14
Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980e2002 Harald Schoen Johannes Gutenberg-Universitat Mainz, Department of Political Science, Saarstraße 21, D-55099 Mainz, Germany Abstract This article examines the effect of German federal election campaigns on citizens’ attitudes toward chancellor candidates. Building on previous research, it puts forward three hypotheses tested using survey data gathered during seven German federal election campaigns. The results confirm that campaigns polarize the voters’ perception of the chancellor candidates. During cam- paigns, voters also bring their opinions about candidates into line with their partisan attitudes. Moreover, during three campaigns candidate preferences become increasingly powerful predictors of vote choice. The findings also suggest that election-specific fac- tors condition campaign effects. The paper concludes by discussing the implications of these findings for campaigns and political science. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: Campaign effects; Candidate evaluations; Priming; Attitude consistency; Voting behaviour; Germany 1. Introduction 1 Candidates have received a great deal of attention in the political science literature due to the leading role they play in electoral politics. Campaign researchers have primarily studied the role candidates play in cam- paign communications (e.g. Swanson and Mancini, 1996). Moreover, scholars have examined candidate strategies in electoral campaigns (e.g. Jacobs and Shapiro, 1994; Druckman et al., 2004). In another sub- field, researchers have analysed voters’ attitudes toward candidates. They have identified several dimensions of candidate perception and have studied how these subdimensions affect overall candidate evaluations (e.g. Miller and Wattenberg, 1985; Kinder, 1986; Bean and Mughan, 1989; Pancer et al., 1999; Funk, 1999; Peterson, 2005). Moreover, many studies have shown that candidate appraisal affects voting behaviour and that candidate effects on vote choice vary across in- stitutional and political conditions (e.g. Granberg and Holmberg, 1988; Vetter and Gabriel, 1998; Brettsch- neider, 2001, 2002). These findings have considerably improved our understanding of how candidates conduct campaigns, how voters perceive and evaluate candi- dates and how candidate appraisal affects vote choice. However, this brief review of the literature also points to some limitations in the previous research. Political science has rarely studied effects of candidate strategies on voters’ attitudes toward candidates. In spite of a few recent studies (see e.g. Mendelsohn, 1996; Gidengil et al., 2002; Druckman, 2004), there continues to be Tel.: þ49 6131 39 3450; fax: þ49 6131 39 2996. E-mail address: [email protected] 1 The author is grateful to the anonymous reviewers, Geoffrey Evans and Christopher Wendt for helpful comments. 0261-3794/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.electstud.2006.06.009 Electoral Studies 26 (2007) 324e337 www.elsevier.com/locate/electstud

Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

Embed Size (px)

Citation preview

Page 1: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

Electoral Studies 26 (2007) 324e337www.elsevier.com/locate/electstud

Campaigns, candidate evaluations, and vote choice: Evidencefrom German federal election campaigns, 1980e2002

Harald Schoen�

Johannes Gutenberg-Universitat Mainz, Department of Political Science, Saarstraße 21, D-55099 Mainz, Germany

Abstract

This article examines the effect of German federal election campaigns on citizens’ attitudes toward chancellor candidates.Building on previous research, it puts forward three hypotheses tested using survey data gathered during seven German federalelection campaigns. The results confirm that campaigns polarize the voters’ perception of the chancellor candidates. During cam-paigns, voters also bring their opinions about candidates into line with their partisan attitudes. Moreover, during three campaignscandidate preferences become increasingly powerful predictors of vote choice. The findings also suggest that election-specific fac-tors condition campaign effects. The paper concludes by discussing the implications of these findings for campaigns and politicalscience.� 2006 Elsevier Ltd. All rights reserved.

Keywords: Campaign effects; Candidate evaluations; Priming; Attitude consistency; Voting behaviour; Germany

1. Introduction1

Candidates have received a great deal of attention inthe political science literature due to the leading rolethey play in electoral politics. Campaign researchershave primarily studied the role candidates play in cam-paign communications (e.g. Swanson and Mancini,1996). Moreover, scholars have examined candidatestrategies in electoral campaigns (e.g. Jacobs andShapiro, 1994; Druckman et al., 2004). In another sub-field, researchers have analysed voters’ attitudes towardcandidates. They have identified several dimensions ofcandidate perception and have studied how these

� Tel.: þ49 6131 39 3450; fax: þ49 6131 39 2996.

E-mail address: [email protected] The author is grateful to the anonymous reviewers, Geoffrey

Evans and Christopher Wendt for helpful comments.

0261-3794/$ - see front matter � 2006 Elsevier Ltd. All rights reserved

doi:10.1016/j.electstud.2006.06.009

subdimensions affect overall candidate evaluations(e.g. Miller and Wattenberg, 1985; Kinder, 1986;Bean and Mughan, 1989; Pancer et al., 1999; Funk,1999; Peterson, 2005). Moreover, many studies haveshown that candidate appraisal affects voting behaviourand that candidate effects on vote choice vary across in-stitutional and political conditions (e.g. Granberg andHolmberg, 1988; Vetter and Gabriel, 1998; Brettsch-neider, 2001, 2002). These findings have considerablyimproved our understanding of how candidates conductcampaigns, how voters perceive and evaluate candi-dates and how candidate appraisal affects vote choice.However, this brief review of the literature also pointsto some limitations in the previous research. Politicalscience has rarely studied effects of candidate strategieson voters’ attitudes toward candidates. In spite of a fewrecent studies (see e.g. Mendelsohn, 1996; Gidengilet al., 2002; Druckman, 2004), there continues to be

.

Page 2: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

325H. Schoen / Electoral Studies 26 (2007) 324e337

a shortage of literature dealing with linkages betweencampaigns and candidate effects on voting behaviour.

This article aims to shed light on how election cam-paigns affect candidate evaluations. Building on previ-ous research, it puts forward hypotheses that concerncorrelations between attitudes toward different candi-dates, relationships between attitudes toward candidatesand parties, and the effect of candidate evaluations onvoting behaviour. The paper then tests these hypothesesusing survey data gathered during German federal elec-tion campaigns. The empirical evidence confirms thatattitudes toward candidates become more polarized dur-ing campaigns. Moreover, voters bring candidate per-ceptions into line with partisan attitudes duringcampaigns, though on a limited scale. The analysisalso confirms candidate priming during three cam-paigns, i.e. candidate evaluations become increasinglypowerful predictors of vote choice. The paper con-cludes with a discussion of implications of the findingsfor campaigns and political science.

2. Theoretical framework

As in many political systems, candidates for head ofgovernment play a leading part in election campaigns inGermany (e.g. Wattenberg, 1991; Brettschneider, 2001,2002). Chancellor candidates speak in election meet-ings, are the main figures in TV ads, and receive a greatdeal of attention in mass media’s campaign coverage.Campaigns thus offer voters an opportunity to learnmuch about the leading candidates (e.g. Fishbein andCoombs, 1974: 116e119; Fishbein and Ajzen, 1981:279e302). By getting to know the candidates, votersmay change their prior evaluations. For example, ifa voter learns that she strongly disagrees with a candi-date on a very important issue, she is likely to increas-ingly dislike that candidate. This paper does not dealwith campaign effects on marginal distributions of atti-tudes toward candidates for chancellor, however. In-stead, it focuses on whether campaigns affect howstrongly attitudes toward competing candidates are cor-related and how strongly opinions about candidates arecorrelated with attitudes toward political parties. It alsoaddresses whether campaigns are successful in influ-encing how strongly candidate evaluations affect votechoice. Three hypotheses are discussed in turn.

