Generations and the changing character of support for European unification in the Netherlands: a research note

This study examines whether younger generations are more likely to associate their support for European unification with cultural, rather than economic issues. The EU has changed from an ‘economic community’ to a ‘political union.’ Because most citizens form relatively stable orientations during their ‘impressionable years,’ we expect recent generations to be more likely to view European unification through a cultural lens. An analysis of 12 waves of panel data from the Netherlands finds the strongest correlation between EU support and cultural attitudes among the newest generations. However, these generations are not less likely to associate EU support with economic attitudes. Moreover, between 2007 and 2019, Euroscepticism became increasingly associated with cultural attitudes among all generations and age groups. These findings indicate that EU support has become more strongly aligned along a cultural dimension and that this realignment will become more pronounced as newer generations replace earlier ones.


Introduction
Over the course of decades, the European Union (formerly European Economic Community) changed in character from an 'economic community' aimed at facilitating international trade, to a 'political union', which has much influence on the public policies of its member states. When the European Union and its predecessors were mainly an economic union, aimed at promoting trade between European countries, support for and opposition to European integration was seen to stem mainly from utilitarian considerations (e.g. Anderson 1998;Eichenberg and Dalton 1993;Gabel 1998). As the EU became increasingly a political union, matters of national identity and other socio-cultural considerations became increasingly important (e.g., Carey 2002;Hooghe and Marks 2018;McLaren 2006). In this brief research note we investigate whether there are generational differences in the drivers of support for European unification. 1 Research on political socialisation demonstrates that people obtain basic political views during the most impressionable years, roughly between 12 and 25, and that these views remain rather stable later in life (for overviews see Van der Brug and Franklin 2018). As a result of differences in the social circumstances during which people were socialized, there are systematic differences between generations in political participation (e.g., Franklin 2004;Neundorf and Smets 2017), party choice (e.g., Dassonneville 2013;Tilley and Evans 2014) and attitudes (e.g., Inglehart 1977). Particularly large generational differences can be observed in support for European unification, where the younger generations are much more supportive of further unification than older generations (e.g., Down and Wilson 2013). Yet, what we do not know is whether generations also differ in the way support for European unification is aligned with other values.
Understanding differences between generations in the factors that drive support for European unification is scientifically as well as societally important. While the scientific literature has shown that generational differences exist in support for European unification, even when controlling for life-cycle effects, we do not know much yet about the underlying mechanisms. In particular we do not know much about generational differences in the way in which attitudes towards the EU are formed. In a broader sense, our study contributes to our knowledge of changes in the way political conflict are structured (see also, Rekker 2016; Van der Brug and Rekker 2021). Besides its scientific relevance, we think it is also important for our understanding of the way in which attitudes towards European unification are becoming politicised. If there are generational changes behind the increasing socio-cultural character of such attitudes, these changes will become more pronounced as newer generations replaces earlies ones.
We study generational differences in the correlates of support for European unification by analysing individual level panel data (N = 12,522; over 60,000 observations over time) from the Netherlands that span the period between 2007 and 2019. This enables us to distinguish between generational differences and life-cycle effects in the ways in which socio-cultural attitudes are related to EU support. We demonstrate that, even in this relatively short period of time, socio-cultural values become more strongly related to EU support. As expected, this relationship is stronger among the youngest generation than among older ones. However, the increase in the strength of the relationship is clearly visible among the older generations as well.

