The subjective cognitive and affective map of extreme right voters: using open-ended questions in exit polls

Electoral Studies 20 (2001) 217–241www.elsevier.com/locate/electstud

The subjective cognitive and affective map ofextreme right voters: using open-ended

questions in exit polls

Marc SwyngedouwInter-University Centre for Political Opinion Research (ISPO), K.U. Leuven, Van Evenstraat 2B, 3000

Leuven, Belgium

Abstract

This article looks to build a motivation profile for extreme right voters of theVlaams Blokin the 1995 general election. Voters’ subjective motivations are studied using an open-endedquestion from the Belgian–Flemish public broadcaster (VRT) exit-poll. After a discussion ofthe advantages and disadvantages of using open as opposed to closed questions, the differentmotives are classified on a functional basis according to existing theoretical voting behaviourmodels. HICLAS, a hierarchical classes analysis method for binary objects (parties) by features(motives) data, is used to create a subjective cognitive and affective map of the extreme rightVlaams Blok(VB) voters. The research strategy reveals the existence of motives to supportthe VB that have previously been mentioned only rarely (if at all). The results also lead usto re-evaluate the relative importance of the different motivations for vote choice (e.g. Flemishnationalism and communitarian issues, crime and law and order, political scandals, …). Fur-thermore, the results make it clear that traditional social groups do not see the VB as the partyrepresenting their interests. The reasons why voters support the VB are to a large extent differ-ent from the reasons that the VB leadership claims that they support it; the VB vote is notsolely a question of economic self-interest or fear (of crime and migrants). This has importantconsequences for solutions designed to “manage” these issues, given the rejections of partypolitics and the anti-political attitude of the extreme right voter. 2001 Elsevier Science Ltd.All rights reserved.

Keywords:Open questions; Extreme right; Motivation profiles; HICLAS

E-mail address:[email protected] (M. Swyngedouw).

0261-3794/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.PII: S0261 -3794(00 )00010-X

218 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

1. Introduction

It would seem that nobody has bothered to ask the electors themselves why theycast a vote for the extreme right. This appears to be a startling omission in view ofthe fact that at least some of these voters can explain their voting behaviour. Expla-nations for extreme right voting patterns have been sought in a variety of theoreticalmodels. Apart from referring to general social trends (the rise of “new conservatism”,the dualization of Western societies, unemployment and so on), socio-demographiccharacteristics (sex, education, occupation, etc.) and socio-psychological character-istics (authoritarianism, pessimism, etc.), these explanatory models usually point tospecific political characteristics which could explain the motivation of the extremeright voter.

We see for example Ignazi and Ysmal (1992); Ivaldi (1999) referring to the crisisof political legitimacy and the party system and the emergence of new issues, suchas “migrants” and “law and order”, which have been “ignored” by the establishedparties. Mayer and Perrineau (1992) see the extreme right voter as a non-rationalprotest voter who has little confidence in social and political processes. Electorsvoting for Le Pen in 1998 were not voting for the Front National, nor for its leader,nor for the extreme right. To quote, “It is more a vote ‘against’ things, againstimmigrants and delinquents, who are the scapegoats for their fears, against the polit-ical establishment and the parties of government” (1992: 134). Regarding the 1995presidential elections, Mayer (1996) reports that the Front National voter is extremelyconcerned with the issues of immigration and safety, that he or she is dissatisfied withthe way democracy works and is extremely critical of the political establishment. Shedistinguishes four kinds of FN voters, the extremists, the traditionalists, the populistsand the detached. The following observations are relevant to the present study. Theextremist voter is attached to the party and has a coherent political ideology. Thetraditional voter used to vote for right wing parties and is against sexual freedom,is against “big” government, and opposes radical change. The populists used to votefor the left and are those most interested in defending the social and economic rightsand privileges of labour. The politically detached have much in common with thepopulists, but are minimally involved with democracy. All these voters share a great“dislike of other (candidates)”. Mu¨ller-Rommel (1995: 5) points to the commitmentof extreme right parties to radically reforming the political system in the short term.Kitschelt (1995) refers to the attractions of the racist, xenophobic and authoritarianmessage, the call to market liberalism, the cynicism regarding established partiesand the desire to “teach them a lesson” and the promise to apply social protectionexclusively to a single clearly defined ethnic group. Betz (1994) mentions the popu-list resentment towards the established political parties, who are generally held tobe responsible for everything which can or does go wrong in society. Swyngedouw(1995, 1998) stresses the importance of the particularist world view aimed at preserv-ing a (long-gone) status quo.

All these analyses implicitly offer motives for the voting behaviour of the extremeright voter. For example some extreme right voters opt for these parties because ofissues such as migrants and safety (law and order); as a protest against the policy

219M. Swyngedouw / Electoral Studies 20 (2001) 217–241

or the absence of policy in this respect; for anti-political reasons (against the politicalestablishment); because they identify with the parties of the extreme right and theirgeneral (authoritarian) ideology; for their traditional conservative views; becausethey defend the interests of labour; or because they want radical reforms of thesystem or precisely because they want to preserve the status quo.

In this article we wish to examine to what extent the motives attributed to votersare actually mentioned by the extreme right voters themselves. The aim is to establisha profile for the extreme right voters, as distinct from the extreme right militantsand the extreme right party ideology (Fennema, 1997).

2. The elections of 21 May 1995 in Belgium–Flanders1

The elections of 21 May 1995 confirmed the political situation which had beenestablished four years before (electoral results of 24 November 1991). Nonethelessthe apparent tranquillity was deceptive in the extreme. Analyses of the shifts basedon the results of the VRT-ISPO Exit Poll2 indicated that roughly 32% of voters hadchanged their voting behaviour in the 1991–1995 period. Contrary to the pre-1991period, when the floating voter flocked to theVlaams Blok(VB), it appeared in1995 that the very considerable numbers of voters voting for another party tendedto neutralize one another. Table 1 reproduces the electoral results for the nationalelections in Flanders without the 19 municipalities of Brussels for the Chamber ofRepresentatives for 1987, 1991 and 1995.

The 1995 national elections had nonetheless promised an upset. Most vulnerablewas the Socialist Party (SP). This party was particularly worried that it would takea beating at the polls. Apart from the continued desertion of its voters to the extremeright VB, the party was involved in the “Agusta scandal” (a criminal investigationof alleged bribes paid to politicians to secure a helicopter contract for Italian firmAgusta). Moreover the credibility of the socialist family was very much in questionbecause of the events which had overtaken its French language sister party, thePartiSocialiste(PS). The unsolved murder of former socialist leader Andre´ Cools andseveral alleged cases of corruption (resulting in the resignation of three key PSministers) had cast a cloud over the party.

During the election campaign, the press and media gave virtually unbroken atten-

1 Voting is compulsory in Belgium. On 21 May 1995 92.5% of voters went to the polls. Of these7.07% cast a blank or invalid vote.

2 On 21 May 1995 ISPO organized an analytical exit-poll survey on behalf of the Dutch languageNews Service of Belgian Radio and Television (VRT) of the voting behaviour of Flemish voters. Thesample of 3748 voters related to the voters who actually voted in the Flemish Community to the exclusionof those living in the 19 Brussels municipalities. They were interviewed in face-to-face interviews immedi-ately upon leaving the voting booth. The interviews took about 5 min each. The field work was carriedout by Field Quality Control, a market research organization, and the interviewers were trained by AnnCarton supervised by Marc Swyngedouw. The analyses were carried out by Marc Swyngedouw andRoeland Beerten. With thanks to Roeland Beerten and Jaak Billiet for their help in the preparation ofthis paper.