Campaigns are designed to communicate the candi-dates’ rivalry for one position. As candidates for chan-cellor are increasingly presented as competitors foroffice during campaigns, voters should increasinglyperceive candidates as opponents of one another (Mar-kus, 1982: 539e546; Finkel, 1989; Bowler et al., 1992:

208e219). Moreover, candidates for chancellor standfor competing political parties that many citizens iden-tify with. Thus, long-standing party loyalties influencevoter attitudes toward candidates. As previous researchhas shown, the effects of party identification on candi-date evaluations increase during campaigns (e.g. Laz-arsfeld et al., 1944; Berelson et al., 1954: 285; Searsand Chaffee, 1979; Markus, 1982; Miller and Shanks,1982; Gelman and King, 1993), so that in turn opinionsabout candidates become more polarized. To put itsomewhat differently, as a campaign progresses, posi-tive evaluations about a candidate should be increas-ingly accompanied by a poor evaluation about his rival.

How strongly campaigns polarize attitudes towardcandidates should depend on two factors. First, the de-sign and tone of the campaign itself may play a role inconditioning polarization effects. A campaign that doesnot work toward portraying the candidates and theirparties as incompatible rivals cannot provoke a strongpolarization of voter attitudes. But even a campaign,which depicts candidates as adversaries, may increasepolarization of candidate evaluations only mildly ifthe candidates presented themselves as rivals even be-fore the campaign. Therefore, a crucial variable is thedifference between candidates’ presentation beforeand during the campaign. Candidate evaluations aremost likely to become increasingly polarized duringa campaign if candidates have not presented themselvesas adversaries in the run-up to the campaign, and if thecampaign is designed to achieve this effect.

Second the pervasiveness of partisan loyalties maycondition polarization effects. As party identificationshapes voter attitudes (Campbell et al., 1960: 128e136),in an electorate made up of party identifiers, leadingpoliticians of competing parties are persistently per-ceived as adversaries, so that campaigns may increasethe polarization of candidate evaluations only mildly,if at all. The pervasiveness of party identifications isthus a limiting factor for the polarization effect of cam-paigns. As mentioned above, however, campaigns mayraise the impact of partisan loyalties on attitudes towardcandidates so that party identifiers may increasinglyperceive candidates for chancellor as opponents (Gel-man and King, 1993). Thus, pervasive party attach-ments may limit rather than preclude polarizationeffects of campaigns on candidate evaluations.

Next, I deal with campaign effects on how stronglyopinions about candidates are correlated with attitudestoward political parties. By means of campaign commu-nications voters can be informed about a candidate’sparty affiliation. For example, in a parliamentary sys-tem a candidate for prime minister may emphasize his

Page 3: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

326 H. Schoen / Electoral Studies 26 (2007) 324e337

party membership, as his success is contingent on hisparty attaining a parliamentary majority. During cam-paigns voters may thus perceive a candidate more andmore as his party’s choice for office and as being op-posed to other parties, particularly to his chief rival’sparty. Put another way, this hypothesis predicts thata good opinion about a candidate will be increasinglyaccompanied by favourable attitudes toward her partyand by negative attitudes toward her rival’s party.

As with the polarization effect, the process of bring-ing candidate evaluations and partisan attitudes into linewith one another may be conditioned by the pervasive-ness of deep-seated partisan loyalties. As party loyaltiescolour political perceptions, party identifiers have anaffinity to like ‘‘their’’ parties’ politicians and tend toreject politicians representing other parties even priorto an election campaign (e.g. Campbell et al., 1960:128e136). Thus, in an electorate made up of dyed-in-the-wool party identifiers, there is little room for thepolarization of attitudes toward candidates duringcampaigns. As mentioned above, however, partisancampaigns may increase the impact of deep-seatedparty identifications on candidate evaluations so thatthe latter may become more strongly correlated withpartisan attitudes even among party identifiers. There-fore, party loyalties limit the effect on relationships be-tween attitudes toward candidates and parties, but theydo not nullify them.

Political communication may play a conditioningrole, as well. A strong campaign effect on correlationsbetween candidate evaluations and partisan attitudesis likely to occur if a nominated politician has not yetacted as a party politician or has antagonised ‘‘his’’party in the past, but now acts as a ‘‘party soldier’’.By contrast, campaigns that do not depict candidatesas opponents and as ‘‘party champions’’ will not exhibitlarge effects on the relationships between candidateevaluations and opinions about parties. What is more,campaigns are not likely to increase correlations be-tween opinions about candidates and parties if candi-dates are widely identified with certain parties even inoff-election times; the former leader of the British Con-servative Party, Margaret Thatcher, and the formerleader of the German CDU, Helmut Kohl, are cases inpoint.

Building on the idea that political communicationscan ‘‘prime’’ voters, I now deal with campaign effectson how strongly candidate evaluations influence votingbehaviour (e.g. Iyengar and Kinder, 1987; Krosnick andKinder, 1990). This idea assumes that there are no fixedcriteria voters rely on when deciding whom to vote for.Instead, voters are assumed to have varied impressions

and perceptions of parties and candidates. When a per-son casts her vote, she uses only those criteria which areeasily available or ‘‘top of the head’’, as Zaller (1992)put it. Whether a criterion is available depends on, interalia, how frequently it has been activated (see e.g. Lau,1989; Valentino et al., 2002). Accordingly, a voter whofrequently considers candidates when thinking aboutvote choice before the election is likely to rely on can-didate evaluations when voting. Which considerationscome to mind before an election depends on, amongother things, the content of political communication,so that voters frequently consider candidates duringcandidate-centered campaigns. As a result, candidate-centered campaigns may increase the impact of candi-date evaluations on voting behaviour (see e.g. Johnstonet al., 1992: 141e167; Mendelsohn, 1996; Gidengilet al., 2002; Popescu and Toka, 2002; Druckman, 2004).

Since chancellor candidates play a leading role inGerman election campaigns, one may expect cam-paigns to considerably increase the impact of candidateevaluations on voting behaviour. This expectation needssome qualification. First, the personalization of poli-tical communication is not unique to campaigns, aspolitical leaders loom large in politics outside of elec-tions. Second, German voters are asked to cast votesnot for candidates but for parties and parties play amajor role in campaign communication. Some votersmay thus consider candidates as party mouthpiecesand vote choice as being chiefly about political parties,so that during campaigns, party attachments increas-ingly affect both candidate evaluations and vote inten-tion (see e.g. Finkel and Schrott, 1995; Miller andShanks, 1996: 190e193; Weßels, 2000). As a con-sequence, depending on to how many voters this appliescandidate priming will be limited during Germanfederal election campaigns.

