Structural and Ideological Voting in Age Cohorts

23
This article was downloaded by: [University of Massachusetts, Amherst] On: 16 September 2012, At: 04:37 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK West European Politics Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/fwep20 Structural and Ideological Voting in Age Cohorts Wouter van der Brug Version of record first published: 12 May 2010. To cite this article: Wouter van der Brug (2010): Structural and Ideological Voting in Age Cohorts, West European Politics, 33:3, 586-607 To link to this article: http://dx.doi.org/10.1080/01402381003654593 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms- and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub- licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Transcript of Structural and Ideological Voting in Age Cohorts

Page 1: Structural and Ideological Voting in Age Cohorts

This article was downloaded by: [University of Massachusetts, Amherst]On: 16 September 2012, At: 04:37Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

West European PoliticsPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/fwep20

Structural and Ideological Votingin Age CohortsWouter van der Brug

Version of record first published: 12 May 2010.

To cite this article: Wouter van der Brug (2010): Structural and Ideological Voting in AgeCohorts, West European Politics, 33:3, 586-607

To link to this article: http://dx.doi.org/10.1080/01402381003654593

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expresslyforbidden.

The publisher does not give any warranty express or implied or make anyrepresentation that the contents will be complete or accurate or up todate. The accuracy of any instructions, formulae, and drug doses should beindependently verified with primary sources. The publisher shall not be liablefor any loss, actions, claims, proceedings, demand, or costs or damageswhatsoever or howsoever caused arising directly or indirectly in connectionwith or arising out of the use of this material.

Page 2: Structural and Ideological Voting in Age Cohorts

Structural and Ideological Voting inAge Cohorts

WOUTER VAN DER BRUG

In most West European countries the effects of long-term determinants of the vote 7 inparticular social class, religion and left–right ideology 7 have slowly weakened sincethe late 1980s. This paper first describes differences between the EU member states inthe extent of structural and ideological voting in the period 1989–2004. It then focuseson the causes behind the changes over time. It is hypothesised that the decline ispartially caused by generational replacement (H1). More specifically, it is assumedthat structural voting is most important for the generation born before 1950, andpolitically socialised in the years of the mass party (H2) and that ideological voting ismost important for the generation born between 1950 and 1970, who were politicallysocialised after the decline of cleavages and before the fall of the Berlin Wall (H3). Totest these hypotheses, the study employs the European Elections Study 1989, 1994,1999 and 2004. The country by year combination provides data from 62 politicalcontexts. To analyse the determinants of party choice across different countries andyears, a methodology is employed that was developed by Van der Eijk and Franklin inChoosing Europe (1996). The study provides support for all three hypotheses.

Introduction

Traditionally, in many West European countries, party choice was determinedto a large extent by structural divisions in societies, most notably class andreligion. These social divisions became politicised in the party systems thatdeveloped in several West European countries around the turn of thenineteenth and twentieth century. For many voters, their party choice was to alarge extent an expression of group identification and loyalty to one’s ownsocial ‘group’. Moreover, for voters who were not always well informed aboutthe programmes of all parties, these social characteristics provided them with acue on the basis of which they could reach their decision on which party tovote for. This changed rapidly in the 1970s. It is now well established that therelation between social structure and party choice has weakened over recentdecades (e.g. Crewe and Denver 1985; Dalton et al. 1984; Franklin et al. 1992).

Correspondence Address: [email protected]

West European Politics,Vol. 33, No. 3, 586–607, May 2010

ISSN 0140-2382 Print/1743-9655 Online ª 2010 Taylor & Francis

DOI: 10.1080/01402381003654593

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 3: Structural and Ideological Voting in Age Cohorts

With the decline of cleavages, voters became more likely to rely onideology, rather than on social categories, to orient themselves on the partysystem. For voters who are often not very well informed about concretepositions of parties on substantive issues, party ideologies provideinformation short-cuts, which provide a generic cue about the maindifferences between parties. The fact that party conflicts on the most salientissues were in the 1970s and 1980s largely structured by on ideological left–right dimension (e.g. Fuchs and Klingemann 1990; Klingemann 1977; vander Eijk and Niemoller 1983), made it relatively easy for voters to orientthemselves on the party system in their country. Left–right provided a frameof reference that enabled voters, even those with little information, to cast ameaningful vote. Because of the fact that party positions on most issues arestructured by left–right (e.g. van der Brug and van Spanje 2009), the effect ofleft–right positions on party choice is a valid indicator of ideological voting(e.g. van der Eijk and Franklin 1996). When the term ‘ideological voting’ isused in this text, I refer to the effect of left–right ideology.

In the 1980s and 1990s many West European countries saw large changesin their party systems. By mobilising support on new issues, such asenvironmentalism (Green parties) and anti-immigration sentiments (farright parties), new parties have entered the political landscape, therebychanging the party systems drastically. As yet, it is uncertain whether thesenew issues are all structured as clearly by left–right as ‘older’ issues are. It isclear that the left–right dimension is still important for structuring partychoice, but in most West European countries the effect of left–right on partychoice slowly weakened during the 1990s (e.g. van der Brug et al. 2007a).

In sum, existing research suggests that long-term factors 7 structuralpositions and ideological orientations 7 have become less important instructuring the vote. The purpose of this paper is twofold. First of all, itdescribes for the EU member states for the period 1989 through 2004 theextent to which party choice is determined by voters’ structural positions(structural voting) and by their ideology (ideological voting). Are thereimportant differences between these countries, both in the extent ofstructural1 and ideological voting, as well as in the decline thereof? Arethere differences in this respect between different types of parties? Tooperationalise the extent of structural voting, the paper focuses on the twocleavages that have historically been most important in Western Europe instructuring the vote: social class and religion. To describe ideological voting,the paper focuses on the effect of left–right distance on party choice.2

Secondly, to the degree that changes occur in structural and ideologicalvoting patterns, the paper focuses on the extent to which generationalreplacement explains these changes. It will be shown that the decline can beexplained to a large extent by generational replacement, in addition tochanges over time that affect all generations. In this sense the paper updatesthe findings of Franklin (1992) and his colleagues, who studied electoralchange between the 1960s and the mid 1980s. It also refines their analyses.

Structural and Ideological Voting in Age Cohorts 587

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 4: Structural and Ideological Voting in Age Cohorts

The dependent variable analysed by Franklin et al. (1992) was thedichotomy between a vote for either a left-wing party or for a centre orright-wing party; a dichotomy which does not reflect the actual choice thatvoters make. In this study a newer approach to studying party choice acrossdifferent countries will be applied, which does not require such redefinitionsof the dependent variable (e.g. van der Brug et al. 2007a, 2007b; van der Eijket al. 2006).

The study employs the European Elections Study 1989, 1994, 1999 and2004. The necessary variables for our analyses were included in ques-tionnaires in 24 political systems in different years. The country*yearcombination provides us with data from 62 political contexts.

Electoral Change

The political cleavages that largely structured electoral processes until thelate 1970s developed by processes at the demand and at the supply side ofthe electoral ‘markets’. At the supply side, political parties claimed torepresent the interests of certain segments of society: workers, Catholics,farmers, etc. In the heyday of structural voting, all sorts of socialorganisations were linked to political mass parties. On the demand side,these social categories provided voters with a sense of group loyalty. This, inturn, made party choice in elections to a large extent an expression of groupidentification and loyalty to one’s own social ‘group’. Most electoralresearchers accept the fact that the ‘frozen’ cleavages of Lipset and Rokkan(1967) started to melt not long after these authors coined the term.According to Franklin et al. (1992: 385): ‘One thing that has become quiteapparent is that all of the countries that we have studied showed adecline . . . in the ability of social cleavages to structure individual votingchoice.’