220 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Table 1Election results for the national elections, Chamber of Representatives, 1987, 1991 and 1995 for Flanders(excluding the 19 municipalities of Brussels)a

1987 1991 1995 1995–1991

CVP 31.6 26.9 27.6 +0.7SP 24.6 19.9 20.7 +0.8VLD 18.4 18.9 20.9 +2.0VU 12.9 9.3 7.3 −2.0VB 3.0 10.3 12.3 +2.0AGALEV 7.5 8.0 7.2 −0.8ROSSEM – 5.2 – −5.2Other 2.0 1.5 4.0 +2.5

a AGALEV: green party; SP: social-democratic party;Vlaams Blok(VB): extrem right party; CVP:Christian-democratic party; VLD: liberal party; VU:Volksunie (people’s union) democratic Flemish-nationalist party; Rossem: populist libertarian party.

tion to these “affairs”. It was thought that this would have a very negative impacton the electoral results of the SP. There was a fear that disaffected SP voters wouldchoose to vote en masse for the extreme right.

The VU was also fighting for its life. Not only had it more or less successfullyimplemented its original programme (with the reform of the state in 1993), it wasconfronted with a powerful competitor for Flemish nationalist sympathies. Analysesof voting shifts revealed that in previous elections VU had lost large numbers ofvoters to VB (Swyngedouw et al. 1990, 1992). The new leader of the VU party,Bert Anciaux, not only brought a young and attractive face to the party, the partyadopted a progressive and “green” programme. The new leader threatened to windthe party up if it proved to be incapable of winning at least 300 000 votes.

VB stuck to its usual themes: migrants, safety, the corruption of the traditionalparties and Flemish independence. For the first time in national elections this partywas able to mobilize sufficient financial resources for a professional campaign. Thismoney came primarily from the government under the party finance law.3. Unlikethe 1991 campaign, however, the VB failed to succeed in dominating the electoraldebate.

Although the political scandals continued to overshadow the election campaign,the debate centred on the social security issue and its reform. Apparently the twocoalition partners, CVP and SP, had already reached agreement on this, even beforethe Agusta scandal broke. In doing so the governing coalition partners attempted toprofit from the so-called “issue ownership” which they could to some extent claimregarding the social security theme. The governing coalition used this ploy to makethe electoral stake its own. Pensions and their future financing, the affordability and

3 In 1995 VB would have received BEF 51 978 650 from the Belgian government, on the basis of its1991 election results (Swyngedouw, 1998).

221M. Swyngedouw / Electoral Studies 20 (2001) 217–241

financing of health care, the fight against unemployment and job creation were tobe the stakes for which the election was played.

In the following we will investigate what motives the VB voter gives for votingfor this party and to what extent these motives differ from those given by othervoters. Further on, we want to know whether voters support the VB for the samereasons that the VB claim.

To make a sensible reply to these questions, we must first introduce the idea of“the subjective cognitive and affective map” of the voter. Voters may, as we shallsee, have the most diverse reasons for voting for a particular party. We will thenexamine the pros and cons of working with open questions and discuss the reliabilityof the information obtained. Finally, we will present the results of a survey carriedout on the basis of an exit poll organized during the elections of 21 May 1995. Thisexit poll sought to find reasons for the voting behaviour of the Flemish voter. Uponleaving the voting booth, the voter was asked for which party he or she had votedin the election for the Chamber of Representatives, and was then asked the openquestion, “Why did you vote for this party or why did you vote blank/invalid?”.

3. The subjective cognitive and affective map of the voter

In their classic book“The American Voter”Campbell et al. (1964) describe theimportance of the subjective perspective of the voter for actual voting behaviouras follows:

If we are to understand what leads the voter to his decision at the polls wemust know how he sees the things to which this decision relates. In casting avote the individual acts toward a world of politics in which he perceives thepersonalities, issues, and the parties and other groupings of a (…) contest. Hisimage of these matters may seem at times exceedingly ill-formed, but his behaviormakes sense subjectively in terms of the way these political objects appear tohim. (…) Indeed, it would be difficult to overstate the importance of perceptionsformed by a mass electorate for the decisions it must periodically render. (…) Itis the cognitive map of national politics held by the (...) electorate that is ourconcern here. The elements of politics that are visible to the electorate are notsimply seen; they are evaluated as well. (...) This mixture of cognition and evalu-ation, of belief and attitude, of percepts and affect is so complete that we willspeak of the individual’scognitive and affective map of politics.(p. 15) (…) Takentogether, the cognitive and affective themes recorded in our interviews suggestthe “sense” the electorate made of the objects it acted toward in (the) elections(p. 27). (…) It might be argued that although the “man in the street” may haveonly a loose idea of what is going on in terms of specific policy, he has a firmsense of the global policy differences between the parties, and relates them withequal firmness to highly generalized values of his own (p. 109).

This theory states that the voter, however badly he/she might be informed about

222 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

politics, has a concrete evaluative scheme regarding the various political actors(parties, candidates, media, pressure groups, …). This scheme is not necessarily andin most cases does not refer to the one used by the political elite (politicians, scien-tists, journalists, ...). In our view even the worst informed voter has a kind of politicalmap in his/her head — the origin of which can be very diverse and sometimes it isformed by general (affective) images from days gone by. But for him/her it is ameaningful guide which gives form to his/her political behaviour. This political mapis not necessarily rational, but can be totally emotional or even purely opportunistic.

The survey objective used here gives the voter/respondent the opportunity to men-tion those elements which in his or her own frame of reference have been decisivein determining his/her voting behaviour.Of course this open method of putting ques-tions has its drawbacks. As Campbell et al. (1964: 16) indicate only the dependentvariable of a causal process is described. The way the subjective motivations andperceptions — the cognitive and affective map — of the voters comes into existence,resulting in these motives and perceptions, is not answered by these questions.

4. Open versus closed questions

Inquiries into the subjective, cognitive and evaluative map of the voter, do notnecessarily have to take the form of an open question in “why the voter voted in aparticular way”. It is not only feasible to put the same question in a closed format,this is what is usually done. Furthermore it is known that the results of both formsof question may vary greatly with respect to the distribution of the replies amongcategories and with respect to the links with other variables. Moreover other closedquestioning methods can map other meaningful aspects of the subjective politicalworld of the voter (Swyngedouw, 1993). Over the years growing numbers ofresearchers have opted to use the closed question. This development to the detrimentof the open question is, according to Van Holsteyn (1994: 46), not so much a matterof the soundness of these questions, but rather the practical drawbacks in applyingthe open question. The greater demands made on the respondent (reflecting andformulation) and on the interviewer (listening and making legible notes), combinedwith the intensive coding work following the interviews, discourage manyresearchers from using open questions.

A discussion of the pros and cons of the open question when asking about opinionsand attitudes can clarify the possibilities and limitations of this form of question.4

The most importantadvantagescompared to the closed question are as follows: Asthere is (1) no limitation on the reply alternatives it is (2) well suited to exploring(new) dimensions in reply behaviour. (3) The replies can be given by the respondentfrom his own world view, definitions and context. Apart from the existence orabsence of answer categories (4) they can also be examined for their salience for

4 We have based ourselves on the following literature: Billiet et al., 1984; Sudman and Bradburn,1974; Schuman and Presser, 1981; Van Holsteyn, 1994; Wilson et al., 1996.

223M. Swyngedouw / Electoral Studies 20 (2001) 217–241

the respondent. Moreover (5) the guiding nature which the closed question maydisplay is less evident.

The maindrawbacksinclude the following: (1) the open question requires therespondent to have verbal skills which those with (very) low levels of education donot possess. (2) In some cases they yield irrelevant information. (3) Respondentsgive an answer that is most accessible and plausible to them (referred to as top-of-the-head information by marketing specialists)5 or does not formulate certain informationsimply because it is too evident. The (4) greater demands made of the interviewermean that there is a greater danger of interviewer effects resulting from inadequaterecording. (5) People do not always know why they feel or do certain things. Theyare incapable of being aware of a significant part of their mental processes. Finally,(6) the method is time consuming and the coding and analysis of the open questionmethod clearly takes more time than the closed question.

