Dealignment, realignment and generational differences in The Netherlands Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=fwep20 West European Politics ISSN: 0140-2382 (Print) 1743-9655 (Online) Journal homepage: https://www.tandfonline.com/loi/fwep20 Dealignment, realignment and generational differences in The Netherlands Wouter van der Brug & Roderik Rekker To cite this article: Wouter van der Brug & Roderik Rekker (2021) Dealignment, realignment and generational differences in The Netherlands, West European Politics, 44:4, 776-801, DOI: 10.1080/01402382.2020.1774203 To link to this article: https://doi.org/10.1080/01402382.2020.1774203 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. View supplementary material Published online: 17 Jun 2020. Submit your article to this journal Article views: 1289 View related articles View Crossmark data Citing articles: 1 View citing articles https://www.tandfonline.com/action/journalInformation?journalCode=fwep20 https://www.tandfonline.com/loi/fwep20 https://www.tandfonline.com/action/showCitFormats?doi=10.1080/01402382.2020.1774203 https://doi.org/10.1080/01402382.2020.1774203 https://www.tandfonline.com/doi/suppl/10.1080/01402382.2020.1774203 https://www.tandfonline.com/doi/suppl/10.1080/01402382.2020.1774203 https://www.tandfonline.com/action/authorSubmission?journalCode=fwep20&show=instructions https://www.tandfonline.com/action/authorSubmission?journalCode=fwep20&show=instructions https://www.tandfonline.com/doi/mlt/10.1080/01402382.2020.1774203 https://www.tandfonline.com/doi/mlt/10.1080/01402382.2020.1774203 http://crossmark.crossref.org/dialog/?doi=10.1080/01402382.2020.1774203&domain=pdf&date_stamp=2020-06-17 http://crossmark.crossref.org/dialog/?doi=10.1080/01402382.2020.1774203&domain=pdf&date_stamp=2020-06-17 https://www.tandfonline.com/doi/citedby/10.1080/01402382.2020.1774203#tabModule https://www.tandfonline.com/doi/citedby/10.1080/01402382.2020.1774203#tabModule Dealignment, realignment and generational differences in The Netherlands Wouter van der Bruga and Roderik Rekkera,b aDepartment of Political Science, University of Amsterdam, Amsterdam, The Netherlands; bDepartment of Political Science, University of Gothenburg, Gothenburg, Sweden ABSTRACT A central question in Western European electoral research is whether elect- oral changes over the past decades should be interpreted in terms of dealignment or realignment. Although many scholars study this question, they have not paid much attention to the role of generational replacement. This ‘age-period-cohort’ (APC) study fills this void by examining the last 10 national elections in the Netherlands (1986–2017). The hypothesis is tested that the determinants of party choice differ systematically across generations. With regard to period effects, it is found that the association between party preference and its predictors has mainly weakened. Over generations, contrar- ily no decreasing associations are found. The effects of religion, social class, partisanship, left–right and redistribution have not structurally decreased with each successive generation, whereas the effects of education, immigration and European unification are stronger for younger generations. Taken together, these findings reveal how dealignment over time can co-exist with realignment over generations. KEYWORDS Political socialisation; generations; realignment; dealignment; party system change It is a well-established fact among social scientists that ‘generational replacement is one of the main driving forces behind social and political change’ (Hooghe 2004: 331). The reason is that most people develop pat- terns of behaviour, basic values and attitudes during their late adolescence and young adulthood, which is roughly between age 15 and 23. During these so-called ‘formative years’ people are politically socialized. As people grow older, they slowly ‘get set in their ways’ (e.g. Franklin 2004). Attitudes and patterns of behaviour then become less likely to change as CONTACT Wouter van der Brug w.vanderbrug@uva.nl Supplemental data for this article can be accessed online at: https://doi.org/10.1080/01402382. 2020.1774203. � 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial- NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way. WEST EUROPEAN POLITICS 2021, VOL. 44, NO. 4, 776–801 https://doi.org/10.1080/01402382.2020.1774203 http://crossmark.crossref.org/dialog/?doi=10.1080/01402382.2020.1774203&domain=pdf&date_stamp=2021-01-06 https://doi.org/10.1080/01402382.2020.1774203 https://doi.org/10.1080/01402382.2020.1774203 http://creativecommons.org/licenses/by-nc-nd/4.0/ https://doi.org/10.1080/01402382.2020.1774203 http://www.tandfonline.com a result of new developments. To the extent that attitudes and behaviours are indeed stable, social change occurs largely as a result of older genera- tions being gradually replaced by younger ones, who are socialized in differ- ent historical periods. The insight that generations may systematically differ from each other as a result of the different periods in which they were socialized has inspired much electoral research, focusing on turnout in elec- tions (e.g. Blais et al. 2004; Dinas 2012; Franklin 2004; Grasso et al. 2019; Lyons and Alexander 2000) and on party choice (Andersen 1976; Goerres 2008; Jennings and Markus 1984; Tilley and Evans 2014). However, hardly any research has been conducted on generational differences in the determi- nants of party choice (see Van der Brug and Franklin 2018). Our study fills this void, by a study of the Dutch case (1986 until 2017). Doing so is of the utmost importance to describe and understand the nature of electoral realignment and dealignment in the Netherlands (and beyond). Realignment was defined by Dalton et al. (1984: 13) as ‘a signifi- cant shift in the group bases of party coalitions, and usually in the distri- bution of popular support among the parties as a result’. While the emphasis on group bases is clearly in line with much of the earlies litera- ture on realignment, more recent studies have broadened the concept to also include shifts in other long-term bases of electoral support, most not- ably ideological predispositions and values (e.g. Kriesi et al. 2008). We follow this way of looking at realignment and define it as a significant shift in the impact of long-term predictors of the vote. Realignment hap- pens when some long-term and stable determinants of the vote are losing their ability to create stable connections between parties and voters, and are being replaced by other stable factors that connect (groups of) voters to parties. Dealignment on the other hand means that the stable and long-term factors that used to be important as determinants of party choice (such as, social class, religion or left–right ideology) have lost their relevance and are not being replaced by other stable long-term predictors. This could mean that short-term determinants of the vote become more important, or that party preferences become less predictable. Since proc- esses of realignment and dealignment refer to the effect of long-term determinants of the vote, these are the focus of our study. The American literature on dealignment has focused mostly on the decline of the number of people who identify with a political party. Comparative scholars have focused mainly on the decline of religious and class voting; a process that is visible across Western Europe in the last quarter of the 20th Century (e.g. Franklin 1992). More recently, several scholars have argued that Western European party systems have realigned, as a consequence of globalization of world markets, European integration and mass migration (e.g. Kriesi et al. 2008). As a consequence of this WEST EUROPEAN POLITICS 777 realignment, party preferences would be based more than in the past on levels of education (e.g. Kriesi et al. 2008; Stubager 2010, 2013), and on ‘new cultural values’ that are reflected in attitudes towards European uni- fication and towards immigration (e.g. Bornschier 2010; Hutter 2014; Kriesi 2010; Van der Brug and Van Spanje 2009). Our argument is that one cannot study such long-term processes without taking into account generational differences. If political socialization plays a role in determin- ing how people choose, we should be seeing new patterns of party choice particularly among the youngest generations. If there is realignment, we should see that the party preferences of the youngest generation is mostly determined by factors related to new lines of conflict (e.g. levels of educa- tion, attitudes towards immigration and European unification). If there is dealignment, long-term predictors of the vote could still be important in determining the party choice of older generations, but not of the younger generations. So, the guiding hypothesis of our study is that electoral changes occur mostly among the youngest generations. If this is indeed the case, processes of realignment or dealignment would only become (fully) visible when younger generations have replaced older ones (whose party preferences are still structured be older cleavages). It basically means we can only observe the first stages of a process that is likely to have an increasing impact over time. To the extent that this is the case, studies that