To be primed by campaign communications a votermust have prior party attachments. Consequently, inan electorate made up of party supporters party primingis likely to occur. In contrast, in an electorate with weakor non-existent partisan loyalties, party priming will beminimal while candidate priming is likely to occur. Be-side the pervasiveness of party attachments, the contentof campaign communication should matter, as well. Theeffect of candidate evaluations on vote choice should in-crease during a campaign if political communicationbefore the campaign is mainly focused on parties, butthe campaign focuses on candidates; by the same token,if parties were completely irrelevant in pre-campaignpolitical communication, a campaign that focuses onparties will increase the impact of party identificationon vote choice.

Page 4: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

327H. Schoen / Electoral Studies 26 (2007) 324e337

Wrapping things up, I expect campaigns to exhibitthree effects. First, campaign communication shouldpolarize candidate evaluations. Second, my theorypredicts that during campaigns voters will perceivea candidate more and more as his party’s choice foroffice and as being opposed to other parties, particularlyto his chief rival’s party. Third, candidate evaluationsshould become increasingly powerful predictors ofvote choice during campaigns. Moreover, the strengthof these effects should be conditioned by some politicalfactors.

3. Hypotheses for German federal electioncampaigns, 1980e2002

In Germany, candidates for chancellor play a consid-erably greater role in election campaigns than in off-election periods (e.g. Hetterich, 2000). However, cam-paigns are less personalized than those in the UnitedStates, as federal elections are held under a system ofproportional representation with party lists (e.g. Farrell,2001: 169e170). Moreover, in Germany politicalparties are considerably stronger than in many othercountries and play a critical role during both campaignsand off-elections periods. This brief overview of the po-litical setting in Germany lends support to the expecta-tion that campaigns will considerably increase bothnegative correlations between candidate evaluationsand effects of attitudes toward candidates on votechoice. At the same, the partisan nature of politicsmay limit campaign effects on the relationship of candi-date evaluations to partisan attitudes.

As discussed in previous sections, the content of po-litical communication and the pervasiveness of party at-tachments should condition campaign effects. To obtainhypotheses on the effects in German federal campaigns,it is thus necessary to take a closer look at election cam-paigns and partisan loyalties. Since the 1970s, Germanyhas experienced a slow but steady decline in party at-tachment (Zelle, 1998; Schoen and Weins, 2005;

Arzheimer, 2006), so that the number of voters consid-ering candidates as adversaries even in off-election pe-riods should have decreased. Thus, campaign effects oncorrelations between attitudes toward competing candi-dates and between candidate evaluations and partisanattitudes are likely to have increased. At the sametime, candidate priming is likely to have becomestronger.

In German campaigns, only the two major parties’(the SPD and CDU/CSU) field candidates with a reason-able probability of becoming Chancellor. I thus concen-trate on discussing candidates and campaigns of SPDand CDU/CSU. The very emotional and heated 1980campaign was dominated by the rivalry between verydissimilar candidates. Incumbent chancellor HelmutSchmidt campaigned for the social-democratic SPD,while Franz Josef Strauss ran for the conservativeCDU/CSU. Both were well-known federal politicians(see Table 1). The campaign preceding the election inMarch 1983 was rather brief, following a premature dis-solution of parliament. Incumbent CDU/CSU chancel-lor Helmut Kohl was a well-known federal politician,whereas SPD candidate Hans-Jochen Vogel was ratherunknown to the general public. In the 1986/1987 cam-paigns, Kohl stood for reelection against JohannesRau, well-known leading Social Democrat and primeminister of Northrhine-Westfalia, Germany’s largestfederal state. In 1990 Kohl, celebrated as the ‘‘reunifica-tion chancellor’’, was challenged by SPD-politician Os-kar Lafontaine, well-known as a party maverick. In the1994 campaign, which lacked polarizing topics to elicitpassionate debates, Kohl’s challenger Rudolf Scharpingwas not very prominent as a party politician at the outsetof the campaign. In the 1998 campaign, there was notmuch of an emotional dispute about the major issues.Kohl’s challenger, Gerhard Schroder, was well-knownright from the beginning of the campaign and, further-more, was reputed as a headstrong master of strategy,though not always in line with his party’s decisions.When Schroder campaigned for reelection in 2002, he

Table 1

Background information about the candidates for chancellor in German federal elections, 1980e2002

CDU/CSU SPD

Candidate Chancellor since Party leader since Candidate Chancellor since Party leader since

1980 Strauss e 1961 (CSU) Schmidt 1974 e

1983 Kohl 1982 1973 Vogel e e

1987 Kohl 1982 1973 Rau e e1990 Kohl 1982 1973 Lafontaine e e

1994 Kohl 1982 1973 Scharping e 1993

1998 Kohl 1982 1973 Schroder e e

2002 Stoiber e 1999 (CSU) Schroder 1998 1999

Page 5: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

328 H. Schoen / Electoral Studies 26 (2007) 324e337

had less of an opportunity to differentiate himself fromthe SPD having become party chair in 1999. His chal-lenger Edmund Stoiber was a well-known conservativepolitician in the CDU/CSU alliance.

In all seven campaigns the CDU/CSU nominatedwell-known, leading federal politicians. The SPD didso in 1980, 1987 and 2002. The SPD candidates nomi-nated in 1983 and 1994, however, became known toa wider public as champions of their parties only aftertheir nomination. In 1990 and 1998, the Social Demo-crats’ candidate was widely known as ‘‘party outsider’’.As a consequence, I expect rather strong polarizationeffects to occur during the 1983, 1994 and 1998 cam-paigns. While the party quickly closed ranks behindSchroder in 1998, the continued inner party controversyover Lafontaine’s 1990 candidature should result inno polarization in 1990. The 1980, 1987, and 2002campaigns should also exhibit no polarization effect,as both the CDU/CSU and SPD nominated well-knownpoliticians for these campaigns.

Now I turn to the relationship of candidate evalua-tions to partisan attitudes. To begin with, as two verywell-known party politicians strove for office in 1980,I anticipate the relationship between attitudes towardcandidates and partisan attitudes to be strong at the startof this campaign and to remain stable during thecampaign. As the Christian Democrats nominatedwell-known representatives in the remaining six elections,I anticipate these campaigns not to raise correlationsbetween opinions about CDU/CSU candidates andpartisan attitudes. As the SPD nominated well-knownpoliticians in 1987 and 2002, I do not expect these cam-paigns to increase the relationship between opinionsabout SPD candidates and partisan attitudes. Regardingthe 1983, 1990, 1994, and 1998 campaigns, the SocialDemocrats nominated candidates who had not madea name for themselves or were known not to be ‘‘partysoldiers’’. Therefore, I expect the relationship betweenopinions about SPD candidates and partisan attitudes tobe weak at the start of these campaigns and to increaseconsiderably during campaign periods.

Partisan dealignment, the rise of television (distinc-tively apt to personalize politics) and the trend towardcandidate-centered campaigns lend support to the ideaof candidate priming having steadily increased at theexpense of party priming (see e.g. Holtz-Bacha andKaid, 1995: 10; Swanson and Mancini, 1996; Norris,1997). This expectation does not go unchallenged, how-ever. In 1983, 1990, 1994, and 1998 the Social Demo-crats nominated candidates who were not well-knownas champions of their party, while in 1980, 1987, and2002 they did not. Accordingly, only the former

campaigns should be expected to have increased candi-date effects on vote choice.