The relatively limited experimental research into comparisons of the two sorts ofquestions, leads, apart from the observation that there is no evidence that the oneor other form of question is superior to the other,6 to the conclusion that open ques-tions are more suitable in the following cases: (1) when the question is “threatening”,(2) if there is a danger of socially desirable answers, (3) when rapidly changingcircumstances may influence the answers, or (4) when there is a suspicion that newcategories of replies may be emerging.

If we accept these conclusions, everything (in the Belgian situation at least) justi-fies using the open question form. First of all, questions about voting behaviour andthe whys and wherefores is still considered to be threatening by many Flemings (incontrast to the Dutch). Secondly the danger of socially desirable replies undoubtedlyexists (witness the under-reporting of VB voting in many surveys). Third it mayindeed be assumed that the election campaign and the associated events are aimedat bringing about rapid changes in the short term among voters. Finally, the changedsocial and political environment justifies the assumption that there may be newlyemerging categories of replies. Making use of the open form of question in thisresearch, therefore appears to us to be self-evident.

However, this does not give an answer to questions about the validity (do I meas-ure what I think I am measuring?) and the reliability (am I measuring properly?) ofthe open question (about voting motives). Van Holsteyn (1994: 59–79, 82–92) exam-ines the Dutch situation in this respect. He uses a “concurrent validity” method toattempt to collect proof of the validity of the open question. He comes to the con-clusion that in most cases expectations about certain relationships are borne out bythe results of his research. The information obtained from the “open” why-questionscan be given a certain validity. To test the reliability of the answers of the respon-dents he checks their consistency on the basis of statements about the voting behav-iour of loyal voters in panel surveys held during the course of 1981, 1982 and 1986.

5 The same criticism is made in connection with the replies to closed attitude questions. See Zaller(1992: 49).

6 Bradburn (1993) quoted in Van Holsteyn (1994: 39).

224 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

A large number of significant links between parallel voting motivations can be foundeven over this relatively long period. There is also support for the reliability of theopen question (about voting behaviour). Most voters apparently do not reply withthe first thing that comes into their heads.

Are there additional things gained by the open-question approach over the closedquestion? First of all, it seems that better estimates of the relative importance of themotives is possible. This is because the inevitable order effects, from using paperand pencil questionnaires and closed questions, are avoided. Further, it is impossibleto include in a closed question all possible given motives. In this research we dis-tinguished 115 different motives categories. Reducing the number of motives toabout twenty in a closed questions, and forcing the respondent to choose one ormore of these, might lead respondents in a number of cases to point to “available”reasons, not to the real (not presented) ones. In other words: the open question leadsto much more detailed information and to the appearance of motives not asked forin closed format.

5. Reliability of data and coder

In the 1995 BRTN/ISPO exit pol1 3547 voters were interviewed. The sample wasdesigned to be representative of the Flemish electorate and respondents were askedabout the way they had voted for the Chamber of Representatives and why theyvoted for that party (or blank/invalid). Voters could give as many reasons as theywished and were prompted to do so. A coding scheme was developed which drewupon as many of the voting theories to be found in the political science literature;and which allowed every substantive answer given by respondents to be uniquelycoded (Beerten et al., 1997).

The following theories can be found in the coding plan: (directional) issue voting,economic voting, programme oriented voting, interest group defence, ideological vot-ing, socialization theories, candidate effects, party identification, habit, patronage,strategic voting, campaign effects, blurring of lines between parties, the desire forchange or continuity in the government, general evaluative voting behaviour, politicalinefficiency, apolitical attitudes, explicitly pillarized voting, prospective and retro-spective evaluations, pocketbook voting, sociotropic voting and anti-political voting.

Two coders were trained to use the coding scheme. They performed the entirecoding independently of one other. The measure of inter-coder reliability used isKappa (Agresti, 1990: 366–370; Popping, 1995).7 A Kappa of 0.82 (95% C.I.: 0.80–0.84; Cronbach’sa: 0.90) was reached. Popping defines a Kappa in the range of0.81 to 1.00 as “almost perfect”.

7 All reported reliability measures are calculated for the first motive. In the training phases we usedthe computer program AGREE 6.0 (Popping, 1995), because its allows only 60 categories we used SPSS6.1.2 for the coding operation as a whole.

225M. Swyngedouw / Electoral Studies 20 (2001) 217–241

6. The cognitive and affective map ofVlaams Blok voters on 21 May 1995

An important advantage of our data is that they were collected at the moment ofthe voting act itself. As a result memory distortion is small if not non-existent. Whatwe think is even more important and this is the unique quality of this data — whenthe voters were interviewed they had no idea of the results and had not been influ-enced by commentators, the statements of politicians, journalists, researchers, etc.speaking on the radio, television or writing in the press. Nor was it possible for anyreinterpretation of the voting behaviour in the light of whether the party for whichthe voter had voted had gained, lost or remained the same.

The response rate was about 60%. A slight under-representation of VB electors,other small parties and blank/invalid voters in particular was noted. In order to rem-edy this under-representation the data was weighted with weighting coefficientsobtained via an iterative proportional fitting to the population figures for Age*Sex*Voting behaviour 1995.8 No voting motives could be recorded for 6.9% of therespondents because of refusal, uncodable (for idiosyncratic reasons) or the illegi-bility of the replies. Most voters therefore (93.1%) were capable of giving a meaning-ful reply in one or another way to the open question.

To provide a guideline to the interpretation of the information we will group thevoting motives in such a way that they yield the most meaningful classification ofvoting motives from the point of view of voting behaviour theory.9

6.1. Ideologies and values

In much electoral research the point of departure of the analysis is the votingbehaviour in terms of ideology and values. Studies of Belgium point to conventionalconflicts based on religious ideology, ethnic and nationalist sentiments, and theleft/right dichotomy, which may or may be not combined with more recent cleavages(Swyngedouw 1992, 1993; Billiet and De Witte, 1995).

In our research 12.2% of respondents used one or other form of ideology or valueto justify their voting behaviour.

From Table 2 it is clear that the VB voter is not or only slightly inspired byvalue motives. The VB voter is significantly more sensitive to people’s nationalistideological values. The people’s nationalist value reason or people’s nationalist ideol-ogy (distinguished from the topic/issue communitarian reasons) is quoted only by1.57% of voters as the reason for voting for a particular party. The VU continuesto be the most significant people’s national party for this small group. Indeed thisis an explicit reason for voting for this party for 17.6% of VU voters. The VB electorrarely quotes his reason, only 3.2% of the VB electorate thinks this is a reason forvoting VB. The voter inspired by people’s nationalism prefers democratic people’s

8 The weighting method employed is identical to that used for the post-electoral election surveys ofISPO 1991 and 1995 (ISPO/PIOP, 1995: 21–23).

9 The attribution of motives to the various paradigmatic theories is purely tentative and is most certainlynot final. Further research is necessary.

226 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Table 2The percentage of voters mentioning a particular value to explain their voting behavioura

AGA SP VB VU VLD CVP BL/I POP

Valuesb 9.61 5.93 3.98 8.65 4.58 6.78 0.00 5.55Christ.or not. 0.77 (0.15) (0.00) (0.00) (0.15) 9.03 1.98 2.54people’s nat. 0.00 (0.00) 3.23 17.60 (0.00) (0.00) 0.00 l.57Ecological 20.25 (0.00) (0.00) 0.25 (0.00) (0.00) 0.00 1.33L-centre-R 0.93 l.97 0.44 2.18 l.87 l.67 0.00 l.47

a Every row in the table is basically the row of a table which indicates the presence or absence of aparticular reason for the various parties. A chi-square test was used to check whether there was an associ-ation between parties and voting motive. Should this be the case (p,0.001) the decomposition of the chisquare statistic in the cells would indicate which parties are over or under-represented. The figures shownin bold type are significant over-representations (p,0.05). The figures between brackets () are significantunder-representations. Remember that respondents may indicate more than one motive, which means thatthe sum total of all the percentages for all the motives may be more than 100. The abbreviation POPindicates an estimated percentage of voters in the total population mentioning a particular motive. Theabbreviation BL/I indicates a blank or invalid vote cast. The same remark applies to all subsequent tablesas well.

b The category “Values” includes reference to motives of the voters stating that “the general values”“the general ideas and ideology” of a certain party was the reason to vote for it. No specifications weremade of the nature of the values, ideas or ideologies.

nationalism to authoritarian people’s nationalism. 76.7% of them vote for the VU,while only 23.3% vote VB.