What explains these changes in the ways in which people arrive at theirelectoral decisions? To a large extent, these changes were the consequence ofde-confessionalisation and individual emancipation, as a result of whichstructural positions became less important as determinants of party choice.Even if voters would still be voting on the basis of their structural positions,the decline of church attendance in the 1970s would lead to a steady loss forChristian Democratic parties. In response to these changes, political partiesbegan to seek support not exclusively on the basis of identification with thesocial group whose interests they claimed to represent, but also among othersocial groups in society. Voters, in turn, voted increasingly on the basis ofpolitical issues, and various scholars noted that the decline of structuralvoting went hand in hand with an increase in issue voting (e.g. Franklin1992; Rose and McAllister 1986).

Issue voting is complex. Even the politically most interested citizens donot have much detailed knowledge of the positions of parties on all kinds ofpolitical issues. Therefore, voters who want to base their vote choice on

588 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 5: Structural and Ideological Voting in Age Cohorts

substantive issues often use party ideology as a cue to decide which party tovote for. Many prior studies have shown that across various Europeansystems the behaviour of parties and voters alike has been structured largelyby a left–right dimension, which until the 1990s remained the dominantideological dimension in many of these countries (e.g. Fuchs andKlingemann 1990; Hix 1999; Klingemann et al. 1994; van der Eijk andFranklin 1996). Not all substantive differences between parties and amongvoters are captured by their positions on the left–right dimension, but inelections not all issues can play a role because it is unrealistic to assume thatthere will be a party with which one fully agrees on all substantive possibleissues. In the words of Sartori (1976: 338): ‘When the citizen speaks, he mayhave many things to say. But when he is coerced into casting a . . . vote, hemay well have to . . . vote for the party . . . perceived as closest on the left–right spectrum.’ Electoral research from the 1980s and 1990s confirms thatthis is how many people reached their electoral decisions in these years (e.g.Granberg and Holmberg 1988; Oppenhuis 1995; Schmitt 2001; van der Eijkand Franklin 1996).

However, there are reasons to expect that the left–right dimension haspartially lost its capacity to structure the behaviour of parties and voters. Animportant development is that the largest parties have converged on the left–right dimension (e.g. Pennings and Keman 2003). Now that Third WaveSocial Democrats have accepted the basic principles of free marketcapitalism, the larger political parties are less distinct in left–right terms.If there are smaller left–right distances between parties, left–right ideology isless useful as a cue for voters on which to base their choice. The meaning ofleft–right is, of course, not just in terms of socio-economic policies. Newissues, such as environmentalism and immigration, have to a large extentbecome integrated in the left–right dimension, at least at the party level.However, the new right-wing populist parties that often successfullymobilised support by politicising immigration policies, often refrain fromusing the left–right dimension to indicate their political position becausethey do not want to be seen as ‘extreme-right’. Moreover, voters’ attitudestowards immigrants are almost uncorrelated with left–right positions (e.g.van der Brug and van Spanje 2009). So, if voters want to affect immigrationpolicies by their vote, they cannot use left–right positions as a cue. Thecombination of these two developments 7 parties converging on the left–right dimension and new issues not being structured by left–right 7 leads tothe prediction that left–right ideology will lose its capacity to structurevoting behaviour.

Yet we also know that these kinds of changes do not take place overnight.People get ‘set in their ways’ as Franklin (2004: 216) put it. Certainbehavioural patterns, deep-rooted values, as well as political attitudes tendto be very stable. Therefore, ‘generational replacement is one of the maindriving forces behind social and political change’ (Hooghe 2004: 331).Aggregate political changes may be the result of new generations entering

Structural and Ideological Voting in Age Cohorts 589

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 6: Structural and Ideological Voting in Age Cohorts

the electoral arena with different attitudes, different party loyalties, differentsets of values than the older generations they replace. Therefore, studies ofpolitical change usually try to distinguish between changes that are affectingall generations, and changes which are due to generational replacement. Theconsequences of generational replacement have been explored for valuechange (e.g. Altemeyer 1996; Flanagan 1987; Inglehart 1984; Tilley 2005),electoral turnout (e.g. Blais et al. 2004; Franklin 2004; Lyons and Alexander2000; Miller 1992; Wass 2007), political trust or political support (e.g.Dalton 2004; Hooghe and Wilkenfeld 2008; van der Brug and van Praag2007), and strength of partisan attachments as well as re-alignment (e.g.Bartels 2000; Beck 1976; Carmines and Stimson 1981; Claggett 1981;Franklin and Ladner 1995; Tilley 2003). However, rather little researchexists on the extent to which determinants of party choice differ acrossgenerations. One of the few examples is the comparative study of Franklinet al. (1992), which showed that the decline of structural voting is to someextent dependent on generational replacement. However, that study is bynow rather outdated, it did not focus on changes in the effect of ideologicalvoting, and it redefined the dependent variable into a dichotomy between avote for either a left-wing party or for a centre or right-wing party, whichdoes not reflect the actual choice that voters make. The purpose of this studyis therefore to update and extend the analyses of Franklin et al. (1992) andto improve upon their approach by means of a new methodology whichdoes not requires us to redefine the dependent variable in a crude andsimplistic manner.

When studying changes over time in relation to generational replacement,three different effects can be distinguished: an age (or life cycle) effect, acohort (or generational) effect and a period effect (e.g. Glenn 2005). Ageeffects pertain to changes that are the result of growing older. In later stagesin life one normally acquires different interests and there is for instance someevidence that partisan attachments grow stronger during one’s life (e.g.Tilley 2003).

While life cycle effects refer to changes in attitudes and behaviours as aresult of ageing, cohort effects refer to the stability of attitudes andbehaviours over the course of people’s lives. People acquire certain valuesand behavioural habits in early stages in their lives, and as a result of thesesocialisation processes some of these values and behavioural habits tend tobe very stable. If these attitudes and behaviours are indeed very stable,aggregate changes would be mainly the result of generational replacement.When scholars speak of generation effects, they refer to differences betweenage cohorts that persist over a long period of time.

Period effects, finally, are changes in behaviours and attitudes that affectall generations. An example of a period effect would be that in a period ofsecularisation all generations go to church less often. A life cycle effectwould mean that, for instance, people go to church more often when theyget older. A generational cohort effect would be that the older and more

590 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 7: Structural and Ideological Voting in Age Cohorts

religious generations die, and are replaced by younger and less religiousgenerations.

The major problem in socialisation studies is that we cannot disentanglethe three effects statistically, because differences over time, differencesbetween generations and differences between age groups are largelyconfounded. The solution to this technical problem is normally found inthe combination of two strategies. The first is to define testable hypothesesthat are theoretically grounded. The second is to evaluate the empiricallyobserved patterns from each of the three perspectives.

On the basis of the results of Franklin et al. (1992) we would expect thatgenerations which were politically socialised in the age of cleavage politicsare most likely to cast their vote on the basis of identification with theirsocial group. Other generations that were politically socialised in the 1970sand 1980s may still identify strongly with ideological predispositions in left–right terms, even if these terms have become less meaningful incontemporary political communication. So, we can now formulate threehypotheses that will be tested in this study. The first hypothesis is that wewill see a decline of structural voting and ideological voting, partially as aconsequence of generational replacement (H1). The second hypothesis isthat structural voting will be found mostly among the generation that wasborn before 1950, who were politically socialised before the era of cleavagepolitics came to an end (H2). The third hypothesis is that left–right votingwill be strongest among the generation that was born between 1950 and1970, who were politically socialised before the fall of the Berlin Wall in1989 (H3).

