Immigration attitudes and the emerging education divide in left-right identification

Abstract

Across the advanced democracies, educational attainment has become increasingly linked to citizens’ political orientations. Education, it seems, polarizes citizens – with the higher educated turning to the left and the lower educated to the right. Here, we contend that this ideological divide between the highly and less educated does not stem from changing political opinions, but rather from a change in the issues that are associated with left and right. Drawing on over two decades of data from the European Social Survey, we demonstrate that Europeans have progressively come to associate left-right positions more with views on immigration. As a result, long-standing educational divides in immigration attitudes have become more strongly reflective of left-right self-identifications. Our findings suggest that the growing education-based cleavage is driven not by opinion change, but by the rising salience of immigration in structuring ideological labels.

Recent research has drawn attention to a growing education-based ideological polarization between citizens in the established democracies (Hooghe and Marks 2025; Stubager 2013). As Garritzmann (2025) has argued, while there has always been conflict and opposition between educational groups – going back as far as the Middle Ages – the postwar spread of mass higher education has resulted in education becoming much more politically important. According to Garritzmann (2025: 11), education now is a real cleavage and shapes citizens’ political identification and behaviour. The growing education-based cleavage is visible in voting behaviour, with lower and higher educated voters supporting different parties. These differences result from the different values that lower and higher educated individuals hold – in particular in terms of the liberal-authoritarian dimension, which includes attitudes on immigration (Stubager 2013).

In this article, we argue that this widening education-based gap – in terms of individuals’ ideological identification – does not reflect over-time changes in the views that the higher and lower educated hold, but instead signifies a change in the issues that are associated with the left-right dimension. We argue that the growing education-based gap in ideology results from the fact that attitudes on immigration have become increasingly predictive of where citizens position themselves on a left-right scale, a change that De Vries et al. (2013) have already documented in the Dutch context.

To study education gaps in ideology, how this gap widens over time, and how it relates to changes in the issues associated with the left-right dimension, we use the individual-level data from the European Social Survey (ESS). We use all available rounds of the ESS, covering the period between 2002 and 2024, and focus on 15 European countries for which data is available in each of the 11 rounds of ESS.

Our empirical analyses reach two conclusions. First, we demonstrate that citizens’ attitudes towards immigration are increasingly connected to how they position themselves on a left-right dimension. Thus, the findings from De Vries et al. (2013) regarding the strengthening of the connection between anti-immigrant attitudes and left-right self-placements in the Dutch context apply to Western Europe more generally. Second, we show that because attitudes on immigration are systematically connected to education, and more so than attitudes about economic conflict, an education-based cleavage in left-right positions emerges and has been increasing as views on immigration became more strongly connected to left-right identification.

These findings advance the literature that has drawn attention to a transformation of party competition and electoral behaviour. This stream of research has argued that change results from the waning of economic conflicts, while second-dimension politics – including conflicts around immigration – have gained in importance (Hooghe and Marks 2018; Kitschelt and Rehm 2023). This transformation, we demonstrate, is also reflected in left-right identification. Our work is also connected to studies that have theorized about the social structure and social identities that underpin party competition on the new sociocultural dimension (Bornschier et al., 2024; Stubager 2013). We find that these group memberships polarize individuals, not only in party support, but also in terms of a more general left-right identification. Finally, our study contributes to a line of research that has pointed out that the content of left-right identification varies over time (De Vries et al., 2013; Steiner 2024) and between individuals (Bauer et al., 2017). Differences in the issues that are connected to left and right, we show, can alter which and how strongly sociodemographic variables correlate with left-right identification.

The puzzle: A growing education-based gap in left-right identification

Individuals’ level of education has come to increasingly shape where they place themselves on a left-right scale. This is shown in Figure 1, which uses data from the 15 countries that have participated in every currently available round of the ESS. ¹ The graph demonstrates that in terms of their average left-right position, higher and lower educated individuals are drifting apart. More precisely, the left-right gap between the least and most highly educated groups was 0.14 points in round 1 (4.97 vs 4.83), but widened to a 0.55 points difference on average in round 11 (5.11 vs 4.56). ²

Figure 1.