ignore these generational differences would seriously underestimate the changes that are taking place. This is why it is important to study these generational differences. When studying the role of political socialisation in electoral change, a simple comparison between generations is not sufficient. It is conceivable that young people show fewer signs of voter–party alignment than older people as a result of their age, rather than as a result of differences in the period when they were politically socialized. Voter–party alignment could become stronger with age, so that such differences might be the result of life-cycle effects, rather than generational (cohort) effects. Moreover, since the political agenda changes over time, different considerations could become more important for all generations and age groups. So, even though we are theoretically mostly interested in period and cohort effects, we can only properly model these by simultaneously considering all three components of change: Age, Period and Cohort (APC). Testing APC- models poses a challenge. Because of the collinearity of age, period, and birth cohort (year of the survey – age ¼ year of birth), APC-models can only be tested by introducing some restrictions. We will discuss the tech- nicalities in the methods section below. Here we simply wish to stress that, as far as we are aware, our study is the first APC-model that focuses on the determinants of party choice. 778 W. VAN DER BRUG AND R. REKKER By merging the Dutch Parliamentary Election Studies since 1986, we study patterns in party choice for a sufficiently long period of time to dis- tinguish the effects of age, period, and generation. Our study provides evi- dence for all three effects. In terms of generations, we find the largest differences between the oldest generation that was socialized during the heydays of ‘pillarization’ and the other generations. But we also find that left–right is most important for the generation that was socialized in the 1970s and 1980s. Moreover, European integration and levels of education are more important for younger generations than for older ones. We also find over-time changes that affect all generations, in particular a gradual decline in the effect of left–right. Finally, we find evidence for age effects. Redistribution of incomes is most important for those who are in the life phase when people are working. Left–right becomes more important as people age, until they reach their late 20s. So, we see evidence of both early adulthood socialization, but also of learning and adaptation later in life. Our study extends the knowledge of electoral changes in three ways. First, while it is well-established that generational replacement is ‘an engine of change’, hardly any research exists on generational differences in the determinants of party choice (Van der Brug and Franklin 2018). Second, the few existing studies on generational differences in the deter- minants of party choice are either cross-sectional (e.g. Wagner and Kritzinger 2012; Walczak et al. 2012), or they are otherwise limited in scope, as will be discussed below (e.g. Maggini 2016; Van der Brug 2010; Van der Eijk et al. 2005). None of these studies contain models which simultaneously estimate 1) stable differences between generations, 2) life- cycle effects, and 3) period effects. Third, by assessing the generational differences in determinants of party choice, we contribute to the literature on realignment and dealignment. On the one hand, we see strong pat- terns of dealignment, as the effect of many long-term predictors of the vote declines in strength among all generations and age groups. On the other hand, we also find evidence of realignment along new conflicts. The fact that there are clear generational differences in the effect of European integration and education, supports the claim of those who argued that Western European societies are realigning along a new cultural cleavage. It suggests that we are only seeing the beginning of this process of realignment, as these factors are likely to become more important when younger generations gradually replace older ones. The structure of the paper is as follows. We first provide a brief over- view of existing research on political socialization and on generational dif- ferences in electoral behaviour. We then discuss changes in the Dutch party system and the consequences thereof for different birth cohorts. We WEST EUROPEAN POLITICS 779 move on to discuss the design of our study, after which we present the results. In a final section we discuss the implications of our findings. Political socialization, cohort and age effects Research on ‘political socialization’ has shown repeatedly that people form their basic values and orientations at a relatively young age. Later in life, when citizens have developed basic values, attitudes, and behavioural habits, these are less likely to change as a result of new developments. People are most likely to be influenced by events that occur before they are ‘set in their ways’ (Franklin 2004), i.e. events that occur during their late adolescence and early adulthood. This phase in life is therefore referred to as ‘the impressionable years’, or ‘formative years’. While schol- ars disagree, for the development of political orientations, the formative years are roughly between 12 and 25. The strongest learning effects take place around the age of 18 (Bartels and Jackman 2014; Rekker et al. 2019; Schuman and Rodgers 2004). Scholars of electoral behaviour have demonstrated that important gen- erational differences exist in political participation, including turnout (e.g. Franklin 2004; Smets and Neundorf 2014), political attitudes and values (e.g. Alwin 1998; Down and Wilson 2013; Inglehart 1977; O’Grady 2019) and party choice (e.g. Dassonneville 2013; Goerres 2008; Tilley and Evans 2014). However, hardly any research exists on generational differences in the determinants of the vote, even though the same processes of political socialization can be expected to operate here. Our guiding hypothesis is that the considerations that motivate people’s party choice during their impressionable years, will continue to be important for their party choice later in life. Our reasons behind this hypothesis follow the logic of political social- ization. When people reach the age when they are allowed to vote, or per- haps a bit earlier, they will learn what the party system in their country looks like. This involves not just the names of the parties and the leaders of those parties. In order to choose between the various parties on offer, a first-time voter will want to understand what the main differences are between those parties. If the party system is structured around cleavages such as social class and religion, voters will learn that parties differ from each other in the representation of groups of citizens distinguished by these cleavages. If the main distinction between parties is in how they position themselves on a left–right dimension, voters will familiarize themselves with this structure of the party system. Their positions on a left–right dimension will then be an important determinant of their party preferences. In view of the overwhelming evidence in support of political 780 W. VAN DER BRUG AND R. REKKER socialization, we expect that the things people learn about the party sys- tem during their ‘impressionable years’ will have a lasting impact on their electoral behaviour later in life. So, voters who learn early in life to vote on the basis of specific social cleavages are expected to keep doing so later in life. People who learn to vote on the basis of left–right are expected to continue to do so later in life, even if the left–right distinction would become less important in party politics. Of course, we do not expect people to stop adapting to changing con- ditions after a certain period in their lives. When new parties enter the stage and/or when new issues are politicized, older generations of voters will adapt to these changing circumstances, at least to some degree. So, societal changes and specific events will exert an effect on all voters, irre- spective of their age or generation. We therefore will have to study period effects. Do certain determinants of party choice exert a stronger or weaker effect on the vote in different elections? This is what most studies of realignment and dealignment do in fact study. However, our main point is that older generations are expected to be less adaptive to these changes than younger ones and that we have to take into account these gener- ational differences. Moreover, studies of socialization need to take into account possible life-cycle effects. As people grow older, other things may become more important and might then exert more weight on people’s vote. We could for instance expect people to start caring more about eco- nomic security as they grow older. While we have no strong hypotheses on these matters, at the very least, these life-cycle effects need to be con- trolled for. Dealignment, realignment, and generations Our study links up to a large research tradition on dealignment. In the US context, dealignment refers mainly to a decline in the number of peo- ple identifying with parties. Party identification (PID) is