Research Design

This study employs data from the European Elections Study 1989, 1994,1999 and 2004. The necessary variables for our analyses were included inquestionnaires in 24 political systems in different years. The country*yearcombination provides us with data from 62 political contexts. I employ thedata from the respondents in each country who answered the questionwhether they voted in the last European election.3 The total sample size is56,491 respondents, which is 911 per context on average.

There are two potential problems when using the European Electionsstudies for a cross-national and over time analysis of electoral change. Thefirst potential problem is that the surveys are conducted in the context ofEuropean elections, which have some specific dynamics that make themdifferent from national electoral contexts. However, the dependent variablein this study does not refer to European elections. Instead, it is a measure ofelectoral preferences for national parties at the time the surveys wereconducted (see below). Since the prime object of this study is to estimatetrends in the effects of long-term determinants of electoral choices (socialstructure and ideology), I do not consider it problematic that these electoral

Structural and Ideological Voting in Age Cohorts 591

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 8: Structural and Ideological Voting in Age Cohorts

preferences are not measured in the context of national election campaigns,which affect the extent to which short-term determinants of the vote areprimed. A more detailed discussion of the use of European Elections Studiesto analyse party choice in national arenas is provided by van der Brug et al.(2007b: 67–9).

A second potential problem is that the questionnaires have beenadministered by different modes in different countries and years, whichmay have led to differences in the biases in the samples. By means ofweighting the cases on the basis of demographics and voting behaviour, Ilargely correct for possible biases arising from self-selection amongrespondents. Moreover, while sampling biases can have a large effect onour ability to describe uni-variate distributions of variables, multi-variaterelationships are much less sensitive to sampling bias. Since the purpose ofthis study is to compare the strengths of relationships, the EuropeanElection Studies are well suited.

In this study, I compare the extent of structural voting – in terms of twoimportant cleavages, class and religion – and of left–right voting acrosscountries, over time and between different generations. Standard methodol-ogies for analysing party choice, such as multinomial logit, do not enable usto answer the research questions and test the hypotheses. In such approachesthe dependent variable (party choice) is a nominal variable, which reflects adifferent choice set in each context. Franklin et al. (1992) solved the problemof non-comparable choices between countries by recoding the dependentvariable in a vote for a left-wing party versus a vote for a centre or right-wingparty. These kinds of redefinitions of the dependent variable lead toestimation errors of unknown magnitude (van der Eijk et al. 2006).

Standard methodologies such as Conditional Logit or Multinomial Logitwould require me to carry out 62 separate country studies, withoutstraightforward means of systematically comparing the results between thedifferent results. Therefore, a method will be employed that was proposed byvan der Eijk and Franklin (1996). In each country respondents were asked,for each party in their political system,4 how likely it was (on a scale of 1 to10) that they would ever vote for it. These questions have been carefullydesigned to yield measures that can be interpreted as the propensity to votefor each of the parties (van der Eijk 2002; van der Eijk and Franklin 1996;van der Eijk et al. 2006). These measures can be regarded for ease ofexposition as preferences, but it turns out that the question is almostdeterministically related to party choice for the national parliament.5

Having measures of vote propensities serves many purposes, but in thispaper the most important function is that these provide a dependentvariable that is comparable across parties (from the same party system, aswell as from different party systems): the propensity to vote for a party.When the data matrix is stacked so that each voter appears as many times asthere are parties for which their utility has been measured (and othervariables have been appropriately transformed as explained below), the

592 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 9: Structural and Ideological Voting in Age Cohorts

question can be posed ‘what is it that makes a vote for a party attractive tovoters?’ We already know that voters virtually always choose to vote for theparty to which they give highest propensity to vote (see note 5). So ananswer to the question ‘what is it that makes a vote for a party attractive tovoters?’ is also an answer to the question ‘what determines which parties arevoted for?’ The use of this measure to analyse the determinants of partychoice has been validated elsewhere (Tillie 1995; van der Eijk et al. 2006).

The problem, of course, is how to define meaningful independentvariables. Social class may have a positive effect on the propensity to votefor party A (meaning that this party attracts its support disproportionatelyfrom the higher classes), and a negative effect on the propensity to vote forParty B. When inserting the variable social class in the stacked data matrix,these effects would single each other out, so that we would wrongly concludethat social class has no effect on the vote. The solution to this problem is toestimate linear regressions for each party separately, before constructing thestacked matrix. In these regressions party preferences are predicted from aset of dummies representing social class categories, and the family income ofthe respondent. The dummy variables were constructed from a questionasking the respondents: ‘If you were asked to choose one of these five namesfor your social class, which would you say you belong to – the working class,the lower middle class, the middle class, the upper middle class or the upperclass?’.6 The resulting predicted scores of these regressions, or y-hats instatistical parlance, are saved and used as the new independent variable.These y-hats are simply linear transformations of the original independentvariables, scaled according to the dependent variable, i.e. the 10-point votepropensity variables.7 Therefore, they are useful for the analysis of thestacked data matrix since they are comparable across parties and countries.The same procedure was carried out for religion, where vote propensitieswere estimated on the basis of the main effects of church attendance and bya set of dummy variables representing different religious affiliations.8 Inaddition, this procedure was used for two additional control variables: EUsupport and priorities given to issues.

In this way, independent variables were created, one at a time, party-by-party and country-by-country, which were included in a stacked data matrixin which the dependent variable is composed of party preferences for allparties across all countries. That is the dataset employed in this study.Datasets such as ours present certain problems of analysis. The main one isa lack of independence between the vote propensity scores given by the sameperson to different parties. The appendix discusses how these problems werehandled, and how the data in each analysis were weighted.

Results

Since the dependent variable, propensity to vote for parties, is measured ona 10-point scale, I employed linear regressions to estimate the effects of

Structural and Ideological Voting in Age Cohorts 593

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 10: Structural and Ideological Voting in Age Cohorts

social class and religion. In order to get an idea of the kind of informationproduced by this approach, Table 1 presents the results of regressions forseparate parties in three different EU member states: Germany, Britain andPoland. In addition, it shows the results of the regressions for all parties ineach of these countries analysed together. It is important to stress that as aresult of the linear transformations of the independent variables social classand religion, the regression coefficients of these variables are necessarilypositively signed. Left–right distance is expected to have a negative effect:the greater the distance between a respondent and a party, the lower thepropensity to support that party.

The exemplary results for these countries show that the effect of left–rightdistance is stronger than the effects of social class and religion in theregressions of 17 parties. Only in the case of the German far right partyRepublikaner is the effect of religion stronger than the effect of ideology.9

Overall, the explained variances of these models are rather low. InGermany, 21 per cent of the variances in party support can be explainedby long-term factors such as class, religion and left–right. In Britain this is18 per cent and in Poland only 14 per cent. The lower proportions ofexplained variance are mainly the consequence of the somewhat weakereffect of left–right ideology in Poland.

TABLE 1

IDEOLOGICAL AND STRUCTURAL VOTING IN BRITAIN, POLAND AND

GERMANY 2004

Country Party

Effect

religion

(beta)

Effect

social class

(beta)

Effect left–right

distance (beta) R2adjust.