6.2. Programme and policy issues

The traditional position model, attributable to Downs,An economic theory ofdemocracy(1957), generally assumes that the voter will vote for a party whosepolicy position on the issues he considers important are closest to his or her viewson the matter. Because the voter is not or cannot always be fully informed he willlimit himself to certain striking questions which he continuously associates withparticular political parties. Over the years numerous variants and refinements havebeen applied to this original model (Table 3).

A total of 12.8% of voters refer to specific policy questions to explain their vote.Moreover slightly more than 6% of voters refer to the general programme of thepolitical party they favour. What is striking is that the VB and the VLD are over-represented here. Regarding the over-representation of VLD and VB it could beargued that the issues “owned” by VLD and VB (VLD: tax, reductions in governmentexpenditure, free enterprise; VB: migrants) did not appear or hardly appeared at allin the election campaign. The voter is then forced to fall back on the general ideashe or she has of the party programmes.

The migrant theme motivated a little more than 4% of the electorate. Only theVB really profited from this issue. 33% of VB voters explicitly mention it as a votingmotive. Agalev, which adopted a pro-integration stance on immigration marginal tothe main political campaign, won few or no voters on this issue.

227M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Table 3Percentage of voters per party referring to the programme or a particular issue as voting motive

AGA SP VB VU VLD CVP BL/I POP

Programa 7.31 5.01 9.75 8.28 11.17 (4.25) (0.00) 6.52Soc. Sec.a (1.33) 8.98 3.17 (1.14) 6.76 (1.20) 0.46 4.24Migrantsa 1.82 (0.38) 32.96 (0.7l) (0.79) (0.09) (0.00) 4.21Environmenta 21.90 l.57 (0.00) (0.00) (0.35) (0.15) (0.00) 1.90“Affaires”a 0.29 5.06 0.45 0.60 (0.36) 0.63 1.11 1.35Economica 0.68 0.54 0.40 0.39 3.40 0.60 0.00 1.11Commun.a 0.25 0.18 2.17 5.32 (0.00) (0.00) 0.00 0.66Crime 0.00 0.19 4.78 0.00 0.15 (0.00) 0.00 0.60

a Social Security: pensions, social security, unemployment and job opportunities, poverty;Migrants:arguments against migrants, for the integration of migrants and mention of migrants without indicationof opinion, voted for migrant;Environment: Environment, traffic problems and the Manure Action Plan(MAP); Affaires —Political Scandals: political corruption+ party in victim role;Economic voting: econ-omic motives, taxes, government spending cuts, savings on government spending;Communitarian issues:further federalization, confederalization, splitting up of social security funds, return to the unitary state,less federalism, solidarity in social security funds;Crime: criminality, police, law and order;Programme:include all general reference to the electoral programme of a party, without pointing to a specific issue.

The communitarian theme too distinguishes the VU from the VB. The number ofelectors mentioning this as a reason are, however, limited (0.66%). The VU hasslightly over half (55.3%) of the voters reporting this as an issue, while VB had37.3%. It seems that the hard-core, which thinks this issue is important, are adoptingmore extreme positions in the direction of separatism. Combined with the people’snationalist orientation it is clear that the VU is still the leading communitarian party.It still attracts 71.0% of the voters explicitly inspired by communitarian/nationalistissues. The VB attracts only 26.7% of these voters. The total size of this electoratecan be estimated at2.2l%.10

In the immediate post-election period there was much speculation about the impactof the political scandals. Our research indicates that only 1.3% of the vote wasdirectly affected by the political scandals. Moreover it appears that 70% of thesevoters voted for the SP because they thought that the SP had been pushed into thevictim role. 5.06% of SP voters on 21 May 1995 referred to this issue whenexplaining their choice. If the Agusta affair had not happened, the SP would havefallen back slightly compared to 1991. What is striking is that the VB did not succeedin turning the issue into electoral gold.

Finally, regard crime, police and law and order issues as motives for voting fora particular party. The VB featured these issues prominently in its programme, andvirtually all the traditional parties echoed on similar themes. From our research it

10 This might seem, at first sight, to be remarkably small. Nonetheless it comes close to the proportionof voters obtained by Maddens et al. on the basis of the 1991 election survey, when those Flemishautonomists who consider the communitarian theme to be important as well are selected, namely 3.12%(Maddens et al., 1994: 60, table 4.1).

228 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

appears that only 0.6% of voters were seduced by this theme, which was given somuch prominence in the media. Normally an analysis of 21 units would be unreliable.Even so, 19 of them — virtually all — voted for the VB. In electoral terms thecrime theme is the exclusive concern of the extreme right and is the motive of 4.8%of the VB’s voters.

6.3. Socialization and identification

Political socialization via education, family, friends and acquaintances used to beregarded, in the consociational democracy that Belgium once was, as the vote defin-ing mechanism above all others. People voted within the political family or pillar(compartment) in which one was born.

US electoral research created the term “party identification”. The Michigan Schoolin particular places party identification at the heart of the models it uses to explainthe voting behaviour of the US electorate. The idea is that voters identify closelywith one of the dominant political currents. The applicability of the Michigan modelto multiple party systems of the European type must, however, be called into question(Thomassen, 1976). Even so statements such as, “I always vote party X”, “I thinkit’s the best party”, and “I’m a paid-up member of party X” or “I’m a Christian-Democrat, Socialist, Liberal, etc.” must be interpreted in the tradition of party identi-fication (Table 4).

9.1% of all voters refer to education, family, partner, friends, and/or job environ-ment as a factor with direct influence on their choice for a particular party. As couldbe expected and as predicted by theory, it is chiefly the “old pillar parties” who winvotes via socialization. Socialization motives are scarce among VB voters, as theparty is a relative newcomer to the political scene (established in 1978).

Identification motives in the broad sense are mentioned by 21.4% of the electorate.If we look at the three categories together (habit, best party and identification in thenarrow sense), the CVP is striking for its consistent over-representation. No less than35.2% of the voters for this party refers to one of these three categories as a votingmotive. The SP is the runner-up with 29.9%. The third and fourth are the VLD(19.4%) and VU (18.4%) respectively. The identification motive is clearly signifi-cantly under-represented among VB voters.

Table 4The percentage of voters reporting socialization or identification as a voting motive

AGA SP VB VU VLD CVP BL/I POP

Socialization (4.09) 11.55 (2.13) 8.20 9.85 14.28 (0.00) 9.11Habitude (2.02) 21.14 (0.45) 9.24 (8.82) 23.31 (0.00) 12.58Best partya 8.89 7.45 (4.23) 8.62 7.91 10.08 (0.00) 7.31Identification 2.42 2.33 0.64 0.93 3.26 3.42 (0.00) 2.28

a Best party: The vague motive of this is the “best party” was interpreted as an indication of identifi-cation with a certain party.

229M. Swyngedouw / Electoral Studies 20 (2001) 217–241

6.4. Parties

We can list four motives under the “parties” heading. Admittedly these all haveto do with the “institution” of political parties, but have nothing to do with the actualparty positions. These four reasons for voting for a particular political party arenonetheless associated with so-called formal or activity elements.

The first voting motive has to do with “party image”, i.e. the marketing aspect ofevery modern party operation. These are references to style, the way in which(according to the voter) the party goes about the business of politics, alternativelyit may be the distinctiveness of the party. This may be expressed in terms such as“the underdog” or “different from the others”. We have also placed here the (rare)references to election campaigns in which no element relating to the programme orvalues can be found.