4. Data and measures

The data in my analysis come from Politbarometera monthly poll of roughly 1000 eligible voters con-ducted by the Forschungsgruppe Wahlen since 1977.2

As my analysis aims at examining campaign effects,only surveys conducted during campaign periods wereincluded, with ‘‘campaign’’ being defined as the inter-val between the nomination of the candidates for chan-cellor and the election.3 In 1988, the survey companychanged the interview mode. Prior to August 1988, re-spondents were selected from the population eligible tovote by multi-stage probability sampling and were in-terviewed face-to-face. From 1990 on, respondentswere interviewed by telephone. In 2002, the interval be-tween consecutive interviews was shorter than previ-ously, with a sample taken fortnightly.

To measure attitudes toward candidates and parties Idraw on 11-point feeling thermometer ratings.4 Fromthe candidate ratings I also constructed variables to cap-ture candidate preference, with the coding depending onthe dependent variable (SPD vote: �1¼ preference forCDU/CSU candidate, 0¼ no candidate preference,

2 For the sake of comparability, the analysis is confined to the West

Germany throughout the whole period from 1980 to 2002. Data sets

were provided by the Zentralarchiv fur Empirische Sozialforschung,

Cologne (ZA). The study numbers are: 1053, 2182, 1276, 1536,

1899, 1920, 2378, 2559, and 3160 (see for details http://www.ge-

sis.org/Datenservice/Wahlstudien/Btw/dnw_studien.htm). Data sets

on the 2002 election were provided by the ZA in a preliminary ver-

sion and were prepared for analysis by Siegfried Buhler and Henning

Laux. Data of the following calendar weeks were used: 3, 8, 10, 12,

15, 17, 19, 22, 24, 26, 28, 30, 32, 33, 34, 35, 36, 37, and 38. Neither

depositors nor the data archive are responsible for any of the reported

analyses or for the interpretation of the outcoming results.3 I include all interviews conducted after candidate nomination, re-

gardless of when politicians announce that they start campaigning

since politicians frequently campaign without announcing it to the

public (see e.g. Shaw, 1999; Druckman et al., 2004). More precisely,

as in all elections from 1980 to 2002 incumbent chancellors cam-

paigned for reelection the analyses start after the challengers’ nomi-

nation. Strauss was nominated on July 2nd 1979, Vogel in October

1982, Rau in December 1985, Lafontaine in January 1990, Scharping

in June 1993, Schroder formally in April 1998, actually, however, at

the state election in Lower Saxony on March 1st 1998, and Stoiber

was nominated on January 11th 2002. Obviously, campaign periods

in this analysis vary in length across elections. Hence, it may be ob-

jected that inter-election differences in campaign effects are a func-

tion of campaign length. The below analyses do not corroborate this

suspicion, however.4 To measure opinions about CDU/CSU CDU and CSU ratings

were averaged.

Page 6: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

329H. Schoen / Electoral Studies 26 (2007) 324e337

Table 2

Effect of attitudes toward CDU/CSU candidates on attitudes toward SPD candidates during German federal election campaigns, 1980e2002 (un-

standardized OLS-coefficients; robust standard errors in parentheses)

Constant CDU/CSU candidate Time Time� candidate R2� 100 N

1980 10.09��� (0.07) �0.19��� (0.01) 0.55��� (0.12) �0.13��� (0.02) 14.0 14 880

1983 8.81��� (0.17) �0.19��� (0.02) 0.57� (0.27) �0.15��� (0.04) 11.7 3695

1987 8.78��� (0.10) �0.24��� (0.02) �0.35 (0.18) �0.03 (0.03) 10.7 12 242

1990 9.66��� (0.13) �0.24��� (0.02) 0.29 (0.23) �0.12��� (0.03) 12.2 8729

1994 7.10��� (0.11) 0.03 (0.02) 0.91��� (0.20) �0.21��� (0.03) 2.1 12 292

1998 8.07��� (0.14) �0.06�� (0.02) 0.73�� (0.22) �0.19��� (0.03) 4.9 6823

2002 7.67��� (0.14) �0.07��� (0.02) 1.72��� (0.19) �0.18��� (0.03) 5.0 17 217

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

þ1¼ preference for SPD candidate; responses werereverse-coded in analyses of voting for the CDU/CSU).To evaluate the effects of party identification I drewon the standard indicator (see e.g. Falter et al., 2000)and constructed two variables, one for analyses of atti-tudes toward the SPD or voting for this party(1¼ strong SPD identifiers, 0.5¼weak SPD identifiers,0¼ adherents of other parties and independents), theother for analyses of opinions about CDU/CSU or cast-ing votes for it (1¼ strong CDU/CSU identifiers,0.5¼weak CDU/CSU identifiers, 0¼ adherents ofother parties and independents).5 To measure votechoice I constructed an SPD variable (1¼ vote forSPD; 0¼ vote for another party, unsure about whomto vote for, does not want to vote) and a CDU/CSU vari-able. Prospective non-voters and undecided voters wereincluded as campaigns may strongly affect persons whoare undecided irresolute or unwilling to vote (e.g. Laz-arsfeld et al., 1944). Finally, for each pooled data set Iconstructed a variable capturing when the interviewswere conducted. It ranges from 0 indicating the first sur-vey within a campaign to 1 referring to the last survey.

5. Findings

5.1. Polarization of candidate evaluations

To examine the effects of campaign communicationon the correlation between attitudes toward competingcandidates, I regressed opinions about one candidateon attitudes toward his rival using OLS-regression. Ad-ditionally, I included time and a candidate evaluation-by-time interaction term. In Tables 2 and 3, the coeffi-cients for candidate evaluation indicate how strongly

5 Since the German multi-party system cannot easily be reduced

to one dimension party identification is not measured with a one-

dimensional 5- or 7-point scale of party identification as it is common

in the United States. Rather, intensity of party identification with

CDU/CSU and SPD is measured with two separate variables (see

e.g. Green et al., 2002: 190).

the opinion about one candidate affects the attitude to-ward his rival at the outset of the respective campaign.Regarding polarization effects, coefficients for the mul-tiplicative term are crucial. Positive coefficients indi-cate that positive evaluations of one candidate areincreasingly accompanied by positive attitudes towardhis rival; by the same token, negative coefficients indi-cate increasingly negative (or decreasingly positive)correlations between candidate evaluations. Finally, toanswer how strongly evaluations of one candidate arecorrelated with attitudes toward his rival, the coeffi-cients for the main candidate term and for the multipli-cative term are simply added (Jaccard, 2001).6

Looking at the results in Tables 2 and 3, we can seethat from 1980 to 1990, attitudes toward candidateswere considerably polarized when candidates startedcampaigning. By contrast, from 1994 on, the polariza-tion of opinions about candidates was almost unnotice-able at the beginning of the campaigns. This finding isin line with the idea of partisan dealignment has reducedthe level of initial polarization of candidate evaluations.The results also suggest that elite actions may be play-ing a noticeable role. For example, the minimal correla-tions between candidate evaluations in 1994 coincidewith the nomination of a not very well-known politicianby the Social Democrats. Regardless of its causes, cor-relations between attitudes toward competing candi-dates had considerably more room to increase duringcampaigns after 1990 than before that year.