not a concept that can be exported to European multi-party systems like the Netherlands, particularly because in this context PID is not the stable atti- tude it should theoretically be and because it is partially endogenous to the vote (e.g. Thomassen and Rosema 2009). While we should thus be cautious not to over-interpret the importance of PID in the Netherlands, we will report the results of PID as a predictor of party preferences. A declining relationship between electoral support and PID could be one indicator of dealignment. In Western Europe, research on dealignment emerged in the 1980s and 1990s and focused mainly on the decline of religious and class voting (Franklin 1992). This decline in class and reli- gious voting has also been reported for the Netherlands, albeit that the WEST EUROPEAN POLITICS 781 there is much election specific variation around the trend lines (e.g. Jansen 2011). A decrease in the explanatory power of these factors on the vote could be the result of three types of changes: changes in the distribu- tion of the independent variable (i.e. a decreasing number of religious citizens), changes at the supply side (new parties no longer mobilizing support on social cleavages like class or religion), and thirdly, a decline in the effect of these variables (see also Best 2011; Evans and Tilley 2012). Our empirical strategy is not geared towards disentangling these three. However, they all reflect dealignment as they all reflect a decreasing impact of these long-term predictors on the vote. The few studies that looked at generational differences in the factors determining the vote have confirmed that ‘cleavage voting’, in terms of social class or religion, is most prominent among the oldest cohorts (e.g. Franklin 1992; Maggini 2016; Van der Brug 2010). While this supports our basic expectation that political socialization matters for the mechanisms that drive party choice, these studies did not distinguish between generational and life-cycle effects. It is theoretically possible that, as people grow older, socio-struc- tural factors become more important factors in their party choice. The observed differences between cohorts could thus reflect life-cycle effects, rather than generational differences. Given the existing evidence, we would expect to find support for dealignment, by testing these two gener- ally formulated hypotheses: H1: The effects of long-term determinants of party preferences become weaker over time across all generations and age groups. H2: After controlling for age and period effects, long-term determinants of the vote have a weaker effect on party preferences among younger generations than among older ones. We also know very little about the extent to which generational differ- ences contribute to patterns of realignment. Several scholars have argued that with the decline of cleavages, electoral behaviour became increasingly structured by left–right in the 1970s and 1980s (e.g. Fuchs and Klingemann 1990; Van der Eijk and Franklin 1996; Van der Eijk et al. 2005). So, we would expect the effect of left–right to become stronger around that period in most Western European countries. However, most of these countries do not have election studies going back far enough in time to systematically analyse whether left–right positions have indeed become more important predictors of party choice in those decades. However, more recent election studies enable us to compare between dif- ferent generations. Van der Brug (2010) showed that ideological left–right distances were most important in determining the voting behaviour of the generations that were politically socialized in the 1970s and 1980s, roughly in the years between the decline of cleavages and the fall of the 782 W. VAN DER BRUG AND R. REKKER Berlin Wall. Van der Eijk et al. (2005) also distinguish between birth cohorts, but they do not find clear differences between generations in the effect of left–right on party choice. This later finding may, however, result from a classification of birth cohorts is not theoretically informed. If dif- ferent circumstances during the formative years lead to different patterns of party choice later in life, the classification of generations need to be based on information about the context in which different generations were socialized. In any case, both studies have not controlled for life-cycle effects, which makes the results of both studies still somewhat preliminary. More recent fundamental changes to European party systems are con- nected to the politicization of issues such as immigration, European unifi- cation, and environmental protection. According to several scholars these changes result from large scale socio-economic changes in post-industrial societies, which are often summarized under the general term ‘globalization’. According to the ‘realignment’ perspective, these develop- ments gave rise to a new socio-cultural political ‘cleavage’, which divides the ‘losers and winners of globalization’ (e.g. Kriesi et al. 2008). The ‘winners’ are people who benefit from the opportunities of open borders. They tend to favour further European integration, as they think that the most important problems of our time, such as global warming, require international collaboration. Moreover, they do not feel threatened by immigrants. The ‘losers’ on the other hand, are those who feel threatened both economically and culturally by the consequences of open borders, and who therefore oppose European integration and migration (e.g. Hooghe and Marks 2018). The increased relevance of issues of European integration, migration, and environmental protection has provided the opportunity for new parties to enter the party system, in particular popu- list radical right and green parties, which in turn have contributed to the further politicization of these issues (e.g. Rooduijn et al. 2016). While most scholars would probably agree that issues like immigration and European unification have become more politicized, so that their impact on vote choice has increased, we know little about generational differences in the effect of these issues on party choice. We expect social- ization during the impressionable years to remain important later in life, so that older generations vote on the basis of considerations that were important when they were young. The implication is that patterns of realignment should be most visible among the youngest generations of voters. So, we would expect ‘socio-cultural’ issues to be particularly important determinants of party choice among the generation of voters that was politically socialized in a period when these issues were politi- cized. Moreover, if there is a new cleavage between higher educated WEST EUROPEAN POLITICS 783 winners and lower educated losers of globalization (e.g. Bovens and Wille 2017; Hooghe and Marks 2018; Kriesi et al. 2008), we expect levels of education to be more clearly reflected in voting patterns of younger gen- erations than older ones. Given the existing evidence, we would expect to find support for realignment, by testing these two generally formu- lated hypotheses: H3: Over time, aspects related mostly to the ‘new cultural cleavage’ (levels of education, attitudes toward immigration, and European unification) are increasingly affecting party preferences across all generations and age groups. H4: After controlling for age and period effects, aspects related mostly to the ‘new cultural cleavage’ have a stronger effect on party preferences of younger generations than those of older ones. We found three studies that asked the question whether socio-cultural issues exerted a stronger effect among younger than among older genera- tions. Two of these use cross-sectional data to compare the strength of different predictors of party choice between generations (Wagner and Kritzinger 2012; Walczak et al. 2012). The third study combines cross-sec- tional studies from four different election years (Gougou and Mayer 2013), which does not enable the authors to distinguish between age, period, and cohort effects. All three studies show that socio-cultural issues are most prominent determinants of party preferences among the youngest gener- ation. However, the failure to distinguish between the life-cycle and cohort effects is problematic. The finding that older cohorts base their party choice more on socio-economic issues than younger cohorts may be the result of the fact that these issues were more politicized when they were in their impressionable years (a generational effect), or it could be the result of the fact that these issues become more important later in life (a life-cycle effect). So, in order to obtain a clear view of patterns of realignment, we wish to disentangle the contribution of age, period, and cohort. The Dutch party system and birth cohorts In the previous sections we developed the idea that people are socialized at a relatively young age into voting on the basis of considerations that are relevant during their impressionable years. This is a general theoret- ical idea that can be expected to apply across different countries. However, the specific predictions of this theoretical model will be context dependent. To have been politically socialized in the early 1970s in France will have different