A second voting motive that can be put in this group is so-called “strategic voting”.This is a calculated choice by the voter in order to achieve certain objectives, or toprevent things from happening, such as the desire to avoid undesirable side-effectsor worse outcomes. We may distinguish two kinds of strategic voting. First of allthere are the voters who do not vote for the party they prefer, but vote for another.They do this to prevent another party from winning the elections (Heath and Evans,1994). Secondly there is expressive strategic voting. This type of voting covers allthe cases in which an individual does not choose for his/her preferred party in orderto send a message to this preferred party or to other parties (Franklin et al., 1994).Such voting is thus not determined by content or party ideology, but by the positionof the party in the political landscape. In this connection we hear arguments suchas the “The best of a bad bunch”, “Don’t vote for a small party”, “It’s a small party”,“To guarantee the continued existence of party X”, “Because it’s an oppositionparty”, “It’s a protest party”.

A third motivation which can be put under the “parties” heading is rooted in the“clientism” system that was once so typical of Belgian political parties. Party loyaltywas established via political patronage, which fitted in with the individual strategyof the candidate or otherwise.

The final party-related voting behaviour is “negative voting”. The voter does notvote for a party, but against a particular party. This motive must be clearly dis-tinguished from apolitical or anti-political voting. The voter disagrees, or is disap-pointed, in one or other political party. There are admittedly no elements of revengeor protest in this motive, but nor are there elements which have anything to do withthe programme of the party itself.

The essential distinction between strategic voting and negative voting is that inthe latter case the vote motive is directednegatively againstanother specified party,while for the first the vote motive lays in some property(ies) of the party vote for(Table 5).

6.3% of voters explain their vote by referring to party image. Although not sig-nificantly over-represented, party image is not an unimportant reason for voting forVB. Nearly 7% of VB voters motivate their vote in this way. The professional (andexpensive) VB election campaign and the stress on the fact that it is different (not

230 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Table 5The percentage of voters in each party reporting party characteristics as a voting motive

AGA SP VB VU VLD CVP BL/I POP

Image 8.29 5.02 7.07 4.81 5.54 9.19 (0.00) 6.34Strategic 12.32 4.54 3.35 9.95 (0.90) 4.12 (0.00) 4.14servicesb. (0.00) 3.64 (0.25) 0.22 2.86 2.35 (0.00) 1.84negative 0.81 1.69 0.73 3.62 2.76 1.63 1.00 1.71

corrupt, uncompromising, principled, etc.) from other parties did pay off to a cer-tain extent.

1.7% of voters voted purely negatively. These voters voted for a particular partybecause they are against another party. The VU voter giving this as a reason is votingagainst the VB. The undemocratic nature of its programme, its intellectual roots inpre-war fascism and its rabidly anti-migrant stance, which threaten to bring demo-cratic Flemish nationalism into discredit, are probably the reason for this.

6.5. Candidate effects

We are saying nothing new when we observe that the candidates in modern elec-tion campaigns play a key role. A striking element here is that the candidates of allthe large parties include photographs of and text about, not only themselves, butalso their partners and/or their children. The boundary between public and private lifeis becoming increasingly vague. The continued advance of professional marketingtechniques in election campaigns certainly has something to do with this (Table 6).The voting behaviour of 8% of voters is influenced by personal acquaintance withcertain candidates or because they sympathize with certain candidates. For VB voters,the VB’s candidates do not constitute a reason for voting for the party.

6.6. Policy

Policy is the central motivation of the voting behaviour of the so-called rationalvoter. This voter is regarded as being fully or adequately informed. He is supposedto rationally weigh up the general political work of the various parties in the past(retrospective) and the expectations regarding future work (prospective), which thenforms the basis of an informed vote. The call for policy renewal and the positive or

Table 6The percentage of voters per party reporting candidates as voting motive

AGA SP VB VU VLD CVP BL/I POP

Candidates (2.46) 7.62 (0.88) 15.24 7.60 13.38 1.27 7.87

231M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Table 7The percentage of voters per party reporting policy work as voting motive

AGA SP VB VU VLD CVP BL/I POP

Renewala 4.01 (0.42) 8.87 3.88 15.77 (0.22) (0.00) 5.33Gooda 0.89 1.83 (0.45) (0.00) 1.48 4.11 0.00 l.77Governmenta (0.00) 0.75 1.45 0.30 1.49 3.56 0.00 1.54

a Renewal: “Small other parties”: 12%, significant. “Good” is the general evaluation of the politicalwork of the various political groupings, as well in opposition as in coalition, without referring to govern-mental work. These are very general statements (“They are doing well”), which are both retrospectiveand prospective. “Government”: Both positive and negative remarks are included in the “government”category and these refer explicitly to government work and/or the government coalition (e.g. the govern-ment should stay, bad government).

negative evaluation of the work of the previous government are further parametersin this policy evaluation model (Table 7).

In total, 8.6% of voters report a policy motive to explain their vote. The desirefor renewal in (government) policy is mentioned by 5.3% of voters. The VLD andVB are over-represented here. 8.9% of VB voters and 15.8% of VLD voters explaintheir vote in such terms. A not insignificant section of VB voters expect generalpolicy renewal by VB. We return to this below.

6.7. Defence of group interests

Lipset and Rokkan (1967) explicitly linked the defence of the interests of thevarious social classes to their cleavage model to explain the birth and developmentof different political parties. In Belgium this is used to explain the existence of theCVP, SP and VLD. Similar defence of interests, based on other social, economicand cultural developments and the cleavages arising from them, are supposed to bethe basis of the growth of other political parties, namely the VU, Agalev and the VB.

Apart from group interest as a reason for voting for a particular party, the defenceof the “general interest” (the so-called altruist vote) or pure self-interest (sometimescalled the pocketbook vote) may be the driving force behind voting behaviour. 8.44%of voters report one of these motives to explain his or her voting behaviour (Table 8).

None of the traditional social groups, workers, or entrepreneurs (self-employed)see the VB as a party representing their specific interests. This contrasts with theVB publicity which claims that the VB is the true defender of the working man.“Self interest” as a reason for voting for a party, but without reference to one or theother social class, is a reason common to voters for all parties. The VB voter isnonetheless heavily over-represented in this group. 4.4% of VB voters report thisargument, compared to 1.4% for all voters. These results challenge the traditionalMarxist interpretation of extreme right as a product of the petty bourgeois interests,at least with regard to the given motives. What is clear, however, is that in thearticulation of their motives a rather small but significant group of VB voters disen-gage self interest from class or group interest. From the point of view of subjective

232 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Table 8The percentage of voters per party reporting the defence of (group) interest as voting motive

AGA SP VB VU VLD CVP BL/I POP

self-employeda (0.00) (0.00) (0.00) (0.39) 13.19 (1.14) (0.00) 2.87working people (0.22) 13.51 (0.44) (0.38) (0.00) (0.18) (0.00) 2.80Gen. Interesta 2.48 l.07 0.34 1.13 0.51 1.05 0.00 1.48Own Interest 0.25 1.29 4.44 0.39 2.01 0.83 0.00 1.43

a The “self-employed” category includes the following groups: self-employed persons, shopkeepers,independent professionals, entrepreneurs, business, and farmers. The “general interest” category includesmotives referring to the “general prosperity, general interest”, as well as women’s interests, children’sinterests and the interests of elder people. These latter three categories were joined with the “generalinterest” category because all four are rather small.

motives, interpreting the extreme right VB vote as the result of social groupingslooking for the defence of their interests must be rejected.

6.8. Pillar membership as a voting motive

Membership of one of the major political families or “pillars” has long givenremarkable stability to voting patterns in Belgium and Flanders (Huyse, 1980; Billiet,1982; Frognier and Gollinge, 1984). However, with the erosion of the power of thepolitical pillar organizations, voters have started to behave more independently. If“pillar membership” is nowadays given as a reason for casting a particular vote, itis almost always employment related. This is employment in the party or in one ofthe many organizations run by the political families (Table 9).