Theory
As people grow up, they learn about the world they live in, which includes politics. Early electoral research emphasised the importance of parental socialisation in the way people derive basic values and partisanship (e.g., Campbell et al. 1960;Jennings and Niemi 1968). This portrayal of politics did not allow for much change.
As each generation would have the same basic attitudes as the generations before them, differences between generations were expected to be small and election outcomes were expected to be stable. Yet, in the 1960s it became clear that the post WW-II baby-boomers did not simply adopt the values of their parents. Inglehart (1977) provided an explanation for differences between generations, which became very influential in the field. According to him, people develop their most basic value orientations during the most impressionable years, largely in response to the social, economic and political circumstances at that time. The most impressionable years for political learning are during adolescence and early adulthood. The basic value orientations that one acquires during those years tend to be rather stable later in life.
Since basic values do not change much over the course of people's lives, generational replacement is an important driver of value changes in society.
While our study does not focus on value change, but on (changing) predictors of support for European unification, Inglehart's insights are highly relevant for our study. In the same way as people develop certain behavioural patterns and basic political values during the most impressionable years, they will learn about how to think about the European Union. When the EU was still called the European Economic Community (EEC), it was clear that its main objectives were to remove import taxes in order to promote international trade. This was mostly supported by right wing parties, while opposition came mostly from labour unions and left leaning parties, which were afraid that this could lead to unfair competition from countries with less generous social welfare arrangements (e.g., Marks 2004;Van Elsas and Van der Brug 2015). Citizens who learned about the EEC at that time, will have based their EU attitudes mostly on these kinds of socio-economic value orientations. As the EEC developed into the EU, its political power and responsibilities increased in stages. With the Schengen agreement and the introduction of the Euro, it has become clear that the EU is now involved in various policies which used to be decided exclusively at the national level, such as migration and the bailout of Greece. As a consequence, attitudes towards the EU have become more strongly linked to attitudes towards immigration and feelings of nationalism, even though socio-economic issues are expected to remain important as well (e.g., Bruter 2003;Carey 2002;Hooghe and Marks 2018;McLaren 2006). So, the guiding hypothesis of this study is that attitudes towards European unification will be more related to socio-economic policy preferences among older generations, who were politicised in a period when the EU was mainly an economic community, while socio-cultural policy preferences (particularly attitudes towards immigration) 2 are more strongly related to European unification attitudes among the younger generations.
The political socialization model can be contrasted with the 'lifelong openness' model, which holds that political views are not so stable over the course of one's lifetime, but rather that people adapt their orientations in light of new developments (Alwin and Krosnick 1991;Tyler and Schuller 1991). So, where the theory of political socialisation predicts differences between generations that reflect their experiences during their most impressionable years, the lifelong openness model does not predict such differences. Of course, the two perspectives do not rule each other out completely. Even if the older generations adapt their attitudes to the changing character of the EU, there might be systematic differences remaining between the generations. Moreover, older generations may adapt their perspectives of the EU more slowly than younger generations. This is simply because people tend to get 'stuck in their ways' as they grow older (Franklin 2004). Consequently, people are most vulnerable to be influenced by those events and developments that occur when they are teenagers or young adults.
One of the major issues in a study of generational differences is that we need to distinguish between life-cycle effects and generational differences. Older people may differ from younger citizens because they are in a different stage of their lives or because they grew up in a different historical period. So, in order to study generational differences, we cannot simply compare younger and older citizens. Moreover, different generations can develop differently over time. Interestingly, this may either take the form of a divergence or a convergence. On one hand, generations may diverge because young generations are still more impressionable and hence adapt more to new political developments. Older people may contrarily adapt less to changes, because they have already learned about the EU. On the other hand, generational research has also documented instanced in which younger and older generations convergence over time (Rekker 2016). This may occur because older voters also learn from new development, albeit at a slower pace than younger citizens. In such instances, young voters can be seen as a 'vanguard of political change' that is eventually followed by older voters who need more time to adapt to a changing political reality (Rekker 2022). In order to separate the mechanisms of political learning from generational differences, we will use panel data (to be discussed below), which allow us to separate changes that occur within individuals from differences between people of different age groups. Theoretical reasoning above leads us to expect the following: H1 Attitudes towards European unification will be more strongly related to socioeconomic attitudes among older generations than among voters who came of age after the establishment of the EU in 1993.
H2 Attitudes towards European unification will be more strongly related to sociocultural attitudes among voters who came of age after the establishment of the EU in 1993 than among older cohorts.
H3 Over time, attitudes towards European unification will become less strongly related to socio-economic attitudes across all generations and age groups.
H4 Over time, attitudes towards European unification will become more strongly related to socio-cultural attitudes across all generations and age groups.