6 This design is apt for examining whether correlations change

from the beginning of a campaign until its end although it is not ap-

propriate to study temporary variations. It is also not capable of iden-

tifying which factors were how influential in changing the dependent

variable. Furthermore, though the data this paper drew on allow me

to compare effects from one election to another, due to the small

number of elections included, effects of campaign-specific factors

cannot be determined adequately. Finally, including interaction terms

is likely to increase the level of collinearity. As statistical tests con-

firmed, however, there were no serious collinearity problems in the

analyses reported below.

Page 7: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

330 H. Schoen / Electoral Studies 26 (2007) 324e337

Table 3

Effect of attitudes toward SPD candidates on attitudes toward CDU/CSU candidates during German federal election campaigns, 1980e2002 (un-

standardized OLS-coefficients; robust standard errors in parentheses)

Constant SPD candidate Time Time� candidate R2� 100 N

1980 10.32��� (0.24) �0.44��� (0.03) 1.13�� (0.39) �0.21��� (0.04) 14.2 14 880

1983 9.56��� (0.29) �0.34��� (0.04) 0.54 (0.37) �0.07 (0.05) 10.6 3695

1987 8.88��� (0.17) �0.39��� (0.02) 0.25 (0.28) 0.00 (0.04) 10.3 12 242

1990 9.64��� (0.18) �0.33��� (0.02) 1.02��� (0.28) �0.06 (0.04) 11.9 8729

1994 4.98��� (0.21) 0.03 (0.03) 3.79��� (0.32) �0.29��� (0.04) 4.3 12 292

1998 6.33��� (0.29) �0.10�� (0.04) 2.74��� (0.43) �0.28��� (0.05) 5.0 6823

2002 7.53��� (0.18) �0.08��� (0.02) 1.63��� (0.26) �0.24��� (0.03) 4.8 17 217

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

As the coefficients for the interaction terms inTables 2 and 3 indicate, there is some evidence in sup-port of the polarization hypothesis. In 1980, althoughcandidate evaluations were considerably negativelycorrelated at the start of the campaign, attitudes towardSchmidt and Strauß became even stronger polarizedup to the election. During the 1983 and 1990 cam-paigns, candidate evaluations became somewhat morepolarized, while the 1987 campaign exhibited nopolarization effect. By and large, from 1994 on, polari-zation effects were somewhat stronger than in the1980s. At the same time, from 1994 on candidateevaluations were mildly correlated (if at all), whencandidates started campaigning. In contrast, prior to1994, initial correlations between attitudes towardcompeting candidates were sizeable. Thus, the findingssuggest that in the 1980s, West Germans’ perceivedcandidates to be opponents of each other at the startof campaigns whereas from 1994 on, candidate evalua-tions were strongly polarized only at the end ofcampaigns.

5.2. Party politicization of candidate evaluations

To examine whether voters bring candidate evalua-tions and partisan attitudes in line I regressed candidateevaluations on partisan attitudes, time, and a partisan

attitudes-by-time interaction term. The results are re-ported in Tables 4 and 5.

The rather strong main effects for partisan attitudesindicate that in all seven elections, opinions about can-didates and parties were strongly correlated when can-didates started campaigning. As hypothesized, initialeffects for CDU/CSU candidates were stronger thanthose for SPD candidates. Moreover, the initial correla-tion between attitudes toward Schroder and the SPD in-creased between 1998 and 2002 significantly. AsSchroder became chairman of the SPD in 1999 andtherefore had less opportunity to oppose his party, thisfinding suggests that political events play a role in influ-encing initial correlations. The finding that in 1983 and1994, when the SPD fielded not very well-known candi-dates, the initial candidate-SPD association was quiteweak also lends support to this idea.

In line with my hypothesis, the candidate-CDU/CSUassociation did not increase appreciably in any cam-paign. One may object to this conclusion, as Table 5shows statistically significant interaction effects in1994 and 2002. This objection does not stand up tocloser scrutiny, however, as these two effects are irrele-vant in substantive terms: they are tiny as compared tothe respective main effects of partisan attitudes. The re-sults on the correlations between opinions about theSPD and its candidates are more mixed. During the

Table 4

Effect of attitudes toward SPD on opinions about SPD candidates during German federal election campaigns, 1980e2002 (unstandardized

OLS-coefficients; robust standard errors in parentheses)

Constant SPD-sympathy Time Time� SPD R2� 100 N

1980 4.52��� (0.14) 0.57��� (0.02) �1.11��� (0.24) 0.11��� (0.03) 52.9 14 828

1983 4.17��� (0.18) 0.45��� (0.02) �2.06��� (0.24) 0.24��� (0.03) 44.7 3681

1987 2.75��� (0.12) 0.62��� (0.01) �1.36��� (0.19) 0.14��� (0.02) 48.9 12 221

1990 3.59��� (0.20) 0.56��� (0.02) �1.42��� (0.30) 0.12��� (0.04) 36.7 8629

1994 4.85��� (0.16) 0.36��� (0.02) �2.97��� (0.26) 0.29��� (0.03) 28.8 12 135

1998 3.62��� (0.24) 0.53��� (0.03) �1.56��� (0.38) 0.14�� (0.04) 36.2 6728

2002 2.92��� (0.14) 0.64��� (0.02) �0.26 (0.21) 0.05� (0.03) 47.5 17 305

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

Page 8: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

331H. Schoen / Electoral Studies 26 (2007) 324e337

Table 5

Effect of attitudes toward CDU/CSU on opinions about CDU/CSU candidates during German federal election campaigns, 1980e2002 (unstandard-

ized OLS-coefficients; robust standard errors in parentheses)

Constant CDU/CSU Time Time�CDU/CSU R2� 100 N

1980 �0.30�� (0.10) 0.98��� (0.01) �0.63��� (0.16) 0.02 (0.02) 61.4 14 791

1983 1.36��� (0.13) 0.83��� (0.02) �0.15 (0.21) 0.04 (0.03) 61.6 3686

1987 0.31��� (0.07) 0.89��� (0.01) �0.01 (0.13) 0.03 (0.02) 64.7 12 257

1990 1.26��� (0.13) 0.83��� (0.02) 0.84��� (0.23) �0.02 (0.03) 56.3 8399

1994 1.19��� (0.10) 0.77��� (0.01) 0.14 (0.17) 0.06�� (0.02) 53.2 12 811

1998 0.88��� (0.12) 0.77��� (0.02) �0.06 (0.22) 0.03 (0.03) 50.0 6569

2002 1.54��� (0.11) 0.79��� (0.01) �0.71��� (0.16) 0.06�� (0.02) 60.0 16 478

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

2002 campaign, the attitudes toward the SPD and itscandidate did increase, while in 1980, 1987 and 1998a modest increase was found. In the remaining cam-paigns, however, the SPD-candidate correlation becameconsiderably stronger. In 1983, the campaign increasedthe correlation by about 50 percent, and in 1994, the as-sociation almost doubled during the campaign. Hence,there are instances of German federal election cam-paigns increasing correlations between candidate ap-praisal and attitudes toward the candidate’s party.What is more, this effect is subject to campaign-specificvariations.