Germany CDU/CSU 0.22 0.11 70.41 0.25SPD 0.17 0.10 70.37 0.18B 90 / Grunen 0.12 0.07 70.38 0.17FDP 0.26 0.11 70.27 0.18PDS 0.19 0.19 70.21 0.11Republikaner 0.28 0.16 70.22 0.15All 6 parties 0.19 0.10 70.40 0.21

Britain Labour 0.08 0.11 70.35 0.15Conservatives 0.13 0.18 70.42 0.25Liberal Democrats 0.05 0.08 70.35 0.14UK Independence Party 0.07 0.08 70.42 0.19All 4 parties 0.08 0.13 70.37 0.17

Poland League of P. Families 0.19 0.11 70.31 0.17Polish Peasant Party 0.14 0.05 70.26 0.10Justice and Law 0.08 0.16 70.29 0.12Civic Platform 0.08 0.23 70.26 0.14Self-Defence 0.17 0.24 70.24 0.16Social Democracy of Poland 0.08 0.13 70.36 0.16Democratic Left Alliance 0.05 0.08 70.44 0.21Labour Party 0.04 0.10 70.32 0.11Freedom Union 0.15 0.23 70.24 0.16All 9 parties 0.11 0.16 70.31 0.14

*All effects are significant at p5 0.05 (two-tailed).

594 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 11: Structural and Ideological Voting in Age Cohorts

In general, social class has somewhat more effect than religion on the votein Britain and in Poland, whereas the reverse is the case in Germany. ForPoland, this may seem counterintuitive, because one would expect religionto be important for most Polish citizens. The explanation for the minoreffect of religion on the vote in Poland is that there is little variation inreligious denomination as well as in church attendance: 95 per cent of therespondents are Roman Catholic, 3 per cent are not religious and 2 per centhave another denomination, while 57 per cent go to church at least once aweek. Only 10 per cent say that they never go to church. If there is littlevariation in religion, it cannot explain variation in electoral preferences forparties.

Table 1 also shows some interesting differences between the parties, suchas the small effect of social class on the support for Labour in Britain. Oneof the causes of its electoral success is that it mobilised support across classcleavages, which is clearly shown by this result. Table 2 takes a moresystematic view on differences between different kinds of parties. If we selectall Christian Democratic parties over this 15-year period, what is the effectof religion, social class and left–right distance? How about parties fromother party families?

The analyses presented in Table 2 demonstrate that in the period between1989 and 2004 left–right distance had a substantially stronger effect on partypreferences than religion and social class. This is true of parties from allparty families. Left–right is a particularly strong determinant of support forSocial Democratic and Conservative parties, whereas it has a substantiallyweaker effect on support for Christian Democratic and Green parties. Theeffects of social class and religion are rather weak, except for the effect ofreligion for Christian Democratic parties. Table 2 also demonstrates thatsocial class is slightly more important than religion as a determinant ofsupport for Liberal and Social Democratic parties and that these structuraldeterminants are hardly important for Green parties and even less for farright parties.

How about differences between countries and developments over time? Iwill now focus separately on social structure and on left–right. Table 3focuses on the effects of social class and religion. It summarises regressions

TABLE 2

STRUCTURAL VOTING AND IDEOLOGICAL VOTING BY PARTY FAMILY

Effect religion

(beta)

Effect social

class (beta)

Effect left–right

distance (beta) R2-adjusted

Social Democratic parties 0.11 0.14 70.45 0.26Christian Democratic parties 0.24 0.11 70.33 0.22Liberal parties 0.12 0.17 70.38 0.21Conservative parties 0.10 0.11 70.51 0.32Green parties 0.12 0.15 70.34 0.16Far right parties 0.10 0.11 70.39 0.19

Structural and Ideological Voting in Age Cohorts 595

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 12: Structural and Ideological Voting in Age Cohorts

for all 62 contexts separately in terms of the proportions of explainedvariance. It also presents these results for specific groups of countries. Thetable shows a general decrease in the extent of structural voting between1989 and 1999, with some minor exceptions. However, on average the extentof structural voting is somewhat stronger in 2004 than it was in 1999, whichshows that the extent of structural voting is not only determined bystructural developments, but that it is also contingent upon specifichistorical circumstances and actions of political parties, at least to someextent.

The extent of structural voting in the new member states, many of whichhave a recent history of Communist rule, is very similar to that in the othercountries that have had rather stable party systems for almost six decades ormore. This suggests that structural voting is not only determined by

TABLE 3

R2s OF STRUCTURAL MODEL TO PREDICT PARTY PREFERENCES (BY YEAR AND

COUNTRY)

1989 1994 1999 2004

Belgium: Flanders 0.116 0.065 0.075Belgium: Wallonia 0.134 0.103 0.058Britain 0.050 0.051 0.017 0.033Denmark 0.103 0.068 0.035 0.046France 0.086 0.057 0.048 0.060Germany 0.083 0.033 0.031 0.053Greece 0.049 0.035 0.033 0.041Ireland 0.063 0.037 0.033 0.053Italy 0.093 0.047 0.017 0.038Luxembourg 0.065 0.042 0.082Netherlands 0.088 0.098 0.065 0.076Portugal 0.051 0.042 0.032 0.026Spain 0.103 0.053 0.042 0.071Austria 0.045 0.045Finland 0.062 0.079Sweden 0.092Cyprus 0.040Czech Republic 0.057Estonia 0.043Hungary 0.019Latvia 0.100Poland 0.046Slovakia 0.041Slovenia 0.089‘Core’ 10 systems* 0.076 0.053 0.037 0.050‘Core’ 13 systems** 0.083 0.058 0.045New members 1999*** 0.066 0.062New members 2004**** 0.055

*Those countries for which data are available in each year (Britain, Denmark, France,Germany, Greece, Ireland, Italy, the Netherlands, Portugal and Spain).

**Same as core 10, but with Flanders, Wallonia and Luxembourg added.

***Austria, Finland and Sweden (except Sweden in 2004).

****Cyprus, Czech Republic, Estonia, Hungary, Latvia, Poland, Slovakia and Slovenia.

596 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 13: Structural and Ideological Voting in Age Cohorts

socialisation, but that people learn about the class and religious orientationsof political parties. In general, though, it should be noted that socialpositions structure party choice only to a limited extent in all of thesepolitical systems.

How about the effects of left–right distances? Table 4 shows that the effectof left–right distance declined in most countries between 1989 and 1999.However, in 2004 it was a bit stronger again than it was in 1999. Acontinuous decline between 1989 and 2004 can only be observed in France,Germany, Ireland and the Netherlands. However, in many countries left–right was a stronger predictor of party preferences in 2004 than it had beenin 1999: Britain, Denmark, Greece, Italy, Spain, Austria and Finland. Atentative explanation could be that the responses to the threat of terrorism,in particular the discussions about the wars in Iraq and Afghanistan,

TABLE 4

R2s OF IDEOLOGICAL VOTING MODEL TO PREDICT PARTY PREFERENCES (BY

YEAR AND COUNTRY)

1989 1994 1999 2004

Belgium: Flanders 0.085 0.087 0.190Belgium: Wallonia 0.188 0.231 0.153Britain 0.185 0.141 0.086 0.153Denmark 0.313 0.211 0.179 0.300France 0.322 0.248 0.238 0.206Germany 0.306 0.254 0.189 0.167Greece 0.342 0.329 0.203 0.223Ireland 0.170 0.153 0.113 0.113Italy 0.185 0.282 0.167 0.238Luxembourg 0.188 0.191 0.166Netherlands 0.208 0.201 0.193 0.173Portugal 0.219 0.263 0.293 0.166Spain 0.227 0.198 0.208 0.337Austria 0.131 0.190Finland 0.144 0.168Sweden 0.264Cyprus 0.372Czech Republic 0.204Estonia 0.090Hungary 0.211Latvia 0.134Poland 0.105Slovakia 0.129Slovenia 0.143‘Core’ 10 systems* 0.247 0.229 0.187 0.202‘Core’ 13 systems** 0.228 0.215 0.183New members 1999*** 0.180 0.179New members 2004**** 0.179

*Those countries for which data are available in each year (Britain, Denmark, France,Germany, Greece, Ireland, Italy, the Netherlands, Portugal and Spain).