Educational attainment and left-right self-placement over time – ESS 2002–2024.

The graph also indicates that this education-based gap in left-right self-placements started to widen roughly around rounds 7 and 8 of the ESS (between 2014 and 2016). From then onwards, the tertiary educated begin to move more to the left on the ideological self-placement scale. At the same time, those with non-tertiary education moved their ideological self-placement somewhat to the right.

The left-right dimension

The left-right dimension has proven to be an effective way to summarize the issue positions of parties and voters, to predict which parties voters prefer, and to better understand how party competition functions (e.g. see Dalton and McAllister 2015; Navarrete 2021). Furthermore, citizens think in terms of left and right, identify as left or right, and individuals’ left-right identifications have even been found to be transmitted between generations (van Ditmars, 2023). While left-right positions are thus a meaningful and important determinant of individuals’ party preferences and political behaviour, how individuals interpret left and right has been found to be context- and individual-specific (Bauer et al., 2017; Jessee 2021).

Our expectations and theoretical argument follow from earlier work that has argued that the ‘meanings’ of left-right orientation are context-dependent, differing between countries (Huber and Inglehart 1995; Tavits and Letki 2009), over time (De Vries et al., 2013; Giebler et al., 2021; Knutsen 1995; Steiner 2024) and between individuals (Bauer et al., 2017). There thus is a large body of research that has made the argument that the issue positions that are reflected in left-right placements vary, which is consistent with the idea of ‘left-right’ being an ‘empty vessel’ that incorporates new issues as they emerge (Bobbio 1994), or that functions as a ‘super-issue’ (Van der Eijk et al., 2005).

In Western democracies, left and right have long been associated with economic issues, reflecting different positions in terms of class conflict and economic redistribution, with the left being in favour and the right being opposed to redistribution (Inglehart 1990; Kitschelt and Hellemans 1990). In recent years, however, the increased salience of postmaterialist issues has led the left-right dimension to reflect positions on postmaterialist values too (Kitschelt and Hellemans 1990; Knutsen 1995). More recently, the conflict over cultural issues, ‘relating to traditional lifestyles, rights of immigrants or Islam’ (De Vries et al., 2013: 223) has increased in salience and had a greater impact on political competition, leading voters to associate left and right more with positions on immigration.

De Vries et al. (2013) theorized that this over time change results from a dynamic of issue bundling, whereby a new policy issue, after it emerges, ‘becomes more and more integrated into the left/right dimension over time’ (p. 228). De Vries et al. furthermore theorized that there are dynamics of issue crowding out as well, ‘whereby traditionally more prominent issues for left/right identification (such as state intervention in the economy) become less important for voters’ left/right identification’ (p. 228). As a consequence of the dynamic of issue bundling, therefore, as immigration has gained salience across Western Europe, immigration attitudes can be expected to have become more strongly associated with left-right identification. Furthermore, issue crowding out would imply that other issues – economic ones in particular – become less strongly associated with left-right identification. Empirically analysing longitudinal survey data from the Netherlands, De Vries et al. (2013) find that between 1980 and 2006, the left-right self-placements of Dutch voters became more connected to their attitudes on immigration. Along the same lines, Steiner (2024) shows that among more recent cohorts in Western Europe, immigration attitudes are more strongly associated with left-right identities.

In recent years, conflict over immigration has increased markedly, with the 2014–2016 refugee crisis representing a watershed moment for the impact of immigration on European politics, leading to increased support for radical right parties, a polarization in terms of immigration attitudes (Van der Brug and Harteveld 2021), more polarization over immigration and an increased salience of immigration (Hutter and Kriesi 2022) as well as increased Euroscepticism (Stockemer et al., 2020). Furthermore, there already is much evidence that anti-immigrant views are an increasingly strong predictor of the vote choice and structure voting for radical right parties in particular (Kriesi et al., 2006, 2012). Here, we are particularly interested in examining whether this increased salience also leads to issue-bundling with regard to left-right self-identification. Given that the salience of immigration shapes how strongly it is associated with the left-right dimension (Giebler et al., 2021), we hypothesize that the association between views on immigration and left-right self-placements has strengthened across European countries. ³

H1

Over-time, attitudes on immigration are increasingly connected to left-right self-placements.