consequences than having been socialized in the same years in Czechoslovakia (under communist rule) or in Greece or Spain (under far-right military regimes). So, in order to specify what 784 W. VAN DER BRUG AND R. REKKER differences we would expect to observe between generations, we need to discuss some basic facts about the developments in the Dutch party sys- tem and the salience of various political issues. Until the mid-1960s the Netherlands was a clear example of a country with a frozen party system (Lipset and Rokkan 1967). The ‘pillarized’ Dutch party system was structured around two cleavages, religion and class, and election outcomes were very stable. The general election of 1967 was the first one in which the two largest governmental parties (the labour party PvdA and the Catholics peoples party KVP) both lost a ser- ious number of seats and a new liberal party D66 entered Parliament with 7 seats, which was unprecedented. During the second half of the 1960s there were many protests, for instance against the Vietnam war, and occupations of university buildings by students who demanded dem- ocratisation. The 1960s also mark a beginning of the secularisation of Dutch society, resulting in a rapid loss in electoral support for the three Dutch Christian parties (KVP, ARP, and CHU), that merged into the CDA in the late 1970s. In the 1970s the Dutch labour party began a strat- egy of ‘polarization’ (in left–right terms), in an unsuccessful attempt to be more attractive for Catholic workers (e.g. Van Praag 1991). Throughout the 1970s and most of the 1980s, left- and right-wing parties were highly divided on matters of redistribution, the government’s budget deficit, and the size of the public sector. So, it seems plausible to argue that genera- tions of voters who had the first opportunity to cast a vote in the turbu- lent elections of 1967, were confronted with a different party system and different patterns of party choice than generations thereafter. The polarization between parties in terms of left and right lasted until the second half of the 1980s. After the elections of 1986 the Labour party changed its leadership and under the new party leader Wim Kok, the party positioned itself at a more moderate position. The first time the party ran a more moderate campaign was in 1989, which was in the same year as the fall of the Berlin Wall. After these elections the labour party entered a new coalition that did not end prematurely. This had not happened since 1956. In the 1990s, a coalition government between social democrats (PvdA), lib- erals (D66), and conservative liberals (VVD) executed a large program of privatization of the public sector, as a result of which economic issues were very much de-politicized. In a way, one could perhaps argue that party pol- itics in the Netherlands was quite de-politicized in general, and that there were very few really contentious issues. So, the generation that was politic- ally socialized after the fall of the Berlin Wall in 1989 was socialized in a different context than the generation that was socialized before. In the 1990s the issue of immigration was sometimes debated, but it never became a very central issue on the political agenda. This changed WEST EUROPEAN POLITICS 785 after the attacks on the Twin Towers in September 2001, when Pim Fortuyn entered Dutch politics and strongly emphasized this issue in his campaign. Fortuyn was murdered just 9 days before the elections, but his party, the LPF, received 17 percent of the votes, making it the second largest Dutch party at that moment. Even though the party soon lost sup- port, mainly as a result of internal conflicts, the issue of immigration retained its prominence in Dutch politics, particularly as a result of the activities of Geert Wilders and the party which he founded in 2004, the PVV. Wilders’ most important theme is immigration, and the threat asso- ciated with the Islam in particular. Yet, he also campaigned against the European Union and the Euro. Partially as a result of these activities, European integration has become a more prominent issue in Dutch polit- ics in the 2000s. Since socio-cultural issues (particularly immigration and European integration) became highly contentious in Dutch politics since the elections of 2002, we expect there to be important differences between generations of voters who were politically socialized before 2002 and those who were socialized after that. To summarize, based on the changes in Dutch post-WW II politics, we can theoretically distinguish four different periods: � until 1967: cleavage politics � 1967–1989: polarization in left–right terms � 1989–2002: de-politicization � 2002–present: politicization of issues related to globalisation We realize that this way of classifying historical periods is somewhat stylized and that in reality the changes from one historical period to the next occurs gradually. Someone who was socialized in the late 1980s was already confronted with a less polarized party system than someone who was socialized in the mid-1970s. However, a more fine-grained classifica- tion of historical periods will lead to too much reduction of power to esti- mate the effects, particularly since we need to distinguish between generations, life-cycle, and period effects. Yet, to the extent that experien- ces in the impressionable years determine patterns of party choice later in life, we should be able to pick up those differences with this distinction between generations. Data and design We estimate generational differences, life-cycle effects, and over-time changes in the predictors of party choice on the basis of ten Dutch Parliamentary Elections Studies of 1986, 1989, 1994, 1998, 2002, 2003, 786 W. VAN DER BRUG AND R. REKKER 2006, 2010, 2012, and 2017. These contain questions regarding issues that were asked repeatedly over time (see Online Appendix 1 for a full over- view of items). One of the major challenges of such a study is that the dependent variable party choice is a different nominal variable at each election. New parties arise, old parties disappear or merge. So, this vari- able does not lend itself to an over-time comparison. Therefore, we decided to employ a different dependent variable, which is derived from a battery of items in which respondents are asked to indicate for each of the relevant parties ‘how likely is it that you will ever vote for party A, party B, etc.’. These ‘propensity to vote’ (PTV) questions are designed to measure current electoral preferences for parties and they are suitable for analysing the causal mechanisms generating party choice (for more details, see van der Eijk et al. 2006). In order to analyse these PTV-questions we generated a ‘stacked’ data matrix in which the records represent respondent�party combinations in long notation. This raises the obvious problem how to define meaningful independent variables. Respondents’ attitudes towards immigration will have a positive effect on support for some parties and a negative effect on support for other parties. When including these variables in the stacked data matrix, these effects would even each other out, so that we would wrongly conclude that these attitudes have no effect on party preferences. The solution to this problem is to estimate linear regressions for each party separately, before constructing the stacked matrix. The resulting predicted scores of these regressions, or y-hats, are saved and used as the new independent variables. These y-hats are simply linear transformations of the original independent variables, scaled according to the dependent variable, i.e. the ten-point vote propensity variables. Therefore, they are useful for the analysis of the stacked data matrix since they are compar- able across parties and countries. To deal with the fact that some parties have a larger average PTV than others, we first centred the propensity to vote score for each party around its year-specific mean. For each survey year, we additionally standardized all scores on y-hats and PTV’s on the entire stacked data matrix. This step was necessary because the y-hat method would otherwise lead to an effect of exactly 1 and an intercept of 0 for every issue in every year. By standardizing the scores, the effects however reflect standardized effects in every year that can be compared across periods. To adjust for the cluster- ing within respondents, we conducted all analyses with cluster-robust standard errors that accounted for clustering within both respondents and survey years. In order to capture the potentially non-linear effect of positions on issues or ideological scales for some parties, we estimated the y-hats by WEST EUROPEAN POLITICS 787 means of a set of dummy variables derived from these scales. In the case of religion, we predicted the y-hats on the basis of dummy variables that distinguish different denominations, as well as by citizens’ church attend- ance. The y-hats for social class and for education were created by regressing the centred probability to vote variables over a set of dummy variables that represent respectively different social class categories and different educational categories. This procedure was carried out in the same way for each of the predictors of party preferences. Because age, period and