All in all only a small group of the electorate report this motive (1.12%). As couldbe expected this argument is irrelevant for VB.

6.9. Political powerlessness and revenge (anti-political sentiment)

Apolitical views, feelings of powerlessness or political inefficiency must be clearlydistinguished from anti-political attitudes.

An anti-political attitude means first of all that the individual rejects or opposesthe dominant political currents of society. At an elite level such feelings are expressedin the contra-revolutionary tradition, nationalism and fascism (Ignazi, 1996). For theindividual it is expressed in the form of hostility towards the role, operation or actionof political parties. An anti-political attitude is expressed as a generally undifferen-

Table 9The percentage of voters per party reporting pillar membership as voting motive

AGA SP VB VU VLD CVP BL/I POP

Pillarization 0.00 1.95 0.34 0.38 0.58 2.30 0.00 1.12

233M. Swyngedouw / Electoral Studies 20 (2001) 217–241

tiated negative feeling about all parties. The individual is disappointed in all politicalorganizations and in politics in general. An anti-political vote is an expression ofthe feeling that nowadays no politician can be trusted. It points to a protest vote.For some this means that one takes revenge on a political system in which the indi-vidual feels entirely irrelevant. The “anti-political” voter is convinced that the “con-ventional” parties are incapable of making him relevant. Some of these protest votersseek to realize their hopes by voting for anti-system parties which reject the basicsof the democratic system. Extreme right parties have consistently profiled themselvesas being “anti-party” throughout West Europe (Betz, 1994). Nonetheless the protestvoter does not explicitly opt for a non-democratic system. There is clearly no specificalternative social structure at the roots of their protest. It is merely an indication ofan emotional rejection, disappointment, revenge and protest against the existing sys-tem and its political organizations. Rapid social change (with the associated rise ofpost-material values), the expansion in information delivery systems, the furthergrowth of participation systems and individualism have led to a situation in whichthe political parties no longer succeed in integrating significant parts of society(Poguntke and Scarrow, 1996).

A clear distinction must be made between anti-political motives and strategic —and negative — motives (see above). To be categorized as an anti-polical motive,the motive must be equally negative toall political parties or to the political systemin general. Where strategic and negative motives are directed (strategically ornegatively) towards one party (see above).

An apolitical attitude on the other hand does not indicate feelings of rejection. Itis more indicative of a lack of political interest. The individual does not understandthe political scene because it has become too complicated. Moreover, he or she thinksit is not worth the bother of learning about it. Alternatively, politics are no longertaken seriously; people laugh about it. This may range from reactions such as “Icouldn’t be bothered” to “You’ve got to vote for something if you are forced to”(Table 10).

About 13.9% of all voters report voting motives based on arguments belongingin the “political powerlessness” and “anti-politics” category. This is an indicationthat the legitimacy of the political system is, for various reasons, called into questionby a significant minority of voters. About 8.7% do this because they “no longerunderstand it” or they “aren’t interested in politics”. Clearly such voters no longerparticipate in politics. 6% of these voters are truly anti-political. They have lost theirfaith in politics. This last category must be characterized as anti-system voters.

Table 10Percentage of voters per party reporting political powerlessness or antipolitical sentiments as voting motive

AGA SP VB VU VLD CVP BL/I POP

powerlessnessa (2.79) (2.76) 6.72 (4.85) (3.84) (4.24) 62.71 8.70Anti-Politics 6.82 (1.36) 13.70 (2.36) (2.64) (0.64) 32.05 5.75

a Political powerlessness, political inefficiency. “Small Other Parties”: 13.12% significant.

234 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

The most striking thing is that 89.3% of blank/invalid voters belong to these twocategories. Broadly speaking, about two thirds of them can be designed as politicallypowerless, while one third can be regarded as anti-political. The VB is the only partywhich is not significantly under-represented in either category. Indeed, it is heavilyover-represented among the “anti-political” voters. According to our findings the VBelectorate consists of 13.7% of anti-politics voters. 6.8% of Agalev voters were anti-political. The “small other political parties” and VB are, as might be expected, theonly parties which have over-representations of “politically powerless” voters.

However, because of the coding scheme used (Swyngedouw et al. 1996), it is notclear if we need to interpret the anti-politics vote as voting for the VB because itis an anti-system party (i.e. its politics will completely change the the system; a questfor renewal) or just as an anti-politics act (i.e. what the VB does is not important, thesymbolic act of voting for it is what counts; a protest vote).

7. Motivation profiles for voting behaviour

7.1. The method: HICLAS

We used the results to create motive profiles for the different parties running inthe elections. This was done in the following way. For each group of motives a chi2

test was performed to test the relationship between the motive (or group of motives)and voting behaviour. A data-matrix of objects (parties) by attributes (the motives)was formed by entering zeros and ones. A one was coded when a party was signifi-cantly over-represented for a given motive(group). This matrix was analysed withHICLAS.11 We used the deterministic disjunct variant of the model.

Our HICLAS analysis links the different motives to the different parties. In thisparticular analysis the motives are hierarchically structured. This is to say that somemotives can be unique for a party, but others have to be shared by two parties andstill other motives have to be shared by three parties. The lines linking the differentblocks of motives from top to bottom indicate which motives are related to whichparties. Global and local fit indices can be used to evaluate the HICLAS-model andto select the most appropriate solution for a matrix (see Appendix).

The result of the analysis is an overview of the more than randomly attributedreasons for voting for the different political parties. Or to put it in managementjargon, we provide an overview of the selling points for each party in the 1995general elections in Flanders for the Chamber of Representatives. Fig. 1 gives theresults of the analysis.

The lightning flash line links parties with voting motives. The blocks on the motiveside represent so-called “motive publics”. By this we mean that their motive has thesame formal object, but the direction of the motive may be different depending on

11 See for example: De Boeck and Rosenberg, 1988; Storms et al., 1994; De Boeck et al., 1993. Withthanks to Gert Stroms (K.U. Leuven) for his help in studying HICLAS.

235M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Fig

.1.

Sig

nific

ant

mot

ives

for

votin

gin

the

1995

Gen

eral

Ele

ctio

n,F

land

ers,

Bel

gium

(HIC

LAS

-sol

utio

n).

236 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

their connection with the parties (see further). The first level of the properties (thebottom classes) gives those voting motives unique to a party. For example, environ-ment and ecological ideas are unique to the green party Agalev. On the second levelon the property side stand those voting motives which two parties share. For example,people’s nationalism (nationalist) and the community issues are shared by the VUand the VB. On the third level are those voting motives shared by three parties. Forexample, the social security issue, which is shared by the SP, the VLD and the othersmall parties. It should be clear that the content of the motive is different for thesupporters of the different parties. SP voters most certainly want to keep the socialsecurity system as much as possible as it is now. The VLD voter is looking for amore liberal (US style) private system. The voter for “other small parties” is mainlyworried about the future of the pension system in Belgium. That the analysis ishierarchical can be seen by looking at the voting motive “renewal” (of the politicalsystem). On the one hand it is linked directly to “other small parties”, while on theother hand it relates to the voting motive “party program” (in general). This in turnis linked to the VB and the VLD. In effect this means that both parties are signifi-cantly over-represented with regard to the motive “party program” and moreoveralso with regard to the motive “renewal”. Such a structure is not found for the “othersmall parties”, these are only directly linked to the motive renewal.

7.2. The motivation profiles of parties

Each of the characteristics shown represents what is a called a “motive popu-lation”. That is say these are voters who mention this characteristic as a motive fortheir voting behaviour. Motive populations can be divided according to the direction(sign) of the motive (e.g. maximum conservation or change social security). Theparties associated with the same motive should therefore share in this “motive popu-lation”. Let us now examine the relationship between the various motive populationsand VB.