While H1 builds on earlier work that has studied the impact of the growing salience of immigration attitudes and its impact for left-right identification in particular (De Vries et al., 2013; Steiner 2024), our main interest is in the implications of a strengthening association between anti-immigrant attitudes and left-right self-placements for the education gap in left-right identification. On this front, we contend that it is the strengthening of the association between attitudes on immigration and left-right identification that gives rise to a widening education gap in left-right self-placements. This expectation is based on the observation that attitudes on immigration correlate more strongly with educational attainment than economic issues such as income redistribution. Much work has already documented the strong connection between education levels and anti-immigrant sentiment (Cavaillé and Marshall 2019; Schäfer and Steiner 2025). This association, Hainmueller and Hiscox (2007) have shown, does not result from the fact that the higher educated are not competing for jobs with immigrants, but reflects the values and beliefs that they hold. As Hainmueller and Hiscox (2007: 437) state, ‘more educated respondents are significantly less racist and place far greater value on cultural diversity in society’. There also is work that has shown that individuals’ level of education correlates with their economic views. Attewell (2022: 1094), for example, finds that the higher educated have ‘more positive attitudes towards the deservingness of welfare state beneficiaries compared to less-educated individuals’, but also that the higher educated prefer a smaller welfare state.

Work that studied both the association between education and anti-immigrant or sociocultural attitudes as well as that between education and economic views finds that the anti-immigrant and sociocultural attitudes correlate more strongly with educational attainment (Stubager 2013). This also holds for the data analysed in this paper. As we document in Appendix B, education correlates weakly with attitudes on income redistribution but is a strong predictor of anti-immigrant views – with the higher educated being significantly and substantially more favourable towards immigration. As a result, as immigration structures left-right views more strongly, the socio-demographic factors that structure attitudes on immigration (i.e. education) will become more connected with left-right positions. ⁴ This leads to the second hypothesis:

H2

Views on immigration explain the widening education gap in left-right self-placements.

Our first hypothesis is in line with earlier literature that has argued and shown that ideological self-placements are context-dependent, and are more strongly associated with positions on issues that are more salient (De Vries et al., 2013; Giebler et al., 2021). Our second hypothesis additionally relies on the assumption that the association between issue positions and left-right self-placements results from a process whereby issue positions shape ideology rather than vice versa. Only if causality runs that way, would we expect the more liberal positions of the higher educated on immigration to translate into more left-wing ideological positions – leading to the emergence of an education gap in left-right self-placements.

Data and methods

To test our hypotheses, we use all currently available rounds (1 to 11) of the ESS. Given our interest in change over time, we restrict the dataset to the 15 European countries that participated in all 11 rounds. Our main outcome variable is citizens’ left-right self-placement, which ranges between 0 and 10 – with 0 corresponding to the left and 10 to the right. To examine educational differences in ideological self-placements, we distinguish between three groups: (1) a lower educated group that has either a primary or lower secondary education; (2) a middle group that has a secondary or post-secondary degree; and (3) respondents who hold a tertiary degree (for details and descriptive statistics, see Appendix A). ⁵

To measure anti-immigrant views, we rely on two sets of indicators that were included in all ESS rounds. First, we use a summary scale that combines citizens’ positions on questions asking them whether immigration is good or bad for the country’s economy, enriches or undermines cultural life, and makes the country a better or a worse place to live. These items correlate strongly (Cronbach’s α = 0.85) and were summed and rescaled to range between 0 and 1, with higher values corresponding to more anti-immigrant positions. Second, we measure how tolerant respondents are of immigrants from poorer countries outside Europe and from a different race or ethnic group than the majority. Both items, which correlate strongly (Cronbach’s α = 0.88), had four-point answer scales, from ‘allow many’ to ‘allow none’. We summed the items and rescaled the indicator to range between 0 and 1, where higher values reflect more anti-immigrant attitudes.

Because left-right is traditionally associated with economic attitudes, we also take these views into account. We rely on a measure of agreement with the statement that the ‘Government should reduce differences in income levels’. Respondents could indicate their level of agreement on a five-point scale, from agree strongly to disagree strongly. We rescaled the item to range between 0 and 1, with higher values corresponding to more economically right-wing views.