cohort have a perfect multicollinearity (i.e. age ¼ period – cohort), APC-models are not identified unless certain con- straints are imposed. In this study, we disentangled age, period, and cohort effects by imposing a theoretically informed functional form on our regression models (Kritzer 1983). This means that age was specified based on life phases: late adolescence (18–21), early adulthood (22–29), middle adulthood (30–65), and late adulthood (65þ). Likewise, cohorts were specified as theoretically informed generations. While the impres- sionable years are between 12 and 25, the peak in ‘political learning’ is at 18, as discussed above. Also, we defined four historical periods in the pre- vious section. We therefore define generations by the historical period in which the respondent turned 18: in other words, whether someone turned 18 before 1967, between 1967 and 1988, between 1989 and 2001, or between 2002 and 2017 (the most recent election study). By imposing such constraints on the functional form of age and cohort, we could esti- mate period effects freely. An overview of the variation in age, period, and cohort is displayed in Online Appendix 2. The standard errors of all analyses were adjusted for clustering within respondents (and survey years) due to the stacked data matrix. Indeed, we are not interested in the direct effects of age, period, and generations on party choice. We want to know whether the effects of pre- dictors of party choice differ across generations, periods, and age groups. So, we will test interaction effects between each predictor of party choice on the one hand and on the other hand the dummy variables for periods, age groups, and generations. Since there are ten election years, four age groups, and four generations, these are distinguished by 9 þ 3 þ 3 ¼ 15 dummy variables. So, for each predictor of party choice, we have to esti- mate its main effect and 15 interaction effects. Obviously, this places very high constraints upon our models. If we had wanted to include control variables, we would have to estimate the same 15 interactions for each control and this is not feasible. Therefore, we estimate our effects without control variables. As a consequence, we may be overestimating the effects of the different predictors of party choice somewhat. However, we are not so much interested in estimating these effects per se, but to assess how 788 W. VAN DER BRUG AND R. REKKER they vary over time, and between age groups and generations. Since period, age, and generation are by nature exogenous, we would only be drawing the wrong conclusion if the overestimation would be systematic- ally different at different moments in time or between different groups of citizens. Of course, the large number of dummy variables in every model also comes with a risk of finding significant effects just by chance (e.g. Type 1 errors). To account for this, we first performed an F-test to deter- mine the joint significance of all dummy variables for respectively age, period, and cohort. We only interpreted significant effects of individual dummy variables if this F-test revealed a significant p-value. Giger and Hug (2016) argued that the two-stage procedure involving y-hats might introduce biases in the estimated coefficients for other varia- bles (Giger and Hug 2016). By including only a single y-hat variable in each model, we could avoid this potential hazard of the research design. While our decision to restrict our analyses to one predictor at the time was borne out of statistical necessity, our results are in in line with the idea of alignment. Alignment is mainly a correlational phenomenon: ‘to what extent are parties tied to certain groups of voters?’. Generations may for example be socialized into supporting parties that match their issues attitudes during their formative years, but young citizens reversely also adopt the issue positions of their preferred party (e.g. Campbell et al. 1960; Rekker et al. 2017). Both causal directions are equally in line with the principle of alignment. Results We tested our models for eight predictors of party choice, three socio- structural variables, partisanship, and four policy scales. The three socio- structural variables are religion, social class, and education. Religion and social class represent the ‘traditional cleavages’, while according to many scholars education is crucial to the ‘new political cleavage’ between win- ners and losers of globalization. Party ID is included, because a decline in the effect of party ID would indicate a process of dealignment. The four policy scales are left–right, the issue of income differences, the issue of immigration, and the issue of European integration. The socio-structural variables, left–right self-placement and the issue of income differences are included in all Dutch election studies that also include the PTV’s ques- tions. The other two issues have been included since 1994. The results are presented in Table 1 below, and they provide evidence for the effects of all three elements of the age, period and cohort (APC)-models. Before going into the details, it is worth noting that there are few age effects. We observe negative coefficients for the late adolescents on all WEST EUROPEAN POLITICS 789 Ta b le 1. Re su lt s of re g re ss io n m od el s. M ai n ef fe ct s of ag e, p er io d an d co h or t ar e om it te d fr om th e ta b le . Re lig io n So ci al C la ss Ed uc at io n Pa rt is an sh ip Le ft -R ig h t Re d is tr ib ut io n Im m ig ra ti on Eu ro p e V ar ia b le 0. 34 (0 .0 2) �� � 0. 22 (0 .0 2) �� � 0. 17 (0 .0 1) �� � 0. 37 (0 .0 1) �� � 0. 52 (0 .0 1) �� � 0. 39 (0 .0 2) �� � 0. 17 (0 .0 2) �� � 0. 08 (0 .0 2) �� � Va ria bl e� Co ho rt (r ef ¼ Be fo re 19 67 ) 18 Be tw ee n 19 67 an d 19 89 0. 03 (0 .0 1) � �0 .0 3 (0 .0 1) �� 0. 01 (0 .0 1) �0 .0 1 (0 .0 0) 0. 03 (0 .0 1) �� � 0. 00 (0 .0 1) 0. 08 (0 .0 1) �� � 0. 03 (0 .0 1) �� 18 Be tw ee n 19 89 an d 20 00 0. 04 (0 .0 2) � �0 .0 3 (0 .0 1) 0. 04 (0 .0 2) � �0 .0 1 (0 .0 1) 0. 01 (0 .0 1) �0 .0 1 (0 .0 2) 0. 10 (0 .0 2) �� � 0. 05 (0 .0 2) �� 18 In 20 01 or la te r 0. 04 (0 .0 2) 0. 01 (0 .0 2) 0. 06 (0 .0 2) � 0. 01 (0 .0 1) 0. 00 (0 .0 2) 0. 01 (0 .0 2) 0. 10 (0 .0 3) �� � 0. 07 (0 .0 3) �� Jo in t p- va lu e .0 79 .0 02 .0 42 .0 08 < .0 01 .7 65 < .0 01 .0 15 V ar ia b le � A g e (r ef ¼ M id d le ad ul ts ) La te ad ol es ce n ts (1 8- 21 ) �0 .0 3 (0 .0 2) �0 .0 6 (0 .0 2) �� �0 .0 8 (0 .0 2) �� � �0 .0 3 (0 .0 1) �� � �0 .0 5 (0 .0 2) �� �0 .0 4 (0 .0 2) � �0 .0 3 (0 .0 2) �0 .0 1 (0 .0 3) Ea rl y ad ul ts (2 2- 29 ) �0 .0 3 (0 .0 1) �0 .0 2 (0 .0 1) �0 .0 1 (0 .0 1) �0 .0 2 (0 .0 1) �� � �0 .0 2 (0 .0 1) �0 .0 1 (0 .0 1) �0 .0 1 (0 .0 2) �0 .0 1 (0 .0 2) La te ad ul ts (6 5- 96 ) 0. 02 (0 .0 1) �0 .0 0 (0 .0 1) �0 .0 0 (0 .0 1) 0. 02 (0 .0 1) �� � �0 .0 1 (0 .0 1) �0 .0 7 (0 .0 1) �� � �0 .0 0 (0 .0 1) 0. 01 (0 .0 1) Jo in t p- va lu e .1 89 .0 22 .0 08 < .0 01 .0 13 < .0 01 .5 66 .8 16 Va ria bl e� Pe rio d (r ef ¼ 19 86 /1 99 4) 19 89 0. 02 (0 .0 2) �0 .0 1 (0 .0 2) �0 .0 1 (0 .0 2) �0 .0 6 (0 .0 1) �� � �0 .0 5 (0 .0 2) �� � �0 .0 8 (0 .0 2) �� � 19 94 �0 .0 1 (0 .0 2) �0 .0 6 (0 .0 2) �� � �0 .0 4 (0 .0 2) � �0 .0 6 (0 .0 1) �� � �0 .1 3 (0 .0 2) �� � �0 .1 1 (0 .0 2) �� � 19 98 0. 01 (0 .0 2) �0 .0 7 (0 .0 2) �� � 0. 00 (0 .0 2) �0 .0 9 (0 .0 1) �� � �0 .1 6 (0 .0 2) �� � �0 .1 4 (0 .0 2) �� � �0 .0 3 (0 .0 2) 0. 01 (0 .0 2) 20 02 0. 08 (0 .0 2) �� � �0 .0 8 (0 .0 2) �� � 0. 01 (0 .0 2) �0 .1 0 (0 .0 1) �� � �0 .0 7 (0 .0 1) �� � �0 .1 5 (0 .0 2) �� � 0. 07 (0 .0 2) �� � 0. 04 (0 .0 2) � 20 03 �0 .0 5 (0 .0 2) � �0 .0 6 (0 .0 2) �� �0 .0 1 (0 .0 2) �0 .0 3 (0 .0 1) � �0 .1 0 (0 .0 2) �� � 0. 06 (0 .0 2) �� � 0. 05 (0 .0 2) �� 20 06 �0 .0 4 (0 .0 2) � �0 .0 4 (0 .0 2) � �0 .0 5 (0 .0 2) �� � �0 .0 5 (0 .0 1) �� � �0 .1 0 (0 .0 1) �� � �0 .1 3 (0 .0 2) �� � �0 .0 2 (0 .0 2) 0. 01 (0 .0 2) 20 10 �0 .0 6 (0 .0 2) �� �0 .0 3 (0 .0 2) �0 .0 2 (0 .0 2) �0 .0 8 (0 .0 1) �� � �0 .1 2 (0 .0 1) �� � �0 .1 5 (0 .0 2) �� � 0. 04 (0 .0 2) � 0. 07 (0 .0 2) �� � 20 12 �0 .0 5 (0 .0 2) �� �0 .0 3 (0 .0 2) �0 .0 2 (0 .0 2) �0 .1 1 (0 .0 1) �� � �0 .1 3 (0 .0 2) �� � �0 .1 5 (0 .0 2) �� � �0 .0 1 (0 .0 2) 0. 07 (0 .0 2) �� � 20 17 �0 .1 0 (0 .0 2) �� � �0 .0 3 (0 .0 2) �0 .0 1 (0 .0 2) �0 .1 8 (0 .0 1) �� � �0 .1 8 (0 .0 2) �� � �0 .1 1 (0 .0 2) �� � 0. 00 (0 .0 2) 0. 11 (0 .0 2) �� � Jo in t p- va lu e < .0 01 < .0 01 < .0 01 < .0 01 < .0 01 < .0 01 < .0 01 < .0 01 M od el Re sp on d en ts 15 ,1 16 15 ,5 12 15 ,4 08 14 ,6 60 16 ,4 00 16 ,9 16 15 ,3 66 13 ,5 72 O b se rv at io n s 14 5, 04 9 14 8, 78 4 14 8, 18 6 14 7, 38 8 16 2, 58 0 16 6, 93 4 14 6, 45 9 13 0, 47 6 R2 11 .4 % 3. 4% 3. 3% 10 .4 % 19 .7 % 7. 6% 6. 9% 3. 