**Same as core 10, but with Flanders, Wallonia and Luxembourg added.

***Austria, Finland and Sweden (except Sweden in 2004).

****Cyprus, Czech Republic, Estonia, Hungary, Latvia, Poland, Slovakia and Slovenia.

Structural and Ideological Voting in Age Cohorts 597

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 14: Structural and Ideological Voting in Age Cohorts

increased the ideological differences between the left and the right, as aresult of which these ideological distances became more salient to voters.This explanation seems quite plausible in Spain and Italy, but not in Britain.Understanding the details of the different patterns in the various countrieswould require an analysis of the specific historical circumstances in each ofthese countries, which is clearly beyond the scope of this study. For thepurpose of this study, the most important conclusion from Table 4 is thatacross most of these countries ideological voting has decreased since 1989,but the pattern is not linear and is different in different contexts.

Having shown the impact of class, religion and left–right ideology on thevote in almost all EU countries between 1989 and 2004, I now turn to thetests of our three hypotheses, beginning by an analysis of the effects of socialstructure in terms of class and religion. The decreases in structural voting inTable 3 may to some extent be the result of generational replacement. It ispossible that social structure would be inconsequential to young genera-tions, but that these remain significant for the older generations (H1). Thesestructural positions are expected to be particularly relevant to thegenerations born before 1950 (H2). In that case, parties would attract theirsupport across all segments of society within the younger generations,whereas they would attract only specific social and religious groups amongthe older generations. In order to assess to what extent cohort effects explain(partially) the decrease in structural voting, separate analyses wereconducted for different age cohorts. Those ten countries were selected forwhich data are available for all four elections.

Since these analyses are conducted on a stacked data matrix, one could inprinciple create dummy variables for the different election years andgenerations, and create interactions with class, religion and left–right toestimate period effects or cohort effects. This strategy will be followed forestimating cohort effects for the variable left–right distance. Since thepredictor variables class and religion were transformed 7 in order toinclude them in the stacked matrix 7 these variables can, unfortunately, nolonger be used for creating interactions. I will therefore follow a differentresearch strategy.

Since European elections are organised every five years, birth cohortswere defined also in five-year intervals. As a result, these birth cohorts canbe traced over time. Even though these are different respondents, they are arandom sample of the same generational cohort. Figure 1 shows for each ofthe birth cohort how strongly their party preferences can be predicted bysocial structure (in terms of social class and religion). Three kinds of changescould conceivably be seen in Figure 1: a generational effect, a life cycleeffect, and a period effect. If structural voting would increase with age, thiswould be indicative of a life cycle effect. In that case, each of the lines wouldbe going up from left to right. This is clearly not the case. Evidence of periodeffects can be found when the patterns within each generation are similar.Even though the patterns are not identical, we can see that the explained

598 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 15: Structural and Ideological Voting in Age Cohorts

variance decreases in most generations between 1989 and 1999 and thenincreases again somewhat in 2004. So, Figure 1 certainly provides evidenceof period effects.

If we compare between each of the lines we see the generationaldifferences. Here we find a very clear and systematic pattern, namely that theolder generations tend to vote more strongly on the basis of their socialposition than younger generations, which supports H1. H2 predicted thatstructural voting would occur particularly among the generation bornbefore 1950. The figure provides some support for H2. In 1989, the R2s in allage groups born before 1950 are higher than those of all age groups bornthereafter. In 1994 there is one exception: the R2 of those born between 1950and 1954 was higher than the R2s of some older age groups. In 1999 there isalso just one exception: the relatively high R2 of those born between 1965and 1969. However, in 2004 there are so many exceptions that the effect ofpolitical socialisation seems to have worn off.

The structure of the stacked data matrix provides the possibility toestimate the extent to which left–right voting changes over time and differsbetween generations, after controlling for the changes over time. To analysechange over time, three dummy variables were created for the election years1994, 1999 and 2004. The year 1989 is the base category. Interactions werecreated between these three dummies and left–right distance. Whenthese interactions are statistically significant, this means that the effect of

FIGURE 1

R2s OF STRUCTURAL MODEL TO PREDICT PARTY PREFERENCES (BY BIRTH

COHORTS)

Structural and Ideological Voting in Age Cohorts 599

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 16: Structural and Ideological Voting in Age Cohorts

left–right was significantly weaker or stronger in the respective election yearthan it was in 1989. In the same way two dummy variables were created fortwo generations: the generation born before 1950 and the generation bornbetween 1950 up until 1970. The generation that was born after 1970 is thebase category. Also here, interactions were created between these twodummies and left–right distance. In order to derive valid estimates of theinteraction effects, the main effects of the dummy variables that distinguishthe generations are also included in the model.10

Table 5 presents the results of two models. Model A is a model thatincludes only the effects of left–right distance, the dummy variables thatidentify election years and generations, and the interactions between left–right distance and these dummies. Moreover, it includes one additionalcontrol variable, party size, measured by the proportion of seats a party hasin the national parliament. Party size represents a strategic considerationthat voters may take into account. When two parties are about equallyattractive otherwise, voters tend to vote for the largest one because it standsa better chance of achieving its policy goals. Previous research has shownthat such strategic considerations are important to voters and that it isimportant to control for this variable (e.g. Oppenhuis 1995, van der Bruget al. 2000, 2007b).

TABLE 5

ESTIMATING COHORT AND PERIOD EFFECTS FOR LEFT–RIGHT DISTANCE ON

PARTY SUPPORT

Model A Model B

b SE Beta B SE Beta

Party size 4.286 0.044 0.241** 4.437 0.042 0.249**Social class 0.577 0.017 0.091**Religion 0.624 0.014 0.115**Importance of issues 0.612 0.012 0.134**EU approval 0.466 0.018 0.071**Distance on left–right 70.635 0.013 70.472** 70.489 0.012 70.364**Year 1994 0.059 0.023 0.008* 0.089 0.022 0.012**Year 1999 70.019 0.024 70.003 0.006 0.024 0.001Year 2004 0.081 0.025 0.011* 0.094 0.024 0.013**Born before 1950 70.156 0.026 70.024** 70.158 0.025 70.025**Born between 1950 and 1970 70.067 0.026 70.010 70.070 0.025 70.011*Year 19946 left–right 0.017 0.010 0.006 70.059 0.009 70.020**Year 19996 left–right 0.155 0.010 0.056** 0.040 0.010 0.014**Year 20046 left–right 0.123 0.011 0.048** 0.017 0.010 0.007Born before 19506 left–right 70.023 0.011 70.011 70.016 0.011 70.008Born between 1950and 19706 left–right

70.036 0.011 70.017* 70.032 0.011 70.015*

R2-adjusted 0.277 0.329Number of clusters (respondents) 33,290 33,290Number of units of analysis 222,512 222,512

*Significant at p5 0.01; **significant at p5 0.001.

Analyses are based only the 10 countries for which data are available in all four years.