H5
The changes predicted by H3 and H4 could either be stronger (H5a) or weaker (H5b) among the youngest than among the oldest cohorts. Respondents were recruited as a probability sample of Dutch citizens aged 16 and over. Panel attrition was handled by selectively recruiting new participants who resemble the respondents who dropped out on key variables. This study analyzes data from a total of 17,267 respondents who participated in at least one wave and in 4.4 waves on average. This panel data allowed us to obtain an exceptionally fine-grained picture of subtle difference between periods and generations by leveraging both between-person and within-person variation with a large sample size.

Measures
Support for European unification was measured with the following survey item: "Where would you place yourself on a scale from 1 to 5, where 1 means that European unification should go further and 5 means that it has already gone too far?" Socio-cultural attitudes were measured with a single item about immigration: "Where would you place yourself on a scale of 1 to 5, where 1 means that immigrants can retain their own culture and 5 means that they should adapt entirely?" Socio-economic attitudes were measured with the following item: "Where would you place yourself on a scale from 1 to 5, where 1 means that differences in income should increase and 5 means that these should decrease?" Ideally, we would have preferred to measure our key variables with multiple items, but for the purpose of our study we need measures that are asked each year in exactly the same format and other items are not available. However, we do believe that these items are at the heart of the concepts that we wish to measure. 3

Analyses
Our hypotheses focus on the question whether the relationships between support for European unification and other attitude scales became stronger (or weaker) over time and whether these relationships differ systematically between generations. We employ a regression design, which enables us to test the hypotheses by means of interaction effects (between period and generations on the one hand and attitude scales on the other), employing support for European unification as the dependent variable. We do not interpret these regression coefficients in causal terms. It is simply a tool to assess whether the relationships between the attitudes differs over time or between generations. This study uses regression models for panel data with random effects. These models leverage both between-person and within-person variation to estimate a single parameter. This results in more efficient parameter estimates (Cameron and Trivedi 2005), which is particularly useful to disentangle subtle and intertwined effects of age, period, and cohort. For example, the effect of ageing can be estimated by simultaneously using differences between younger and older respondents in the same survey year, as well as changes that occur within respondents over time as they grow older. Parameters are estimated using generalized least squares (GLS) with standard errors that are robust to heteroscedasticity and clustering within respondents. Respondents' gender and educational level are included in the models as control variables.
Because age, period and cohort have a perfect multicollinearity (i.e., period = cohort + age), APC-models cannot be identified without imposing certain constraints. In this study, we identify the models by imposing a theoretically informed functional form on the effects of age and cohort (Kritzer 1983). Specifically, age was specified based on life phases: late adolescence (18-21), early adulthood (22-29), middle adulthood (30-65), and late adulthood (65 +). Following a similar approach as Down and Wilson (2013), we specify cohorts based on the phase of European integration during which respondents reached the age of 18. People typically form their political attitudes between about age 12 and 25, with the most impressionable years around the age of 18 (Rekker et al. 2019). We distinguish between three historic phases of European unification during which respondents had reached this age: the period of the 'European Coal and Steal Community' (ECSC), the period of the 'European Community' (EC) that started when the Brussels treaty went into effect in 1967, and the 'European Union' (EU) period that started when the Maastricht treaty took effect in 1993. Whereas the ECSC and EC where predominantly economic projects, the EU expanded this cooperation to other policy areas. More generally, cultural issues related to globalization also became much more politicized during this phase of European integration. We therefore expect that citizens who came of age after 1992 associate European integration more with cultural issues and less with economic issues compared to both earlier cohorts. Although we do not hypothesize differences between the ECSC-and the EC-cohort, we nonetheless distinguish between these cohorts to obtain a more complete picture of generational differences in the correlates of support for European unification. Because we impose these constraints on the functional form of age and cohort, period effects can be estimated freely.