5% 790 W. VAN DER BRUG AND R. REKKER predictors of party support. Five of these are statistically significant. This is indicative of a learning effect. At the age of 18 to 21, voters are still learning how their political attitudes and social positions can be related to party preferences. Beyond that age, few differences can be seen in the way people of different age groups relate their social positions and attitudes to their party preferences. A notable exception is partisanship, which becomes more strongly related to party choice as people grow older. This finding is much in line with Dinas (2014), who finds a similar pattern, which he explains by ‘habit formation’. Also, we find an age effect for the issue of income differences, which will be discussed below. The period and cohort effects require a more detailed discussion, to which we now turn. All age, period, and cohort differences are depicted in Figure 1 (socio-structural variables and partisanship) and Figure 2 (left–right self- placement and issue attitudes). Of the three socio-structural variables in our model, religious affilia- tions clearly have the strongest effect on party preferences. While there is a downward trend across time in the effect of religion, this variable had a stronger effect on party support in 2002, the year of the rise of the list Pim Fortuyn. This overall downward trend is in line with the expectation that religion has become a less important determinant of party choice in the Netherlands. However, we did not find the expected generational decline in the role of religion. An F-test revealed no significant gener- ational differences in the association between religion and party choice. In the past 30 years, social class has not been a strong predictor of party choice in the Netherlands. It is slightly more important for the old- est generation than for later generations, except for the youngest gener- ation for whom subjective social class is as important as it is for their grandparents. When looking at the over-time trends, we see that subject- ive social class has become less important over time, but that this trend has not continued. Actually, across different generations and age groups, there is no significant difference in the effect of social class in 1986 and 2019. Given the low percentage of explained variance, one could argue that the dealignment of the class cleavage had already taken place before 1986. Yet, we do not find evidence for a further dealignment over the past three decades. Moving to the third and final socio-structural predictor of party prefer- ences, education, the generational differences are in line with the realign- ment perspective. The effect of education becomes increasingly strong with each subsequent generation. There is no noticeable over-time trend within generations. Yet, as younger generations replace older ones, the effect of education on party preferences is likely to increase. Given the low R2 for education, we should not exaggerate the extent to which WEST EUROPEAN POLITICS 791 education structures party choice, but it certainly is more important for younger generations than for the older ones. Partisanship is the only predictor of party choice that becomes more important as people age. This is in line with findings of Dinas (2014), who interprets this as indicative of habit formation. Yet, it is interesting to see that there is also a significant difference between the group of mid- dle adults (30 to 64) and the oldest age group (65 and older). So even later in life, partisanship becomes more aligned with the actual vote. Generations do not differ in the effect of partisanship, but over time its importance decreases substantially across generations and age groups. This is indicative of dealignment. Having discussed the findings for the socio-structural predictors, we now move to political attitudes. Left–right remains to be the strongest predictor of party choice in the Netherlands. As expected, the effect of Figure 1. Age, period and cohort differences, the association of party preferences with socio-structural variables and partisanship. 792 W. VAN DER BRUG AND R. REKKER left–right is stronger among the generation that was politically socialized in the 1970s and 1980s. Yet, while statistically significant, this difference is quite small. There is also a modest life-cycle effect, which we also observe for the other scales: the effect of left–right is weaker among voters under age 22 than among the older age groups. Also, the effect of left–- right becomes weaker among all generations and age groups over time. Yet the decline in effect is certainly not linear. It is for instance particu- larly strong in 2002 and 2003. When looking at the effect of income differences, we find no evidence of generational effects. Yet, there is a life-cycle effect: the issue carries most weight for people between 22 and 65, roughly the years when one is expected to be active in the labour market. Also, the effect of redistribu- tion is strongest in 1986, after which it weakens somewhat until 1998. Since 1998 the effect is quite stable. Figure 2. Age, period and cohort differences, the association of party preferences with left-right self-placement and issue attitudes. WEST EUROPEAN POLITICS 793 For socio-cultural issues, we contrarily find clear generational effects. However, the generational patterns are different for immigration than for European unification. In the case of immigration, the main difference is between the oldest generation and the other generations. The effect is weakest for the generation that was socialized in the period of cleavage voting. The issue has a stronger effect for all other generations, but the difference between these three generations is not statistically significant. This is not in line with our theoretical expectation, because we had expected this issue to be most important for the generation that was socialized after 2002. There are also differences between the years. The issue of immigration has the strongest effect in 2002, 2003, and 2010. Unexpectedly, however, we find little evidence that the immigration issue has structurally become more important over time. Turning to European unification, we find that the issue has the stron- gest effect for the youngest generation. This is clearly in line with our prediction, as well as with the realignment perspective. This effect is grad- ually weaker for each of the older generations. In addition to that, the issue also becomes more important over time for all generations and age groups. So, as a consequence of the gradual increase in the effect of European integration, in combination with generational replacement, this issue is likely to become increasingly important over time. In order to assess whether the period and generational effects are (par- tially) driven by changes in the party system, we repeated all analyses for the two parties that were most dominant in the 1980s when our analysis starts: PvdA and CDA. These results are displayed in Online Appendix 3. Although both parties showed somewhat different results (as may be expected), the overall patterns for these individual parties is in line with the general pattern presented here. The only substantive difference is that we do not find period or generational effects for education regarding PvdA and CDA. So, the increasing importance of education over time and for new generations partially reflects changes in the party system. Online Appendix 4 finally explores additional contrasts between the sur- vey years by displaying all period effects with alternative reference catego- ries. This does not change the substantive conclusions. Conclusion and discussion As far as we are aware, this is the first study of realignment that disentan- gles the contribution of age, period, and generation to the over-time change in the determinants of party support. Indeed, research has been conducted on differences in party choice between generations, and age groups, particularly in two party systems such as the UK and US (e.g. 794 W. VAN DER BRUG AND R. REKKER Jennings and Markus 1984; Meffert et al. 2001; Tilley and Evans 2014). However, little research exists on generational differences in the predic- tors of party choice. Recent existing studies have demonstrated that socio- cultural issues are more important for younger generations of voters than older ones (Gougou and Mayer 2013; Wagner and Kritzinger 2012; Walczak et al. 2012). However, these studies did not disentangle life-cycle and generational differences. While the present study did indeed provide support for the theoretical expectation that generations differ in the consid- erations they use to evaluate parties, we found different patterns for differ- ent issues and for left–right. Moreover, we find evidence for the effects of all three components of APC-models: age, period, and cohort effects. Of the three components of the APC-model, age (or life-cycle) is the least important for patterns of party choice. The most important pattern is that all predictors exert a slightly weaker effect among the youngest age group than among older ones. Five of the eight interactions are also stat- istically significant. Apparently, it takes some voters a while to figure out how their attitudes and socio-structural positions might be linked to their party preferences. Also, young voters may need some time to understand and internalize the meaning of left and right and relate these to their party choice. Only in the case of partisanship did we find a gradual increase in the effect on party choice as people age (see also Dinas 2014; Van der Brug and Franklin 2018). We also found some interesting age differences for the effect of the issue of income differences. This issue matters the most for the two age groups who are in the life phase in which they are expected to work. In principle, this is an