The evidence also lends support to the notion ofcampaigns rendering correlations between opinionsabout a candidate and attitudes toward his rival’s partymore negative, which may be called a negative party po-liticization of candidate attitudes. Looking at Tables 6and 7, by and large attitudes toward a candidate andhis competitor’s party were negatively correlated atthe beginning of campaigns. What is more, excepting1987 and 1990, this correlation became considerablystronger during campaigns. Thus, the results concerningnegative party polarization parallel the findings on can-didate polarization though the latter effect is somewhatstronger than the former one. This pattern suggests thatduring campaigns candidates were more and more per-ceived as representatives of their parties and this causedcandidate evaluations to become increasingly polarized.

5.3. Priming of candidate evaluations and partyattachments

While in the previous sections I dealt with campaigneffects on correlations between attitudes, I now addresswhether election campaigns affect the impact of atti-tudes on vote choice. To examine priming effects I firstregressed vote intention on candidate preference, time,and a candidate preference-by-time interaction term.The results are reported in Table 8.

As regards voting for SPD, during the 1980, 1987,1990 and 1998 campaigns, attitudes toward candidatesdid not become a stronger predictor of voting behaviour.The 1983, 1994 and 2002 campaigns, however, in-creased the effect of opinions about candidates onvote choice substantially. Turning to the decision tovote for the Christian Democrats, except 1998, candi-date evaluations became a stronger predictor of votechoice during campaigns. While the increase was rathermodest in 1980 and 1987, it was quite sizeable in the re-maining elections. In summary, excepting 1998, candi-date priming appears to have occurred in regard to atleast one party. This finding is surprising since many ex-perts regard the 1998 campaign as the most ‘American-ized’ and personalized German federal campaign in the20th century.

Next, I address if campaigns also increase the impactof party attachments on vote choice. As the results in

Table 6

Effect of attitudes toward SPD on opinions about CDU/CSU candidates during German federal election campaigns, 1980e2002 (unstandardized

OLS-coefficients; robust standard errors in parentheses)

Constant SPD-sympathy Time Time� SPD R2� 100 N

1980 10.43��� (0.16) �0.53��� (0.02) 0.79�� (0.26) �0.17��� (0.03) 23.2 14 818

1983 10.38��� (0.21) �0.45��� (0.03) 0.52 (0.28) �0.07 (0.04) 21.2 3697

1987 9.58��� (0.15) �0.48��� (0.02) 0.35 (0.26) �0.01 (0.03) 16.2 12 289

1990 9.34��� (0.19) �0.30��� (0.02) 0.88��� (0.30) �0.02 (0.04) 8.4 8789

1994 5.34��� (0.18) �0.01 (0.02) 3.73��� (0.30) �0.27��� (0.04) 4.4 13 413

1998 7.46��� (0.28) �0.25��� (0.03) 1.82��� (0.43) �0.14�� (0.05) 6.9 6852

2002 8.27��� (0.16) �0.20�� (0.02) 1.12��� (0.23) �0.18��� (0.03) 8.3 16 972

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

Page 9: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

332 H. Schoen / Electoral Studies 26 (2007) 324e337

Table 7

Effect of attitudes toward CDU/CSU on opinions about SPD candidates during German federal election campaigns, 1980e2002 (unstandardized

OLS-coefficients; robust standard errors in parentheses)

Constant CDU/CSU Time Time�CDU/CSU R2� 100 N

1980 10.70��� (0.11) �0.26��� (0.02) 0.39� (0.18) �0.08�� (0.03) 12.7 14 797

1983 8.74��� (0.18) �0.18��� (0.02) 1.07��� (0.27) �0.23��� (0.04) 13.5 3672

1987 9.18��� (0.12) �0.29��� (0.02) �0.54� (0.21) �0.01 (0.03) 10.7 12 189

1990 9.99��� (0.13) �0.29��� (0.02) �0.28 (0.23) �0.07� (0.03) 12.1 8264

1994 7.18��� (0.12) 0.01 (0.02) 1.03��� (0.21) �0.23��� (0.03) 2.6 11 641

1998 8.41��� (0.15) �0.11��� (0.02) 0.59 (0.25) �0.14��� (0.04) 5.0 6467

2002 7.78��� (0.15) �0.09��� (0.02) 1.69��� (0.20) �0.17��� (0.03) 4.9 16 740

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

Table 9 indicate, party identification turned out to bea powerful predictor of vote choice even at the outsetof all seven campaigns. Since initial effects of partyidentification on vote intention decreased somewhatfrom 1980 to 2002, one may expect a parallel increasein party priming. In regard to voting for the CDU/CSU,however, there was no indication of party priming.Turning to the SPD vote, the results were somewhatmore mixed. While in 1983, 1987, 1990 and 1998, theeffect of party identification on the intention to casta vote for the Social Democrats was constant duringcampaigns, it changed in 1980, 1994 and 2002. Duringthe 1994 campaign, the effect of party attachments onthe decision to vote for the Social Democrats increasedsomewhat. In contrast, during the 1980 and 2002 cam-paigns the effect of party identification on vote choicefor the Social Democrats dropped considerably. In sum-mary, it is hard to find evidence of party priming inGerman federal campaigns since 1980.

So far, the evidence lends considerable support to theidea of candidate priming while priming of party attach-ments appears to be non-existent. However, the evidenceis only preliminary as campaigns may have strengthenedeffects of party loyalties on attitudes toward candidates.In a further step I thus analysed whether the impact ofparty attachments on candidate preferences increasedduring campaigns. To examine this hypothesis I re-gressed candidate preferences on party identification,time and a party identification-by-time interactionterm. The results are reported in Table 10.

As the coefficients for party identification indicate,party attachments had a considerable effect on candi-date preferences at the beginnings of campaigns. The1980, 1983 and 2002 campaigns increased the effectof identifying with CDU/CSU on candidate prefer-ences, though the total result was only modest (an addi-tion of 15% at most). The lack of strong campaigneffects is unsurprising as in all cases well-known

Table 8

Effect of candidate preference on vote choice during German federal election campaigns, 1980e2002 (unstandardized logit coefficients; robust stan-

dard errors in parentheses)