The unique selling propositions of VB are negative. The VB voter is “against”something and if the voter is for something, then it is the voter’s self interest. Thecontent of the basic class associated with VB is against “migrants” (33% of VBvoters), against (rising) crime (4.8%), for self-interest (4.4%) and “revenge” on thepolitical system (13.7%). In view of the marginal numbers of voters reporting crimeand self-interest as the prime motives, the anti-immigrant theme and anti-politics(revenge) are the main reasons for voting VB. The VB shares the anti-political voterswith the “blank/invalid” option, but it is, as the only anti-system party, the only onewhich tries to win the votes of these electors. Like the VLD voters, VB voters make(vague) references to its “programme” (9.75%) and to “renewal” (8.9%). The lattervoters are also shared with the “small other parties” (12%).

VB shares the communitarian (3.2%) and people’s nationalist voters (2.2) withthe VU. The latter group is probably the hard-core of ideologically inspired VBsupporters. We have already pointed out that this is a relatively limited market seg-ment (2.21% of all voters). The VU is far and away the preferred party of this motive

237M. Swyngedouw / Electoral Studies 20 (2001) 217–241

group. We do nonetheless have the impression that the VU in 1995 has lost groundin this respect compared with 1991.

8. Discussion

The use of the open questions to determine the motives of voting behaviourappears to be a success. Wilson et al. (1996) show that this question must comeafter the question about actual voting behaviour in order to avoid unwanted contexteffects. The open question method also appears to work in an exit poll, 93% ofrespondents appeared to be capable of giving relevant and codeable answers. Themain advantage of using this method in an exit poll is that the respondents still arereally replying in their own reference frame and have not yet been influenced bythe statements of all kinds of specialists (experts, politicians, reporters) in the media(TV, radio, newspapers and magazines) about the causes and reasons for their votingbehaviour (the message of the elections). Nor are they influenced by knowingwhether their party has gained, lost or stayed the same compared to previous elec-tions.

If we look at the results which we obtained from the motive profile structure, wesee that nearly all important voting behaviour models are referred to in replies givenby the respondents. The models used by the voters themselves range from partyidentification and socialization to issue voting, economic voting, defence of groupinterests, strategic voting, ideological and value voting, etc. What is clear is thatsome voting behaviour models are more suitable for explaining the results of someparties than of others. The advantage of the open question is that the researcher doesnot try to explain the results of all parties on the basis of a single model.

It also appears to be possible to establish profile motives via HICLAS for thevarious parties. In our method a motive is held to belong to a party specific profileif it appeared more than expected in the motives indicated by voters for a particularparty (given the marginal size of the motive in the entire electorate and given themarginal size of the party support). This definition, for example, results in the con-clusion that the left/right dichotomy should not be attributed to the profile elementsof just any party in Belgium Flanders. Similarly it indicates that for the VB the partyimage (7.1%) is not part of its profile for its supporters, in view of the fact that instatistical terms it is not more or less present than in the entire electorate. Partyimage thus has no significant influence on voting or not voting VB.

Compared to the closed questions strategy, the methodology used here (open ques-tion and HICLAS analysis) leads to some undeniable advances. It is nearly imposs-ible with closed questions to construct and fit one overall multivariate model linkingall the different motives to the different parties. Secondly, the strategy reveals theexistence of motives to support the VB which have previously be mentioned onlyrarely if at all. Above all, the results lead us to re-evaluate the relative importanceof the different motives (e.g. Flemish nationalism and communitarian issues, crimeand law and order, political scandals, etc.).

The motive profiles suggested by the VB voters themselves in their own subjective

238 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

framework, on the one hand confirm a number of known trends, but on the otherhand, throw new light on neglected profile elements. All the explanatory modelsreferred to in the introduction recognize the sensitivity of the migrant and law andorder issues. The migrant issue is by far the most important reason for voting VB.Law and order is also important, but the number of Flemish voters who explain theirvoting behaviour in terms of this issue is so small that it can only have a veryminor influence on the VB’s results. People’s nationalism and the associated typicallyBelgian communitarian issues are part of the VB motive profile. In contrast to theamount of attention given to these issues in the media, we note that only 2.1% ofall Flemish voters fall into this motive population. In total they account for about5% of the VB’s support. Compared with the immigrant issue, therefore, it is relativelyunimportant. Clearly in 1995, it was not party identification, nor its candidates, northe defence of group interests, nor its liberal free market economics which movedelectors to vote for the far right. The defence of personal self-interest was, however,a reason for 4% of the VB’s voters. This result makes it clear that traditional socialgroups do not see the VB as the party representing their interests. Neither do theworkers, nor the self-employed.

The new element is the finding that some of the VB’s supporters vote for its entireprogramme (9.8%) and that the VB would renew the Belgian political system (8.9%).The programme support comes probably from the hard line supporters of the party,as well as from people who prefer for reasons of social desirability to refer to thevaguer and more neutral programme than the hotly disputed “migrant issue”. As forthe renewing motive it may be feared that apart from the extremist support, there aremany voters who object to the political methods of the established parties (clientism,ideological blurring, etc.) and the corruption (scandals) in which they are involved.The VB appears to be, to some extent, reliant on the anti-political voter seekingrevenge on the system (13.7%). It is the only party recruiting voters in this group.In view of the events of 1996 and 1997 in Belgium (the Dutroux affair and theevident failing of the Belgian justice system, the various corruption scandals and theinability of the political classes to prevent factory closures) it may be expected thatthe group of anti-political voters will grow and that confidence in political organiza-tions will be further undermined. In other words, the legitimacy crisis of politics andthe party system will become even worse. It may therefore be expected that thisgroup of voters could be a future source of further electoral gains of the far rightVB party.

This research makes it clear that the reasons why voters support the VB are to alarge extent different from the reasons that the VB claims that they support it. Fur-thermore, it reveals that the VB vote is not solely a question of economic self-interestor fear (of crime and immigrants). This has important consequences. Solutionsdesigned to “manage” these issues (e.g. by targeting assistance to economically mar-ginalized people or by taking a “hard line” on immigration and crime) will ultimatelynot completely undermine the support for the extreme right. Given the rejections ofparty politics and the anti-politcal attitude of the extreme right voter, dealing withthe support of the extreme right becomes highly complex.

239M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Appendix A. HICLAS

The deterministic disjunct version of HICLAS used in this study (see Stroms etal, 1994) is a “set theoretical” model for binary (0/1) objects by property data. Itsearches for the best possible set theoretical structure for the objects and their proper-ties (Table 11).

It shows simultaneously (1) the classes of objects and the classes of properties(2) the hierarchical relations between the objects on the one hand and between theproperties on the other and (3) the association of the structure that connects theobject classes with the property classes. The goal of the analysis is to develop themost parsimonious model possible to describe the binary data, and which is substan-tially meaningful. There are (Jaccard) goodness-of-fit measures to evaluate the rep-resentation of the different objects and properties as well as for the evaluation ofthe entire model.

Formally HICLAS is a member of the family of clustering techniques. Contrary toalmost all hierarchical clustering techniques, HICLAS does a simultaneous structuralanalysis and representation of the features and objects, reflecting their relatedness(association). In classic clustering separate analysis of the features and the objectshas to be done. Because of this simultaneous representation of features and objects,one can see the model as linked to some MDS-models.

HICLAS allows concidering hierarchical classes models that do not fit perfectlywith the data. It uses a minimisation algorithm that searches for the best fitting model,given the user specified fixed number of bottom classes (the rank of the model).These fixed numbers of bottom classes (or the rank) can be seen as the number of“factors” in the solution. In this sense the model is related to Boolean factor analysis.A scree test can help to determine the rank of the solution.

Formally, (Storms et al., 1994) disjunctive HICLAS is defined by two componentsand their concatenation. The first component is three set-theoretical relations ofobjects (parties) and their features (motives). The second component is a Booleandecomposition of binary objects by features matrix.