interesting result, which makes sense. Yet, given the fact that we had no clear hypotheses on these age differences and since we are the first to estimate them, we want to be cautious in not drawing strong conclusions until these results have been replicated in other countries. Turning to generational and period effects, the results for education and those for the issue of European unification were clearly in line with the idea that a realignment is taking place. Across different generations and age groups, the issue of European unification has gradually become increasingly important for party choice (a period effect), as predicted by H3. Also, the issue is more important for younger generations than for older ones, in line with H4. This latter is also the case for education. So, in combination, this suggests that generational replacement fuels the increasing importance of education and European unification for party choice. We had expected that social class and religion would be most relevant for voters who were politically socialized in the years of ‘frozen’ cleavages. However, we do not find support for this expectation. Religion has become less important for party choice across all generations (H1), but religion does WEST EUROPEAN POLITICS 795 not have the strongest effect for the oldest generation, as we had expected (H2). Also in the case of subjective social class, we do not find the strongest effect for the oldest generation. One explanation could be that the Dutch party system has changed over the past three decades. People who were pol- itically socialized before the elections of 1967, learned their way around a party system that has been almost entirely overhauled. Of the 13 parties that are currently represented in the Dutch parliament, only three were repre- sented in that same parliament before 1967 (the VVD, PvdA, and SGP). So, if social class and religion are important for these voters as determinants of party choice, they would have to know how each of these newly entered parties represent social groups. A more stable party system may be needed to find such stable effects of social cleavages. Left–right turns out to exert the strongest effect among the generation that was politically socialized in the 1970s and 1980s. This is in line with expectations as well as previous findings (Van der Brug 2010). Moreover, across all generations and age groups, left–right has become less import- ant in structuring the vote. The effect of left–right is not decreasing in a linear manner, however. The effect of left–right was actually particularly strong in 2002 and 2003; the years when Dutch party politics was in tur- moil as a result of the entry of the List Pim Fortuyn (LPF) to parliament and the politicization of the immigration issue. This may be explained by the fact that left–right had already become strongly correlated with atti- tudes towards immigration by this time (e.g. De Vries et al. 2013). Rekker (2016) shows that there are generational difference in how citizens inter- pret left and right. For the younger citizens it is more closely correlated with the immigration issue, while traditional socio-cultural issues (i.e. secularism and civil liberties) are more important for older generations. This might partially explain why we find that left–right is also a strong predictor of party support among younger generations. We also find generational differences in the effects of the immigration issue. This issue has become a more important driver of party choice for younger generations, although only the oldest generation differs sharply from other generations. Surprisingly, the issue of immigration does not have a gradually increasing impact on the vote over the years. Rather, the effect is already quite substantial in 1994 and then becomes temporarily more important in elections that were characterized by the breakthrough of anti-immigration parties, such as the rise of the List Pim Fortuyn in 2002 and 2003 or the success of Geert Wilders’ PVV in 2010. The issue is an important predictor of party choice for most generations, but it has not become more important since the 1990s. Another unexpected finding is that the issue of income differences is equally important for different generations. While its effect has become 796 W. VAN DER BRUG AND R. REKKER less important for all generations since 1986 (in line with H1), we find no significant differences between generations (H2). The lack of generational differences for economic issues may also be linked to findings from ado- lescent research that economic attitudes develop at a later age than socio- cultural orientations and remain more open to change (Rekker et al. 2015). By 2017, the last year included in our data, the issues of income differences and immigration exert an almost equally strong effect of the vote for most generations, except the oldest one. If we summarize the evidence, we find that the effects of the strongest predictors of party choice, partisanship, left–right, religion, and the issue of income differences, have become weaker over time for all generations. This suggests dealignment. The effects of immigration and education meanwhile have not become stronger over time. Only one the effect of attitudes on European unification on party support have increased. Overall, the changes that take place within generations over time are therefore in line with an image of dealignment between parties and voters. However, the differences that exist between generations are more in line with the idea of realignment. The effects of religion, social class, partisan- ship, left–right, and redistribution have not structurally decreased with each successive generation, whereas the effects of education, immigration, and European unification are stronger for the youngest generations than for the oldest. This raises the final question of whether these results for the Netherlands can be generalized to other countries. The general theoretical idea that the period in which people are socialized will remain to have an impact on how people form their party preferences later in life, should apply in other countries as well. However, the specific predictions regard- ing generational differences would be dependent upon the specific histor- ical circumstances in which different generations of voters were socialized in each country. The party system in the Netherlands changed fundamen- tally in 2002, so we expect differences between those socialized before and after that election year. However, for instance, the biggest change in the Italian party system was in 1994 and in Greece in 2014. The nature of these changes were also very different. So, while the theoretical idea is generalizable, the predictions would be more specific for different genera- tions in different countries. Acknowledgements We are grateful for the constructive criticism of the anonymous reviewers of West European Politics. We also received valuable suggestions when presenting earlier versions of this paper in the workshop on electoral democracy at the WEST EUROPEAN POLITICS 797 ECPR joint sessions in April 2019 in Mons and at the University of Nottingham and the University of Gothenburg. Disclosure statement No potential conflict of interest was reported by the authors. Funding This research was supported by a grant from the Dutch Research Council (NWO). Grant number: VI.Veni.191R.039. Notes on contributors Wouter van der Brug is Professor of Political Science at the University of Amsterdam. His research interests focus on comparative research in collective political behaviour, in particular electoral behaviour, right-wing populism and political parties. [w.vanderbrug@uva.nl] Roderik Rekker is a postdoctoral researcher at the University of Amsterdam and at the University of Gothenburg. He is a political scientist and a psychologist. His research interests focus on stability and change in political attitudes across time, generations and the lifespan. [r.s.b.rekker@uva.nl] References Alwin, Duane F. (1998). ‘The Political Impact of the Baby Boom: Are There Persistent Generational Differences in Political Beliefs and Behavior?’, Generations, 22:1, 46–54. Andersen, Kristi (1976). ‘Generation, Partisan Shift, and Realignment: A Glance Back to the New Deal’, in Norman H. Nie, Sidney Verba, and John R. Petrocik (eds.), The Changing American Voter. Cambridge, MA: Harvard University Press, 74–95. Bartels, Larry M., and Simon Jackman (2014). ‘A Generational Model of Political Learning’, Electoral Studies, 33:1, 7–18. Best, Robin E. (2011). ‘The Declining Electoral Relevance of Traditional Cleavage Groups’, European Political Science Review, 3:2, 279–300. Blais, Andre, Elisabeth Gidengil, Neil Nevitte, and Richard Nadeau (2004). ‘Where Does Turnout Decline Come from?’, European Journal of Political Research, 43:2, 221–36. Bornschier, Simon (2010). ‘The New Cultural Divide and the Two-Dimensional Political Space in Western Europe’, West European Politics, 33:3, 419–44. Bovens, Mark, and Anchrit Wille (2017). Diploma Democracy: The Rise of Political Meritocracy. Oxford: Oxford University Press. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes (1960). The American Voter. New York: Wiley. Dalton, Russell J., Scott C. Flanagan, and Paul A. Beck (1984). Electoral Change in Advanced Industrial Democracies: Realignment or Dealignment? Princeton: Princeton University Press. 