600 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 17: Structural and Ideological Voting in Age Cohorts

Model A shows evidence of period effects, as well as of generationaleffects. To understand the interaction effects of left–right distance on the onehand and years and generations on the other, we have to compare theseinteractions with the main effect of left–right distance. This main effect isnegative: when the ideological distance increases, the propensity to supporta party decreases. The interaction effects with all years are positive. Since themain effect is negative, a positive interaction effect means that the effects ofleft–right are weaker in these years than they were in 1989. The interactioneffect for 1994 is not significant, but in 1999 and in 2004 the effect of left–right distance had significantly decreased in comparison to 1989.

Of the two interaction effects involving generations, only the interactionwith the generation born between 1950 and 1970 is statistically significant.This interaction effect is negative, which means that for this generation, theeffect of left–right is significantly (but not much) stronger than for theyoungest generation. These findings thus support both H1 (the decline inleft–right voting is partially the result of generational replacement) as well asH3 (the effect of left–right is strongest for the generation born between 1950and 1970).

In Model B the same interactions are tested, but this time with theaddition of more control variables. This affects the results for the periodeffects. When controlling for EU attitudes, priorities given to issues, classand religion and generational effects, there is no structural decline in left–right voting. However, we still find a significant interaction between left–right distance and the generation born between 1950 and 1970. So, eventhough the interaction effects are weak, the support for H1 and H3 seemsquite robust.

Conclusion and Discussion

The purpose of this paper was twofold. First of all, it provided descriptionsof the extent to which social structural positions (social class and religion)and ideology (left–right) determine party choice in the various countries ofthe EU in the period between 1989 and 2004. Differences were exploredbetween countries and between types of parties. Secondly, the study testedthree hypotheses regarding the extent to which generational replacementcontributes to changes in the extent of structural and ideological voting.

The analyses show that social class and religion exert significant effects onparty choice in all contexts. These variables also affect the electoral supportfor parties from different ideological complexions. On the other hand, theeffects of these variables are rather weak, and they have generally declinedafter 1989. This is pretty much in line with findings of Franklin et al. (1992),who predicted a different dependent variable (the vote for a left-wing party)and concluded that structural voting had declined. This study shows that thedecline has prolonged after the period of their study (until the mid 1980s).11

This study also showed that the decline is to some extent dependent upon

Structural and Ideological Voting in Age Cohorts 601

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 18: Structural and Ideological Voting in Age Cohorts

generational replacement, which is also in line with the patterns found byFranklin et al. (1992).

The decline in structural voting is not linear over time, though. In 2004the effect of social structure on the vote was somewhat stronger than in1999, which is conceivably the consequence of a stronger effect of religion onthe vote. Analysing the effects of social class and religion separately isbeyond the scope of this paper, but the stronger effect of social structure in2004 is most likely due to an increased political relevance of religion in theearly 2000s (for more details see van der Brug et al. forthcoming).

Another finding is that the effect of social structure on the vote is equallystrong in former communist countries as in more established democracies inWestern Europe. In the scholarly literature, two opposite predictions havebeen made about the effects of social structure on the vote in post-communist countries. Since Communist regimes did their best to eliminateor suppress religious and class differences, some have predicted thatstructural voting would be weak in post-communist societies (Muelemann2004; White et al. 2000). Others have argued that in countries where partysystems are new and especially where they are in flux, voters are in desperateneed of other cues which may be ‘cheaply’ provided by social and religiousdistinctions (Evans 2006; Evans and Whitefield 1993, 1998, 2000; Gijsbertsand Nieuwbeerta 2000; Shabad and Slomczynski 1999; Szelenyi et al. 1997;Tomka and Zulehner 1999). Since the effects of social structure on the voteare equally weak in both sets of countries, the analyses do not support eitherview (see also van der Brug et al. 2008).

This paper tested three hypotheses pertaining to the role of generationalreplacement in electoral changes. The first hypothesis was that we wouldfind a decline of structural voting and ideological voting, partially as aconsequence of generational replacement (H1). The second hypothesis wasthat structural voting would be found mostly among the generation that wasborn before 1950, who were politically socialised before the era of cleavagepolitics came to an end (H2). The third hypothesis was that left–right votingwould be strongest among the generation that was born between 1950 and1970, having been politically socialised after the decline of cleavage politics,but before the fall of the Berlin Wall in 1989 (H3). The study generatedsupport for all three hypotheses.

Studies of electoral change have often looked at differences betweengenerations. Most of these studies look at differences in the values differentgenerations cherish (e.g. Inglehart 1977). Franklin (2004) shows thatdifferences in turnout can to some extent be traced back to the salience ofthe first election when these generations were allowed to vote. This paperunderlines the main thrust of his argument, that people ‘get stuck in theirways’. They do not only get stuck in their ways in terms of party preferences,values, or the propensity to turn out to vote. They also get used toevaluating parties by certain criteria. For those who were socialised in theyears of the mass party, their social positions, in terms of class and religion,

602 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 19: Structural and Ideological Voting in Age Cohorts

are still more important than for younger generations. For those raised inthe years in which parties competed more in ideological terms, ideologymatters more. For the youngest generation, structural positions and left–right distance matter the least.

This raises the question how the youngest generation decides which partyto vote for. Since the long-term determinants of the vote are less importantfor them, it seems plausible that their vote will be more and more decided byshort-term considerations, such as issues, evaluations of governmentperformance and candidate evaluations. This is somewhat beyond thescope of this paper, but should be on the agenda for future research.

Having concluded all this, it must also be emphasised that the differencesbetween generations are small. Left–right distance is the strongestdeterminant of the vote in each generation and structural voting is relativelyweak in all years and all generations. Moreover, there is also a period effect:the effect of social structure on the vote has decreased somewhat within allgenerations, including the eldest. It is not plausible, therefore, thatgenerational replacement will lead to a further decrease in structural votingafter 1999. The effect of left–right may, however, decline somewhat more asa result of generational replacement. However, there are different patternsof ideological voting in different countries. In some countries ideologicalvoting has decreased, but in some countries a decrease has been followed byan increase. These differences are most likely due to changes at the supplyside. When parties compete more strongly in left–right terms, the effect ofleft–right will become stronger. Yet, this study showed that underneaththese different patterns, and underneath the various period effects, there aresome small but clear differences between different generations.

Notes

1. According to Bartolini and Mair (1990), cleavages have three components: social structure,

values structure and organisation (see also Enyedi 2008). According to them, cleavages exist

only when the three elements coincide and reinforce each other. Whenever I use the term

cleavage in this text, I refer to cleavages in the same way as defined by Bartolini and Mair

(1990: 216). However, the analyses in this paper focus only on one of the three aspects of a

cleavage: the effect of social structure on the vote. In line with Enyedi (2008), I therefore use

the term structural voting, rather than cleavage voting.

2. I realise that other structural positions may also have become important, besides social

class and religion. Also, other ideological dimensions than left–right may have become

important to voters. However, social class and religion were cleavages on the basis of

which parties have been founded in many West European countries. Also, left–right is

still the dominant dimension along which party systems are organised 7 even though

left–right has a different meaning in different historical contexts. Since class, religion and

left–right are important for structuring the supply side, they have always been important

determinants of the vote. After all, voters can only choose among the parties on offer.

For that reason, a study of changes in the extent to which social class, religion and left–

right determine the vote is important in its own right. Whether other structural positions

have become important as well, is also a relevant question, but is beyond the scope of

this paper.

Structural and Ideological Voting in Age Cohorts 603

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 20: Structural and Ideological Voting in Age Cohorts

3. We do this, even though we are not specifically interested in respondents’ vote in the

European Parliament elections, in order to be able to weight our data to the actual (known)

election outcome, enabling us to correct for most forms of sampling bias that might be

present in our different national samples.