Constant Candidate preference Time Time� candidate PR2� 100 N

SPD vote

1980 �1.60��� (0.12) 2.14��� (0.12) �0.24 (0.21) 0.24 (0.22) 27.5 14 880

1983 �0.48��� (0.07) 1.44��� (0.07) �0.37�� (0.12) 0.77��� (0.13) 36.6 3646

1987 �0.91��� (0.06) 1.69��� (0.06) �0.31�� (0.11) 0.16 (0.12) 31.2 11 204

1990 �0.87��� (0.06) 1.40��� (0.06) 0.11 (0.10) 0.06 (0.10) 24.6 8729

1994 �1.39��� (0.06) 1.11��� (0.07) 0.54��� (0.10) 0.37�� (0.11) 18.4 12 292

1998 �1.06��� (0.09) 1.62��� (0.09) 0.09 (0.14) �0.13 (0.15) 22.8 6823

2002 �1.39��� (0.07) 1.60��� (0.08) �0.36�� (0.11) 0.45��� (0.12) 28.0 17 217

CDU/CSU vote

1980 �0.14� (0.06) 1.83��� (0.07) �0.10 (0.10) 0.24� (0.11) 36.0 14 880

1983 �0.68��� (0.08) 1.76��� (0.08) 0.08 (0.13) 0.63��� (0.14) 43.4 3646

1987 �0.70��� (0.06) 1.91��� (0.06) �0.07 (0.11) 0.25� (0.12) 42.0 11 204

1990 �0.55��� (0.06) 1.59��� (0.06) �0.61��� (0.11) 0.32�� (0.11) 32.3 8729

1994 �1.09��� (0.06) 1.35��� (0.06) �0.20 (0.11) 0.47��� (0.11) 28.2 12 292

1998 �0.80��� (0.07) 1.62��� (0.08) �0.21 (0.12) 0.09 (0.13) 31.2 6823

2002 �0.66��� (0.05) 1.50��� (0.06) �0.11 (0.08) 0.31��� (0.08) 32.2 17 217

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

Page 10: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

333H. Schoen / Electoral Studies 26 (2007) 324e337

Table 9

Effect of party identification on vote choice during German federal election campaigns, 1980e2002 (unstandardized logit coefficients; robust stan-

dard errors in parentheses)

SPD vote Constant SPD party ID Time Time� SPD-PID PR2� 100 N

1980 �1.77��� (0.07) 6.80��� (0.30) 0.01 (0.11) �1.66��� (0.45) 48.3 13 795

1983 �1.57��� (0.09) 5.54��� (0.34) �0.23 (0.14) 1.28 (0.69) 45.8 3419

1987 �1.33��� (0.06) 5.36��� (0.28) �0.47��� (0.11) 0.86 (0.56) 37.5 10 508

1990 �1.71��� (0.07) 5.55��� (0.27) �0.18 (0.11) �0.57 (0.47) 40.4 8963

1994 �1.77��� (0.06) 3.71��� (0.18) 0.08 (0.10) 0.87�� (0.31) 29.8 13 761

1998 �1.03��� (0.06) 4.04��� (0.25) �0.26� (0.11) 0.07 (0.39) 25.6 7092

2002 �1.82��� (0.06) 4.39��� (0.18) 0.14 (0.09) �0.54� (0.25) 29.6 17 737

CDU/CSU vote Constant CDU/CSU party ID Time Time�CDU/CSU-PID PR2� 100 N

1980 �1.94��� (0.07) 7.21��� (0.38) �0.48��� (0.12) �0.61 (0.61) 54.2 13 795

1983 �1.80��� (0.09) 6.03��� (0.47) 0.24 (0.14) 0.54 (0.71) 47.8 3419

1987 �1.74��� (0.06) 6.18��� (0.31) 0.16 (0.12) 0.29 (0.64) 43.4 10 508

1990 �2.00��� (0.07) 5.62��� (0.26) 0.39�� (0.12) �0.13 (0.45) 43.0 8963

1994 �2.45��� (0.07) 5.00��� (0.21) 0.75��� (0.12) �0.53 (0.35) 34.3 13 761

1998 �2.25��� (0.09) 4.49��� (0.26) 0.49�� (0.14) �0.38 (0.42) 32.4 7092

2002 �1.44��� (0.06) 4.45��� (0.20) �0.22�� (0.08) �0.05 (0.29) 31.6 17 737

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

politicians ran for CDU/CSU. The notion that public’sfamiliarity with the candidates plays a role in condition-ing campaign effects was also supported by the resultsconcerning SPD attachments. During the 1980 and2002 campaigns, when SPD incumbent chancellorsran for reelection the effect of party identification oncandidate preference did not increase during the cam-paign; in 2002, it declined somewhat. By contrast, dur-ing the remaining five campaigns, when SPDcandidates were challengers, the effect of party loyaltieson candidate preference increased considerably. What

is more in 1983 and 1994, when the SPD candidateswere not well-known before the campaign, the effectof party identification on party preference increasedby far stronger than during the remaining campaigns.Thus, the evidence suggests that election-specific fac-tors affect how strongly campaigns influence the effectof party attachments on candidate preferences.

Having shown that by and large, party attachmentsbecome more powerful predictors of candidate prefer-ence during campaigns, I next examined whether candi-date priming was proved after controlling for party

Table 10

Effect of party identification on candidate preference during German federal election campaigns, 1980e2002 (unstandardized OLS-coefficients;

robust standard errors in parentheses)

Constant SPD party ID Time Time� SPD-PID R2� 100 N

1980 0.08��� (0.02) 1.02��� (0.03) 0.01 (0.04) �0.05 (0.05) 23.3 13 705

1983 �0.28��� (0.03) 1.09��� (0.05) �0.14�� (0.04) 0.44��� (0.07) 31.8 3415

1987 �0.12��� (0.02) 1.21��� (0.03) �0.17��� (0.03) 0.15�� (0.06) 24.4 11 353

1990 �0.11��� (0.02) 1.02��� (0.04) �0.37��� (0.04) 0.29��� (0.07) 23.8 8619

1994 0.22��� (0.02) 0.78��� (0.04) �0.58��� (0.04) 0.47��� (0.06) 19.6 12 006

1998 0.19��� (0.03) 0.82��� (0.04) �0.28��� (0.04) 0.29��� (0.06) 17.2 6691

2002 �0.21��� (0.02) 1.28��� (0.03) 0.13��� (0.03) �0.15�� (0.04) 22.3 16 837

Constant CDU/CSU party ID Time Time�CDU/CSU-PID R2� 100 N

1980 �0.72��� (0.01) 1.33��� (0.04) �0.13��� (0.02) 0.16�� (0.06) 43.0 13 705

1983 �0.38��� (0.03) 1.28��� (0.05) �0.04 (0.04) 0.19�� (0.07) 36.3 3415

1987 �0.48��� (0.02) 1.49��� (0.03) 0.10�� (0.03) �0.05 (0.05) 37.7 10 993

1990 �0.51��� (0.02) 1.31��� (0.04) 0.36��� (0.04) �0.07 (0.06) 29.5 8619

1994 �0.61��� (0.02) 1.38��� (0.04) 0.38��� (0.03) �0.06 (0.06) 27.0 12 006

1998 �0.64��� (0.02) 1.23��� (0.05) 0.21��� (0.04) 0.12 (0.08) 27.4 6691

2002 �0.32��� (0.02) 1.28��� (0.03) �0.16��� (0.03) 0.12� (0.05) 29.7 16 837

Significance levels: �p< 0.05; ��p< 0.01; ���p< 0.001.