Results
The results of our regression models are displayed in Table 1 and depicted graphically in Fig. 1. Rejecting our first hypothesis (H1), Model 1 reveals no generational differences in the association between support for European unification and attitudes on redistribution. Our second hypothesis (H2) was however supported, because voters who came of age after 1992 indeed revealed a stronger association than earlier cohorts between EU support and socio-cultural attitudes on immigration. With regard to period effects, the results did not support our hypothesis that attitudes towards socio-economic policies would have become less strongly related to support for European unification. Instead, the time trend in Fig. 1 is rather flat with, interestingly, a small uptick during the years of the eurozone crisis. The results however provide strong support for our hypothesis (H4) that attitudes towards socio-cultural policies would have become more strongly related to support for European unification. Between 2007 and 2019, the standardized effect of immigration attitudes on support for European unification surged from r = 0.08 to 0.21.
The hypothesis on the potential divergence or convergence of generations over time was tested in a separate model (Model 2) that replaced the year dummies with a single linear term for period. This adjustment was necessary because modelling three-way interactions of 11 period dummies with attitudes and generations would have resulted in an overparameterized model. This model provided no support for either divergence (H5a) or for convergence (H5b) of generations over time. Although a non-significant pattern cannot be interpreted, Fig. 1 however hints at an over-time convergence for the effect of socio-cultural attitudes on support for European unification (p = 0.155 for the three-way interaction of the post-1992 cohort with immigration attitudes and period), which would be in line with the idea that young voters are a vanguard of political change that is eventually followed by older citizens. Our theoretical framework did not hypothesize any life-cycle effects because we see no obvious theoretical reason why people would start to associate the EU with different issues merely as a result of growing older. Indeed, our models revealed no significant effects for ageing.

Discussion
This research note examined generational differences in the types of issues that are associated with support for European unification. The results provided support for our main idea that citizens who came of age after 1992 would base their EU support more on socio-cultural attitudes such as immigration, because during their 'impressionable years' the EU was seen as having more responsibilities for policies in this field than in previous periods. However, the analyses challenged our expectation Note. Gender, education, and main effects for age, period, and cohort are omitted from the table. *p < .05; **p < .01; ***p < .001

Fig. 1
Period and cohort differences in the association of Euroscepticism with attitudes on redistribution and immigration that older generations would show a stronger connection between EU support and socio-economic issues, because these citizens grew up in the period of the ECSC or EC when European integration was still mainly an economic endeavor. One obvious explanation could be that socio-economic policy considerations remain to be important for attitudes towards European unification. After all, many of the issues that the EU decides upon pertain to economic policies, as has become highly salient to the European public in the context of the financial crisis (see e.g., De Vries 2017; Hutter and Kriesi 2019). Another explanation could be that the impressionable years model is less applicable for socio-economic than for cultural attitudes. Indeed, this finding is consistent with earlier studies that economic attitudes develop at a later age than cultural orientations and remain more open to change (Rekker et al. 2015).
The period effects in our study revealed that the association between EU support and socio-cultural attitudes on immigration indeed increased sharply between 2007 and 2019. We found no support however for our hypothesis that the effect of socioeconomic attitudes on EU support would have decreased during this period. Instead, the effect of socio-economic attitudes on EU support remained remarkably constant over time, with only a small uptick during the eurozone crisis. This null-result may be explained by the fact that the association between socio-economic attitudes and EU support was already very weak during the start of our examined time period in 2007. Potentially, the hypothesized decrease in this association had already taken place before 2007. Finally, this study did not find any ageing effects or interactions between generations and period, which emphasizes the idea that generational differences remain rather constant over time.
A limitation of this study is that it focused exclusively on the Dutch case. Although this approach allowed us to leverage highly suitable panel data and obtain a fine-grained picture, we cannot be sure to what extent these findings generalize to other EU member states. We think that the general mechanism will be the same. If people grow up in a context in which debates on European unification are linked to sociocultural issues like immigration, these attitudes will strongly correlate with each other among generations who are then in their most impressionable years. Yet, the political consequences of this mechanism will differ across EU-member states. Our expectation is that EU attitudes will be strongly linked to matters of civil liberties in Poland and Hungary among the youngest generation, while socio-economic issues will be more important for people who were in their most impressionable years around 2011-2014 in the bailout countries. We believe this to be an important avenue for future research.
The overall picture that emerges from our findings is that, for Dutch voters, the EU has become strongly aligned along a socio-cultural attitude dimension during the 2010s. Even in 2007, Dutch voters moreover did not associate EU support strongly with socio-economic attitudes. This process of realignment of EU support as a cultural issue may increase even further in future decades, as older voters are replaced by younger generations who came of age in a period when European unification had already become a cultural and political project, rather than just an economic endeavor.