798 W. VAN DER BRUG AND R. REKKER Dassonneville, Ruth (2013). ‘Questioning Generational Replacement. An Age, Period and Cohort Analysis of Electoral Volatility in The Netherlands, 1971–2010’, Electoral Studies, 32:1, 37–47. De Vries, Catherine E., Armen Hakhverdian, and Bram Lancee (2013). ‘The Dynamics of Voters’ Left/Right Identification: The Role of Economic and Cultural Attitudes’, Political Science Research and Methods, 1:2, 223–38. Dinas, Elias (2012). ‘The Formation of Voting Habits’, Journal of Elections, Public Opinion & Parties, 22:4, 431–56. Dinas, Elias (2014). ‘Does Choice Bring Loyalty? Electoral Participation and the Development of Party Identification’, American Journal of Political Science, 58: 2, 449–65. Down, Ian, and Carole J. Wilson (2013). ‘A Rising Generation of Europeans? Life-Cycle and Cohort Effects on Support for ‘Europe’’, European Journal of Political Research, 52:4, 431–56. Evans, Geoffrey, and James Tilley (2012). ‘How Parties Shape Class Politics: Explaining the Decline of the Class Basis of Party Support’, British Journal of Political Science, 42:1, 137–61. Franklin, Mark N. (1992). ‘The Decline of Cleavage Politics’, in Mark N. Franklin, Thomas Mackie, and Henry Valen (eds.), Electoral Change: Responses to Evolving Social and Attitudinal Structures in Western Countries. Cambridge: Cambridge University Press, 381–402. Franklin, Mark N. (2004). Voter Turnout and the Dynamics of Electoral Competition in Established Democracies since 1945. New York: Cambridge University Press. Fuchs, Dieter, and Hans-Dieter Klingemann (1990). ‘The Left–Right Schema’, in M. Kent Jennings and Jan van Deth (eds.), Continuities in Political Action: A Longitudinal Study of Political Orientations in three Western Democracies. Berlin: De Gruyter, 203–234. Giger, Nathalie, and Simon Hug (2016). ‘An Error Grave of Consequence. Empirical Models to Study Propensity to Vote for Parties’, Conference paper at ‘European Political Science Association’ (Brussels, June 23–25, 2016). Grasso, Maria T., Stephan Farrall, Emily Gray, E. Colin Hay, and Will Jennings (2019). ‘Thatcher’s Children, Blair’s Babies, Political Socialization and Trickle- down Value Change: An Age, Period and Cohort Analysis’, British Journal of Political Science, 49:1, 17–36. Goerres, Achim (2008). ‘The Grey Vote: Determinants of Older Voters’ Party Choice in Britain and West Germany’, Electoral Studies, 27:2, 285–304. Gougou, Florent, and Nonna Mayer (2013). ‘The Class Basis of Extreme Right Voting in France: Generational Replacement and the Rise of New Cultural Issues (1984–2007)’, in Jens Rydgren (ed.), Class Politics and the Radical Right. London: Routledge, 156–72. Hooghe, Liesbet, and Gary Marks (2018). ‘Cleavage Theory Meets Europe’s Crises: Lipset, Rokkan, and the Transnational Cleavage’, Journal of European Public Policy, 25:1, 109–35. Hooghe, Marc (2004). ‘Political Socialization and the Future of Politics’, Acta Politica, 39:4, 331–41. Hutter, Swen (2014). Protesting Culture and Economics in Western Europe: New Cleavages in Left and Right Politics. Minneapolis: University of Minnesota Press. WEST EUROPEAN POLITICS 799 Inglehart, Ronald (1977). The Silent Revolution: Changing Values and Political Styles in Advanced Industrial Society. Princeton: Princeton University Press. Jansen, Giedo (2011). Social Cleavages and Political Choices: Large-Scale Comparisons of Social Class, Religion and Voting Behavior in Western Democracies. Nijmegen: Radboud University Nijmegen. Jennings, M. Kent, and Gregory B. Markus (1984). ‘Partisan Orientations over the Long Haul: Results from the Three-Wave Political Socialization Panel Study’, American Political Science Review, 78:4, 1000–18. Kriesi, Hanspeter (2010). ‘Restructuration of Partisan Politics and the Emergence of a New Cleavage Based on Values’, West European Politics, 33:3, 673–85. Kriesi, Hanspeter, Edgar Grande, Romain Lachat, Martin Dolezal, Simon Bornschier, and Timotheos Frey (2008). West European Politics in the Age of Globalization. Cambridge: Cambridge University Press. Kritzer, Herbert M. (1983). ‘The Identification Problem in Cohort Analysis’, Political Methodology, 9:1, 35–50. Lipset, Seymour M., and Stein Rokkan (1967). ‘Cleavage Structures, Party Systems and Voter Alignments: An Introduction’, in Seymour M. Lipset, and Stein Rokkan (eds.), Party Systems and Voter Alignments. New York: Free Press, 1–64. Lyons, William, and Robert Alexander (2000). ‘A Tale of Two Electorates: Generational Replacement and the Decline of Voting in Presidential Elections’, The Journal of Politics, 62:4, 1014–34. Maggini, Nicola (2016). Young People’s Voting Behaviour in Europe: A Comparative Perspective. London: Palgrave Macmillan. Meffert, Michael F., Helmuth Norpoth, and Anirudh V. Ruhil (2001). ‘Realignment and Macropartisanship’, American Political Science Review, 95:4, 953–62. O’Grady, Tom (2019). Age Cleavages across European Countries and Issue Dimensions since the 1980s. Retrieved from https://doi.org/10.31235/osf.io/ r8xbg Rekker, Roderik (2016). ‘The Lasting Impact of Adolescence on Left–Right Identification: Cohort Replacement and Intracohort Change in Associations with Issue Attitudes’, Electoral Studies, 44:1, 120–31. Rekker, Roderik, Loes Keijsers, Susan Branje, and Wim Meeus (2015). ‘Political Attitudes in Adolescence and Emerging Adulthood: Developmental Changes in Mean Level, Polarization, Rank-Order Stability, and Correlates’, Journal of Adolescence, 41:2, 136–47. Rekker, Roderik, Loes Keijsers, Susan Branje, and Wim Meeus (2017). ‘The Dynamics of Political Identity and Issue Attitudes in Adolescence and Early Adulthood’, Electoral Studies, 46:2, 101–11. Rekker, Roderik, Loes Keijsers, Susan Branje, and Wim Meeus (2019). ‘The Formation of Party Preference in Adolescence and Early Adulthood: How and When Does It Occur in the Multiparty Context of The Netherlands?’, Young, 27:1, 48–68. Rooduijn, Matthijs, Wouter van der Brug, and Sarah M. de Lange (2016). ‘Expressing or Fuelling Discontent? The Relationship between Populist Voting and Political Discontent’, Electoral Studies, 43:1, 32–40. Schuman, Howard, and Willard L. Rodgers (2004). ‘Cohorts, Chronology, and Collective Memories’, Public Opinion Quarterly, 68:2, 217–54. 800 W. VAN DER BRUG AND R. REKKER https://doi.org/10.31235/osf.io/r8xbg https://doi.org/10.31235/osf.io/r8xbg Smets, Kaat, and Anja Neundorf (2014). ‘The Hierarchies of Age-Period-Cohort Research: Political Context and the Development of Generational Turnout Patterns’, Electoral Studies, 33:1, 41–27. Stubager, Rune (2010). ‘The Development of the Education Cleavage: Denmark as a Critical Case’, West European Politics, 33:3, 505–33. Stubager, Rune (2013). ‘The Changing Basis of Party Competition: Education, Authoritarian–Libertarian Values and Voting’, Government and Opposition, 48: 3, 372–97. Thomassen, Jacques J. A., and Martin Rosema (2009). ‘Party Identification Revisited’, in John Bartle and Paolo Bellucci (eds.), Political Parties and Partisanship: Social Identity and Individual Attitudes. London: Routledge, 42–59. Tilley, James, and Geoffrey Evans (2014). ‘Ageing and Generational Effects on Vote Choice: Combining Cross-Sectional and Panel Data to Estimate APC Effects’, Electoral Studies, 33:1, 19–51. Van der Brug, Wouter (2010). ‘Structural and Ideological Voting in Age Cohorts’, West European Politics, 33:3, 586–607. Van der Brug, Wouter, and Mark N. Franklin (2018). ‘Generational Replacement: Engine of Electoral Change’, in Justin Fisher, Edward Fieldhouse, Mark Franklin, Rachel Gibson, Marta Cantijoch, and Christopher Wlezien (eds.), The Routledge Handbook of Elections, Voting Behavior and Public Opinion. London: Routledge, 429–42. Van der Brug, Wouter, and Joost van Spanje (2009). ‘Immigration, Europe and the ‘New’ Cultural Dimension’, European Journal of Political Research, 48:3, 309–34. Van der Eijk, Cees, and Mark N. Franklin (1996). Choosing Europe? The European Electorate and National Politics in the Face of Union. Ann Arbor: University of Michigan Press. Van der Eijk, Cees, Hermann Schmitt, and Tanja Binder (2005). ‘Left–Right Orientations and Party Choice’, in Jacques Thomassen (ed.), The European Voter. Oxford: Oxford University Press, 167–91. Van der Eijk, Cees, Wouter van der Brug, Martin Kroh, and Mark Franklin (2006). ‘Rethinking the Dependent Variable in Electoral Behavior—On the Measurement and Analysis of Utilities’, Electoral Studies, 25:3, 424–47. Van Praag, Philip (1991). Strategie en Illusie: Elf jaar intern debat in de PvdA (1966–1977). Amsterdam: Het Spinhuis. Wagner, Markus, and Sylvia Kritzinger (2012). ‘Ideological Dimensions and Vote Choice: Age Group Differences in Austria’, Electoral Studies, 31:2, 285–96. Walczak, Agnieszka, Wouter van der Brug, and Catherine de Vries (2012). ‘Long- and Short-Term Determinants of Party Preferences: Inter-Generational Differences in Western and East Central Europe’, Electoral Studies, 31:2, 273–84. WEST EUROPEAN POLITICS 801 Abstract Political socialization, cohort and age effects Dealignment, realignment, and generations The Dutch party system and birth cohorts Data and design Results Conclusion and discussion Acknowledgements Disclosure statement References