume that left-right self-placement captures the same issue positions -be it across space, time or cohorts.

Notes
1.I thereby follow an understanding of party-level issue politicisation as being the joint product of issue salience and issue polarization (see, e.g., Hutter and Kriesi 2022).
2. Respondents who answered 'don't know' are excluded, following this contribution's focus on the issue correlates of left-right orientations among those who hold such an orientation.For research on why respondents do (not) hold left-right orientations, see Otjes and Rekker (2021), who also emphasize political socialization.In all countries studied here, the share of 'don't know' responses is below a fifth (see Figure A1 in the online appendices).3. The decision to use all nine ESS waves forces me to use those single items consistently included.Only in the case of immigration attitudes would it have been possible to combine several items.For reasons of consistency, I took a single item in this case as well.A factor built from several immigration items correlates highly with this item (over 0.8) and using it instead leads to similar main findings (see Figure C1 in the online appendices).4. Controlling for education as an interactive confounder is important because, first, left-right positions are more strongly associated with issue attitudes the higher an individual's level of education (Inglehart and Klingemann 1976;Lesschaeve 2017).Second, individuals in more recent cohorts tend to hold higher levels of formal education (see Table A1 in the online appendices for a cross-tabulation).Thus, we might see stronger associations between issue attitudes and left-right positions among more recent cohorts just because these tend to be more educated.Equation (1) adjusts for education to get at a 'pure' cohort effect that is informative about the effects of political socialization. 5. Age, or life-cycle, effects would mean that associations between issue attitudes and left-right positions regularly change over the life course.It is not apparent why life-cycle experiences should affect the issue basis of left-right positions (cf.Rekker 2016: 125), yet it cannot be ruled out on theoretical grounds.In the cumulated ESS data, the relatively short observation period of 16 years introduces high collinearity between cohorts and age: 86% of the variance in age (in years) is accounted for by cohort membership.I thus use a rough distinction of age groups that keeps collinearity with cohorts at an acceptable, while inevitably high, level (see Table A1 in the online appendices for a cross-tabulation).6. Figure 1 also shows that more recent cohorts are more open to immigration and more tolerant of homosexuality, in line with theories of generational value change (Inglehart 1977;Norris and Inglehart 2019).Despite this, there remains sufficient variation in attitudes within cohorts (see Figure A2 in the online appendices).7. Figure B1 in the online appendices presents conditional effects by levels of education, showing that all four issue attitudes are much more strongly related to left-right orientations for those with higher levels of formal education.This is in line with previous research (cf.Inglehart and Klingemann 1976;Lesschaeve 2017) and vindicates the rationale for adjusting for education.

Figure 1 .
Figure 1.issue attitudes and left-right positions across cohorts.note: scatterplot of left-right positions and issue attitudes within generational cohorts with jitter and linear regression line added.pearson correlation listed in the title.

Figure 2 .
Figure 2. effects of issue attitudes on left-right positions across generational cohorts from baseline model.

Figure 3 .
Figure 3. conditional effects of issue on left-right positions from model with age X issue interactions included.

Figure 4 .
Figure 4. effects of issue attitudes on left-right positions by cohort across countries.