4. In practice the parties asked about included only those with representation in the national

parliament or those which are high enough in the polls to obtain such representation.

5. The vote propensities are most strongly related to the intended vote in national elections

(the question ‘if there were elections today for the national parliament, which party would

you vote for’). About 93 per cent of the respondents would vote for the party to which they

give the highest vote propensity score.

6. I employ this subjective measure of social class, rather than categorisations based upon

profession, because this variable is available in all surveys in the same form and thus most

comparable across time.

7. These scores present problems of analysis unless they are centred around the same mean for

all parties. In practice we subtract the mean value for each party, turning all of them into

deviations from zero.

8. I previously checked whether the inclusion of interactions between religious affiliations and

church attendance would add substantially to the explained variance, but did not find this

to be the case. By adding many new predictors to these regressions that each exert a minor

effect, this procedure runs the risk of capitalising on chance. Therefore, I did not include

these interactions when predicting the y-hats for religion.

9. The original effect is negative: frequent church-goers are less likely to support the

Republikaner than those who never go to church. As a consequence of the linear

transformation of the independent variable ‘religion’, the direction of the effect (positive or

negative) is no longer visible.

10. I checked for multicollinearity. The tolerance of the interaction terms is never below 0.20,

and all VIF statistics of the interactions are below 5. The only somewhat problematic

estimate is the main effect of left-right distance, where the VIF is 11 and the tolerance 0.09.

However, these collinearity statistics must always be evaluated in relation to the sample

size. With a sample of more than 30,000 respondents, there is so much statistical power that

one still obtains highly reliable estimates even with these levels of multicollinearity. The low

standard errors and high significance of the effect of left–right distance bears evidence to

this.

11. The most recently available survey differed per country (between 1984 and 1987).

References

Altemeyer, Robert (1996). The Authoritarian Specter. Cambridge, MA: Harvard University

Press.

Bartels, Larry M. (2000). ‘Partisanship and Voting Behavior 1952–1996’, American Journal of

Political Science, 44:1, 35–50.

Bartolini, Stefano, and Peter Mair (1990). Identity, Competition and Electoral Availability: The

Stabilization of European Electorates, 1885–1985. New York: Cambridge University Press.

Beck, Paul A. (1974). ‘A Socialization Theory of Partisan Realignment’, in Richard G. Niemi

(ed.), The Politics of Future Citizens: New Dimensions in the Political Socialization of Children.

San Francisco: Jossey-Bass.

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.

Carmines, Edward G., and James A. Stimson (1981). ‘Issue Evolution, Population

Replacement, and Normal Partisan Change’, The American Political Science Review, 75:1,

107–18.

Claggett, William (1981). ‘Partisan Acquisition v. Partisan Identity: Life Cycle, Generation, and

Period Effects’, American Journal of Political Science, 25:2, 193–214.

604 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 21: Structural and Ideological Voting in Age Cohorts

Crewe, Ivor, and David T. Denver (1985). Electoral Change in Western Democracies. New

York: St. Martin’s.

Dalton, Russell J. (2004). Democratic Challenges, Democratic Choices. The Erosion of Political

Support in Advanced Industrial Democracies. Oxford: Oxford University Press.

Dalton, Russell J., Scott C. Flanagan, and Paul A. Beck, eds. (1984). Electoral Change in

Advanced Industrial Democracies: Realignment or Dealignment? Princeton, NJ: Princeton

University Press.

Enyedi, Zsolt (2008). ‘The Social and Attitudinal Basis of Political Parties: Cleavage Politics

Revisited’, European Review, 16:3, 287–304.

Evans, Geoffrey (2006). ‘The Social Bases of Political Divisions in Post-Communist Eastern

Europe’, Annual Review of Sociology, 32, 245–70.

Evans, Geoffrey, and Stephen Whitefield (1993). ‘Identifying the Bases of Party Competition in

Eastern Europe’, British Journal of Political Science, 23:4, 521–48.

Evans, Geoffrey, and Stephen Whitefield (1998). ‘The Structuring of Political Cleavages in Post-

Communist Societies: The Case of the Czech Republic and Slovakia’, Political Studies, 46,

115–39.

Evans, Geoffrey, and Stephen Whitefield (2000). ‘Explaining the Formation of Electoral

Cleavages in Post-Communist Democracies’, in Hans-Dieter Klingemann, Ekkehard

Mochmann, and Kenneth Newton (eds.), Elections in Central and Eastern Europe: The First

Wave. Sigma: Berlin, 36–67.

Flanagan, Scott C. (1987). ‘Value Change in Industrial Societies’, American Political Science

Review, 81:4, 1303–19.

Franklin, Mark (1992). ‘The Decline of Cleavage Politics’, in Mark Franklin, Thomas Mackie,

and Henry Valen (eds.), Electoral Change. Responses to Evolving Social and Attitudinal

Structures in Western Countries. Cambridge: Cambridge University Press.

Franklin, Mark (2004). Voter Turnout and the Dynamics of Electoral Competition in Established

Democracies since 1945. New York: Cambridge University Press.

Franklin, Mark, and Matthew Ladner (1995). ‘The Undoing of Winston Churchill:

Mobilization and Conversion in the 1945 Realignment of British Voters’, British Journal of

Political Science, 25:4, 429–52.

Franklin, Mark, Thomas Mackie, and Henry Valen (1992). Electoral Change: Responses to

Evolving Social and Attitudinal. Structures in Western Countries. Cambridge: Cambridge

University Press.

Fuchs, Dieter, and Hans-Dieter Klingemann (1990). ‘The Left–Right Scheme: Theoretical

Framework’, in M. Kent Jennings and Jan W. van Deth (eds.), Continuities in Political

Action: A Longitudinal Study of Political Orientations in Three Western Democracies. Berlin:

De Gruyter, 233–74.

Gijsberts, Merove, and Paul Nieuwbeerta (2000). ‘Class Cleavages in Party Preferences in the

New Democracies in Eastern Europe: A Comparison with Western Democracies’, European

Societies, 2, 397–430.

Glenn, Norval D. (2005). ‘Cohort Analysis (Second Edition)’, Sage University Paper Series:

Quantitative Applications in the Social Sciences 07-005. London: Sage Publications.

Granberg, Donald, and Soren Holmberg (1988). The Political System Matters. Social

Psychology and Voting Behavior in Sweden and the United States. Cambridge: Cambridge

University Press.

Hix, Simon (1999). ‘Dimensions and Alignments in European Union Politics: Cognitive

Constraints and Partisan Responses’, European Journal of Political Research, 35:1, 69–106.

Hooghe, Marc (2004). ‘Political Socialization and the Future of Politics’, Acta Politica, 39: 331–

41.

Hooghe, Marc, and Britt Wilkenfeld (2008). ‘The Stability of Political Attitudes and Behaviors

across Adolescence and Early Adulthood: A Comparison of Survey Data on Adolescents and

Young Adults in Eight Countries’, Journal of Youth Adolescence, 37:2, 155–67.

Inglehart, Ronald (1977). The Silent Revolution: Changing Values and Political Styles Among

Western Publics. Princeton: Princeton University Press.

Structural and Ideological Voting in Age Cohorts 605

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 22: Structural and Ideological Voting in Age Cohorts

Inglehart, Ronald (1984). ‘The Changing Structure of Political Cleavages in Western Society’, in

Russell J. Dalton, Scott C. Flanagan, and Paul Allen Beck (eds.), Electoral Change in

Advanced Industrial Democracies: Realignment or Dealignment? Princeton: Princeton

University Press, 25–69.