To test H1, we estimate OLS regression models in which left-right self-placement, ranging from 0 to 10, is the outcome variable. We regress left-right positions on anti-immigrant views, an indicator for the ESS round and interactions between ESS rounds and anti-immigrant views. To account for the role of economic views, we add to the model an indicator of citizens’ views on whether the government should reduce income differences (also interacted with ESS rounds). We account for differences between countries through country fixed effects and cluster standard errors by country. We test H2 by comparing the estimated education-based gaps in left-right self-placements from models that do and models that do not account for respondents’ anti-immigrant views. We expect gaps to be substantially smaller when we control for anti-immigrant views.

Results

We start by examining whether anti-immigrant attitudes are increasingly connected with citizens’ left-right self-placement (H1). We summarize the results in Figure 2, with a focus on the association between the perceived value of immigration and left-right (left-hand panel) and the association between positions on how many immigrants can be admitted to the country and left-right positions (right-hand panel). For both independent variables, higher values correspond to a more anti-immigrant position.

Figure 2.

The association between anti-immigrant attitudes and left-right positions over time (a) Value of immigration for the country (b) Admitting immigrants.

The evidence supports our hypothesis. For both indicators, we observe a strong increase in the strength of the association between anti-immigrant attitudes and left-right self-placement, starting around round 8 (fielded in 2016). The effects are substantial. In terms of citizens’ views about the value of immigration (left-hand panel), in round 1 the most anti-immigrant respondents placed themselves on average 1.9 points further to the right compared to the least anti-immigrant respondents. In round 11, a similar difference in views on immigration results in respondents placing themselves 3.0 points further to the right, corresponding to a 58% increase in the strength of the association between anti-immigrant views and left-right self-placements.

For views on how many immigrants from outside Europe and from a different race or ethnic background can be admitted into the country (right-hand panel), the results show an even stronger increase in the association with left-right positions. Between rounds 1 and 11, the effect of moving from the most to the least in favour of admitting immigrants increased from 1.4 to 2.4 – corresponding to a 71% increase in the effect of anti-immigrant attitudes on left-right positions. These results are in line with what De Vries et al. (2013) observed in the Dutch context, indicating that the strengthening of the association between anti-immigrant attitudes and left-right self-placements characterizes public opinion in Europe more generally. We also note that, as can be seen from Appendix B, the association between left-right positions and economic views has remained largely stable over time.

Finally, we verify whether differences in immigration attitudes between the higher and lower educated account for the widening of the education-based gap in left-right ideology. Panel a in Figure 3 plots the expected left-right self-placements of respondents in different educational groups, by ESS round. These estimates are based on models in which we only account for time, country differences, respondents’ level of education, as well as their economic attitudes (detailed estimates are reported in Appendix D). The estimates show a picture that closely resembles Figure 1, indicating an over-time widening left-right gap between the lowest and highest educated. As can be seen from panel b, this education-based ideological gap all but disappears when we account for anti-immigrant attitudes. In fact, the left-right positions of higher and lower educated individuals can no longer be distinguished up until round 9 of the ESS, and they are much reduced in rounds 10 and 11. In line with H2, therefore, we find that views on immigration explain the widening of the education-gap in left-right positions.

As can be seen from Appendix E, our results are substantively the same when we account for respondents’ views on how many immigrants can be admitted rather than the indicator of the extent to which individuals value immigration. Over time, there has been an expansion of tertiary education, with many more tertiary educated among the youngest age groups than among older generations. Given that age also correlates with political attitudes and ideology (O’Grady 2023; Tilley 2002), it might be important to account for the effect of age when estimating the effect of education (Schäfer and Steiner 2025). In Appendix K, we show that our conclusions still hold when we control for respondents’ age.