The three set-theoretical relations are: (1) equivalence relation: Objects are equiv-alent if and only if they have identical sets of features. Features are equivalent ifand only if they characterise identical sets of objects. (2) association relation: An

Table 11Example: Fictitious motives by parties binary matrixa

Motives/Parties CVP SP VB VLDProperty/Object

Issue 1 1 0 1 0Value 1 0 1 0 1Left–right 0 1 0 0Anti-politics 0 1 1 0

a A 1 means a significant overrepresentation of the motive at stake for the respective party.

240 M. Swyngedouw / Electoral Studies 20 (2001) 217–241

object class (equivalent objects) is associated with a feature class (e.g. equivalentfeatures) if and only if the feature class is in the set of feature classes possessed bythe given object class. The association is symmetric. (3) Order relation: An objectclass is below another object class if and only if its associated feature class(es) is(are) a proper subset of associated feature class of the other. This is partial orderrelations implying overlapping and non-overlapping hierarchies.

Boolean Decomposition: A Boolean decomposition of anm objects byn featuresbinary matrixM exits and the three set-theoretical relations are restrictions on thisdecomposition.

HICLAS uses a minimisation algorithm to look for the best fitting model whichis not necessarily perfectly homogeneous at the level of the data (De Boeck andRosenberg, 1988). In other words discrepancies between model and data are allowed.The necessary fit measures are available to evaluate these discrepancies and to selectthe most appropriate parsimonious model given the observed data.

References

Agresti, A., 1990. Categorical Data Analysis. Wiley, New York.Beerten, R., Billiet, J., Carton, A., Swyngedouw, M., 1997. 1995 general election study Flanders-Belgium,

Codebook and questionnaire, Leuven, Department of Sociology, ISPO, 217 p.Betz, G., 1994. Radical Right-Wing Populism in Western Europe. St. Martin’s Press, New York.Billiet, J., DeWitte, H., 1995. Attitudinal dispositions to vote for a “new” extreme right-wing party. The

case of “Vlaams Blok”. European Journal of Political Research 27 (2), 181–202.Billiet, J., 1982. Verzuiling en politiek; theoretische beschouwing over Belgie¨ na 1945. Belgisch Tijd-

schrift voor Nieuwste Geschiedenis 13 (1), 110–122.Billiet, J., Loosveldt, G., Waterplas, L., 1984. Het Survey-interview onderzocht. SOI, Leuven.Campbell, A., Converse, P., Miller, W., Stokes, D., 1964. The American Voter. An Abridgement. Wiley,

New York.De Boeck, P., Rosenberg, S., Van Mechelen, L., 1993. The hierarchical classes approach: a review. In:

Van Mechellen, I. et al. (Eds.), Categories and Concepts. Theoretical Views and Inductive Data Analy-sis. Academic Press, New York, 265–286.

De Boeck, P., Rosenberg, S., 1988. Hierarchical classes: model and data analysis. Psychometrica 53,361–381.

Fennema, M., 1997. Some conceptual issues and problesm in the comparison of anti-immigrant partiesin Western Europe. Party Politics 3, 473–492.

Franklin, M., Niemi, R., Whitten, G., 1994. The two faces of tactical voting. British Journal of PoliticalScience 24 (4), 549–557.

Frognier, A., Gollinge, M., 1984. La Proble´matique des “Mondes Sociologiques” en Belgique. Les Cahiersdu CAGES 114.

Heath, A., Evans, G., 1994. Tactical voting: concepts, measurement and findings. British Journal of Polit-ical Sciene 24 (4), 557–561.

Huyse, L., 1980. De Gewapende Vrede. Kritak, Leuven.Ignazi, P., 1996. The intellectual basis of right-wing anti-partyism. European Journal of Political Research

29 (3), 279–296.Ignazi, P., Ysmal, C., 1992. Extreme right-wing parties in Europe. European Journal of Political

Research 22.ISPO/PIOP, 1995. General Election Study Belgium, Codebook and Questionnaire. Leuven, Louvain-la-

Neuve.Ivaldi, G., 1999. L’extre`me droite ou la crise des syste`mes de partis. Revue Internationale de Politique

Compare´e 6 (1), 201–246.

241M. Swyngedouw / Electoral Studies 20 (2001) 217–241

Kitschelt, H., 1995. The radical right in Western Europe. University of Michigan Press, Michigan.Lipset, M., Rokkan, S., 1967. Cleavage structures, party systems and voter alignment. In: Lipset, M.,

Rokkan, S. (Eds.), Party Systems and Voter Alignments. Free Press, New York.Maddens, B., Beerten, R., Billiet, J., 1994. 0 dierbaar Belgie¨? Het Natiebewustzijn van Vlamingen en

Walen, ISPO/SOl, Leuven.Mayer, N., 1996. The National Front vote and right-wing extremism, 1988–1995. Research on Democracy

and Society 3, 197–222.Mayer, N., Perrineau, P., 1992. Why do they vote Le Pen? European Journal of Political Research 22

(1), 123–141.Muller-Rommel, F., 1995. The electoral success of anti-establishment parties in Western Europe: a frame-

work for comparative explanation. Paper presented at the Conference on ‘Party Politics in the year2000’, Manchester, not published.

Poguntke, T., Scarrow, S., 1996. The politics of anti-party sentiment: introduction. European Journal ofPolitical Research 29 (3), 257–262.

Popping, R., 1995. User’s Manual AGREE 6.0, Computing Agreement on Nominal Data. Pro-GAMMA, Groningen.

Schuman, H., Presser, S., 1981. Questions and Answers in Attitude Surveys. Experiments on QuestionForm, Wording and Context. Academic Press, New York.

Storms, G., Van Mechelen, I., De Boeck, P., 1994. Structural analysis of intention and extention ofsemantic concepts. European Journal of Cognitive Psychology, 6 (1), 43–75.

Sudman, S., Bradburn, N., 1974. Response Effects in Surveys. A Review and Synthesis. Aldine, Chicago.Swyngedouw, M., 1995. Les nouveaux clivages dans la politique flamande. Etude empirique. Revue

francaise de science politique 5 (45), 775–790.Swyngedouw, M., Billiet, J., Carton, A., 1992. Van Waar Komen Ze, Wie Zijn Ze? Stemgedrag en

Verschuivingen op 24 november 1991, Leuven, ISPO-Bulletin 3.Swyngedouw, M., De Winter, L., Schulpen, L., 1990. De Verkiezingen voor het Europees Parlement

1989. Dimarso-Gallup, Brussel.Swyngedouw, M., 1993. Nieuwe Breuklijnen in de Vlaamse Politiek, De Politieke Ruimte van de l8- tot

65-jarige Vlaamse Kiezer na de Verkiezingen van 24 November 1991. In: Swyngedouw, M., Billiet,J., Carton, A., Beerten, R. (Eds.), Kiezen is Verliezen: Onderzoek naar de Politieke Opvatting vanVlamingen. Acco, Leuven/Amersfoort.

Swyngedouw, M., 1998. The extreme right in Belgium - of a non-existing front national and an omnipres-ent Vlaams Blok. In: Betz, G.H., Immerfall, S. (Eds.), The New Politicis of the Right. Neo-populistParties and Movements in Established Democracies. St. Martin’s Press, New York.

Swyngedouw, M., 1992. Waar voor je Waarden. ISPO/SOI, Leuven.Thomassen, J., 1976. Party identification as a gross-national concept: its meaning in the Netherlands. In:

Budge, I., Crewe, I., Farlie, D. (Eds.), Party Identification and Beyond. Representations of Voting andParty Competition. John Wiley, New York.

Van Holsteyn, J., 1994. Het woord is aan de kiezer, doctoraatsthesis, Rijksuniversiteit Leiden. DSWO-Press.

Wilson, T., Lafleur, S., Anderson, D., 1996. The validity and consequences of verbal reports about atti-tudes. In: Schwarz, N., Sudman, S. (Eds.), Answering Questions. Methodology for Determining Cogni-tive and Communicative Processes in Survey Research. Jossey Bass, San Francisco, 91–114.

Zaller, J.R., 1992. The Nature and Origins of Mass Opinion, Cambridge University Press.