Klingemann, Hans-Dieter (1977). Ideologisches Denken in der Bevolkerung westlicher

Industriegesellschaften. Mannheim: Universitat Mannheim.

Klingemann, Hans-Dieter, Richard I. Hofferbert, and Ian Budge (1994). Parties, Policies, and

Democracy. Boulder: Westview.

Lipset, Seymour Martin, and Stein Rokkan (1967). ‘Cleavage Structures, Party Systems and

Voter alignments: an Introduction’, in Seymour Martin Lipset and Stein Rokkan (eds.),

Party Systems and Voter Alignments. Glencoe, IL: The 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’, Journal of Politics, 62:4,

1014–34.

Meulemann, Heiner (2004). ‘Enforced Secularization – Spontaneous Revival? Religious Belief,

Unbelief, Uncertainty and Indifference in East and West European Countries 1991–1998’,

European Sociological Review, 20:1, 47–61.

Miller, Warren E. (1992). ‘The Puzzle Transformed: Explaining Declining Turnout’, Political

Behavior, 14:1, 1–40.

Oppenhuis, Erik (1995). Voting Behavior in Europe. A Comparative Analysis of Electoral

Participation and Party Choice. Amsterdam: Het Spinhuis.

Pennings, Paul and Hans Keman (2003). ‘The Dutch Parliamentary Elections of 2002 and 2003:

The Rise and Decline of the Fortuyn Movement’, Acta Politica, 38, 51–68.

Rogers, William H. (1993). ‘Regression Standard Errors in Clustered Samples’, Stata Technical

Bulletin, 13, 19–23.

Rose, Richard, and Ian McAllister (1986). Voters Begin to Choose. Beverly Hills: Sage.

Sartori, Giovanni (1976). Parties and Party Systems. A Framework for Analysis. Cambridge:

Cambridge University Press.

Schmitt, Hermann (2001). ‘Zur vergleichenden Analyse des Einflusses gesellschaftlicher

Faktoren auf das Wahlverhalten: Forschungsfragen, Analysestrategien und einige Ergeb-

nisse’, in Hans-Dieter Klingemann and Max Kaase (eds.), Wahlen und Wahler. Analysen aus

Anlaß der Bundestagswahl 1998. Opladen: Westdeutscher Verlag.

Shabad, Goldie, and Kazimierz M. Slomczynski (1999). ‘Political Identities in the Initial Phase

of Systemic Transformation in Poland: A Test of the Tabula Rasa Hypothesis’, Comparative

Political Studies, 32:6, 690–723.

Szelenyi, Ivan, Eva Fodor, and Eric Hanley (1997). ‘Left Turn in Postcommunist Politics:

Bringing Class Back In?’, East European Politics and Societies, 11:1, 190–224.

Tilley, James R. (2003). ‘Party Identification in Britain: Does Length of Time in the Electorate

Affect Strength of Partisanship?’, British Journal of Political Science, 33:2, 332–44.

Tilley, James R. (2005). ‘Libertarian–Authoritarian Value Change in Britain, 1974–2001’,

Political Studies, 53:2, 442–53.

Tillie, Jean (1995). Party Utility and Voting Behaviour. Amsterdam: Het Spinhuis.

Tomka, Miklos, and Paul M. Zulehner (1999). Religion in den Reformlaendern Ost(Mittel)

Europas. Gott Nach dem Kommunismus. Schwabenverlag: Ostfildern.

Van der Brug, Wouter, and Philip van Praag (2007). ‘Erosion of Support for Political Institutions

in the Netherlands: Structural or Temporarily? A Research Note’, Acta Politica, 42:4, 443–58.

Van der Brug, Wouter, and Joost van Spanje (2009). ‘Immigration, Europe and the ‘‘New’’

Cultural Cleavage’, European Journal of Political Research, 48:3, 309–34.

Van der Brug, Wouter, Meindert Fennema, and Jean Tillie (2000). ‘Anti-immigrant Parties in

Europe: Ideological or Protest Vote’, European Journal of Political Research, 37:1, 77–102.

Van der Brug, W., Cees van der Eijk, and Mark Franklin (2007a). ‘EU Support and Party

Choice’, in Wouter van der Brug and Cees van der Eijk (eds.), European Elections and

Domestic Politics: Lessons from the Past and Scenarios for the Future. Notre Dame, IN:

University of Notre Dame Press, 168–88.

606 W. van der Brug

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012

Page 23: Structural and Ideological Voting in Age Cohorts

Van der Brug, Wouter, Cees van der Eijk, and Mark Franklin (2007b). The Economy and the

Vote: Economic Conditions and Elections in Fifteen Countries. Cambridge: Cambridge

University Press.

Van der Brug, Wouter, Mark Franklin, and Gabor Toka (2008). ‘One Electorate or Many?

Voting Behavior in New and Established Democracies in Europe’, Electoral Studies, 27:4,

589–600.

Van der Brug, Wouter, Sara Hobolt and Claes de Vreese (forthcoming). ‘Religion and Party

Choice in Europe’, West European Politics.

Van der Eijk, Cees (2002). ‘Design Issues in Electoral Research: Taking Care of (Core)

Business’, Electoral Studies, 22, 189–206.

Van der Eijk, Cees, and Broer Niemoller (1983). Electoral Change in the Netherlands. Empirical

Research and Methods of Measurement. Amsterdam: CT Press.

Van der Eijk, Cees, and Mark N. Franklin, eds. (1996). Choosing Europe? The European

Electorate and National Politics in the Face of Union. Ann Arbor: University of Michigan

Press.

Van der Eijk, Cees, Wouter van der Brug, Martin Kroh, and Mark Franklin (2006).

‘Rethinking the Dependent Variable in Voting Behavior: On the Measurement and Analysis

of Electoral Utilities’, Electoral Studies, 25, 424–47.

Wass, Hanna (2007). ‘Generations and Socialization into Electoral Participation in Finland’,

Scandinavian Political Studies, 30:1, 1–19.

White, Stephen, Bill Miller, Ase Grodeland, and Sarah Oates (2000). ‘Religion and Political

Action in Postcommunist Europe’, Political Studies, 48:4, 681–705.

Williams, Rick L. (2000). ‘A Note on Robust Variance Estimation for Cluster-Correlated

Data’, Biometrics, 56:3, 645–46.

Appendix

The stacked matrix, combining party preferences for 429 parties from 62political systems, has a total of 288,607 units of analysis, after deletion ofmissing cases. To estimate the parameters of the regression models, units ofanalyses are weighted in two steps. As a result of the weight factor applied inthe first step, respondents in each system are weighted in such a way thattheir party choice in the European Elections reflect exactly the actualelection results. In the second step this weight variable is multiplied by a(different) constant for each system, so that the 62 systems in the stackedmatrix contain the same number of cases.

Because we stacked the data, the unit of analysis is no longer theindividual respondent, but the respondent/party combination. Since theseare not independent observations, panel corrected standard errors werecomputed. Statistical significance is reported on the basis of these tests. Tobe precise, these analyses were estimated in STATA, using the robustestimate of variance (known as the Huber/White/Sandwich estimate ofvariance) and the ‘cluster’ option to adjust for the dependency amongobservations pertaining to the same respondent (Rogers 1993; Williams2000). Each of the 44,343 respondents was defined as a separate cluster.

Structural and Ideological Voting in Age Cohorts 607

Dow

nloa

ded

by [

Uni

vers

ity o

f M

assa

chus

etts

, Am

hers

t] a

t 04:

37 1

6 Se

ptem

ber

2012