Figure 3 suggests that the emerging educational cleavage in left-right ideology results from an over-time change in the issues that correlate with left-right positions in European democracies, not from a shift in the political views of the lower and higher educated. That it is not shifts in opinion that drive change is substantiated by supplementary analyses that are reported in Appendix F. These analyses indicate that there has been a depolarization in terms of the economic views of different educational groups, while differences in the anti-immigrant attitudes of the lower and higher educated have remained stable over time. If anything, the political views of higher and lower educated respondents have grown more similar over time. ⁶

Figure 3.

Educational attainment and left-right positions, with and without controls for anti-immigrant attitudes (a) No control for anti-immigrant attitudes (b) Accounting for anti-immigrant attitudes.

Conclusion

For half a century, left-right self-placement has been a core variable in political science and has acted as a proxy measure for mass policy preferences. While it has been the subject of frequent methodological and theoretical criticisms (for reviews, see Bauer et al., 2017; Lo et al., 2017), it remains the single best indicator of where citizens position themselves on their country’s major policy debates. The observation that tertiary educated voters are moving to the left and non-tertiary educated voters are moving to the right, therefore, implies that mass publics are increasingly ideologically polarized based on education.

We have argued that this apparent change in the ideological views of those with higher and lower education is less reflective of polarization in the attitudes that different educational groups hold and instead is caused by a change in the issues that are associated with left-right positions. Using all 11 rounds of the European Social Survey conducted since 2002, we find that attitudes towards immigration have become more connected to left-right identification. Since immigration attitudes are strongly related to tertiary education, this accounts for the increasing ideological polarization around education. We also find that the change occurred at a specific point in time – around 2014 and 2016 – and has continued since then. Why it occurred at this point in time is a matter of speculation, but at least two factors may be important. First, the refugee crisis of 2015–16, when large numbers of migrants entered Europe fleeing conflict in the Middle East and North Africa, heightened the public’s awareness of immigration (Economou and Kollias 2024; Van der Brug and Harteveld 2021). Second, parties have responded to this growing public awareness by giving more attention to immigration in their party platforms (Dancygier and Margalit 2020), potentially further solidifying the salience of immigration in political debates.

In studying change over time, we made the assumption that the issue of immigration has gained in salience over time. In line with the findings of Giebler et al. (2021), who showed that issues that are more salient to individuals structure the perceptions of parties’ left-right positions more, we expect the increased salience of immigration to lead to a stronger association between anti-immigrant views and left-right identity too. Unfortunately, because the ESS does not include indicators of issue importance, we cannot test this assumption at the individual level. As we show in Appendix H, however, the salience of immigration at the party system level does condition the association between immigration attitudes and left-right self-placements. Especially in contexts where political parties talk more about immigration, these supplementary analyses show, immigration attitudes are more strongly connected to individuals’ left-right identifications. ⁷

In this paper, we studied changes in the association between left-right self-placements and individuals’ attitudes on more specific policy issues. In line with De Vries et al. (2013) and Steiner (2024) we infer from over-time changes in the strength of associations between left-right self-placement and other policy attitudes that left-right is being redefined, due to variation in the importance of specific issues. However, these measures only provide a partial view of changes in the meanings of left-right. Ideally, we would also have measures of respondents’ perceptions of parties’ positions on a left-right dimension and on specific policies – to verify if there is also evidence of change in the issues that individuals associate with parties’ left-right positions (Giebler et al., 2021). Another alternative would be to use information from questions asking respondents to indicate more directly how they understand left and right, for example, by means of open-ended survey questions (Bauer et al., 2017). Unfortunately, because of a lack of longitudinal survey data that includes such measures, we cannot pursue such analyses.

Our findings suggest that the much debated potential for the realignment of European party systems around an education divide may be overstated. While there is clear evidence that the cleavage structures of the major party systems are gradually moving away from their traditional bases in class and religion (Hooghe and Marks, 2025), it would appear that the education divide in political views has existed for sometime, but since 2015 has simply become more visible – as immigration became more salient than was the case in the past. While education provides a lens through which citizens view immigration, it is less education that divides them than the immigration debate, and how the main parties approach the issue.

Supplemental material

Supplemental Material – Immigration attitudes and the emerging education divide in left-right identification

Supplemental Material for Immigration attitudes and the emerging education divide in left-right identification by Ruth Dassonneville, Ian McAllister in Research & Politics