Page 11: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

334 H. Schoen / Electoral Studies 26 (2007) 324e337

T

able

11

Eff

ecto

fca

nd

idat

ep

refe

ren

cean

dp

arty

iden

tifi

cati

on

on

vote

cho

ice

du

rin

gG

erm

anfe

dera

lele

ctio

nca

mpa

ign

s,1

98

0e2

00

2(u

nst

and

ard

ized

log

itco

effi

cien

ts;r

obu

stst

anda

rder

rors

inp

aren

thes

es)

SP

Dvo

teC

onst

ant

Can

didat

epre

fere

nce

SP

Dpar

tyID

Tim

eT

ime�

cand

idat

eT

ime�

PID

PR

2�

10

0N

19

80

�2

.39���

(0.1

2)

1.4

5���

(0.1

2)

5.7

0���

(0.2

9)

�0

.37

(0.2

2)

0.3

7(0

.22

)�

1.5

2��

(0.4

4)

55

.81

37

05

19

83

�1

.53���

(0.1

0)

1.1

3���

(0.0

9)

4.8

6���

(0.3

6)

�0

.22

(0.1

6)

0.5

4���

(0.1

5)

0.3

6(0

.79

)5

7.5

33

81

19

87

�1

.54���

(0.0

7)

1.2

8���

(0.0

7)

4.1

3���

(0.2

8)

�0

.45��

(0.1

3)

0.2

4(0

.14

)0

.86

(0.5

6)

49

.71

04

11

19

90

�1

.88���

(0.0

8)

1.0

3���

(0.0

7)

4.8

2���

(0.2

7)

0.1

7(0

.13

)0

.04

(0.1

2)

�0

.77

(0.4

6)

47

.98

61

9

19

94

�2

.03���

(0.0

7)

0.7

9���

(0.0

8)

3.2

7���

(0.1

9)

0.4

2���

(0.1

2)

0.2

7�(0

.12

)0

.36

(0.3

2)

36

.21

20

06

19

98

�1

.54���

(0.0

8)

1.3

0���

(0.0

9)

3.3

1���

(0.2

6)

0.1

0(0

.14

)�

0.1

5(0

.15

)�

0.2

4(0

.42

)3

5.9

66

91

20

02

�1

.94���

(0.0

8)

1.2

0���

(0.0

8)

3.3

0���

(0.1

8)

�0

.29�

(0.1

2)

0.4

1��

(0.1

3)

�0

.50

(0.2

6)

41

.31

68

37

CD

U/C

SU

vote

Con

stan

tC

andi

dat

epre

fere

nce

CD

U/C

SU

par

tyID

Tim

eT

ime�

cand

idat

eT

ime�

PID

PR

2�

10

0N

19

80

�1

.21���

(0.0

9)

1.3

6���

(0.1

0)

6.3

6���

(0.4

5)

�0

.24

(0.1

6)

0.0

6(0

.17

)�

1.1

8(0

.70

)6

1.7

13

70

5

19

83

�1

.77���

(0.1

0)

1.4

0���

(0.1

0)

4.9

4���

(0.4

8)

0.3

2�(0

.16

)0

.47��

(0.1

6)

�0

.40

(0.7

5)

62

.33

38

1

19

87

�1

.53���

(0.0

7)

1.5

6���

(0.0

7)

4.9

8���

(0.3

6)

�0

.03

(0.1

4)

0.0

5(0

.14

)�

0.4

0(0

.68

)5

8.2

10

41

1

19

90

�1

.75���

(0.0

8)

1.2

1���

(0.0

8)

4.8

2���

(0.2

8)

�0

.17

(0.1

4)

0.1

9(0

.13

)�

0.6

3(0

.44

)5

3.1

86

19

19

94

�2

.18���

(0.0

9)

0.9

4���

(0.0

8)

4.5

0���

(0.2

5)

0.3

1�(0

.15

)0

.47���

(0.1

3)

�1

.36��

(0.3

9)

45

.01

20

06

19

98

�1

.79���

(0.1

0)

1.3

0���

(0.0

9)

3.5

8���

(0.2

8)

0.1

4(0

.16

)0

.06

(0.1

5)

�0

.72

(0.4

5)

44

.76

69

1

20

02

�1

.31���

(0.0

6)

1.1

1���

(0.0

6)

3.3

3���

(0.2

1)

�0

.11

(0.0

9)

0.3

1��

(0.0

9)

�0

.27

(0.3

0)

43

.91

68

37

Sig

nifi

canc

ele

vels

:� p<

0.0

5;��

p<

0.0

1;��� p<

0.0

01

.

identification. To test this relationship, I ran logistic re-gressions with vote choice as the dependent variable.As explanatory variables I included candidate preference,party identification, time, a candidate preference-by-timeand a party identification-by-time interaction term.

The evidence reported in Table 11 shows that cam-paigning did not increase the impact of party loyaltieson vote choice in the German federal elections since1980. Apart from that, during the 1980 and 1994 cam-paigns the effect of party attachments on vote choice de-clined considerably. As the predicted probabilities inTable 12 show, during the 1980 campaign, the effectof being a weak SPD identifier on vote choice declinedimmensely. As a result, the Social Democrats’ voteshare dropped by about 4 points. Arguably, this lossmay partially explain why the SPD did worse in the1980 election than many observers had expected beforethat election. During the 1994 federal campaign, re-gardless of the strength of identification, the effect ofCDU/CSU identification on vote choice decreasedclearly, resulting in a 2.5 points loss of CDU/CSU.

Turning to candidate priming, the evidence is con-siderably more encouraging. While the 1980, 1987,1990, and 1998 campaigns did not affect how stronglycandidate preference influenced vote choice, the re-maining three campaigns did. As the predicted proba-bilities reported in Table 12 indicate, candidatepriming varied somewhat among groups of voters. Ob-viously, the effect of candidate preference on vote inten-tion for a certain party increased considerably amongcitizens who did not identify with the respective party.In 1983, for example, at the start of the campaign, inthis group adherents of the SPD candidate were by 32points more likely to vote for the SPD than adherentsof the CDU/CSU candidate while at the end of the cam-paign the difference was 42 points (Table 12). As re-gards weak identifiers, things are quite different. Atthe end of the 1983 campaign adherents of both candi-dates differed less in their likelihood of voting for bothSPD and CDU/CSU than at the outset of the campaign.During the 1994 campaign, the difference did notchange at all, while during the 2002 campaign itincreased somewhat. Among strong party identifierscandidate effects were quite stable or declined some-what during campaigns.

Candidate priming changed the parties’ vote sharesand in some cases even altered the partisan balance.In 1983, candidate priming increased the vote sharesof both SPD and CDU/CSU by roughly 1 point, leavingthe partisan balance unchanged. In 1994, the vote shareof CDU/CSU was considerably stronger due to candi-date priming than the SPD’s share. However, this bonus

Page 12: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

335H. Schoen / Electoral Studies 26 (2007) 324e337

Table 12

Predicted probabilities of voting for SPD and CDU/CSU depending on party identification, candidate preference, and time of interview

SPD-vote 1980 CDU/CSU-vote 1994

Start of campaign End of campaign Start of campaign End of campaign

No, other party ID 13 3 10 3

Weak identification with

party voted for

72 36 53 22

Strong identification with

party voted for

98 91 91 73

Effect on aggregate outcome �4.1 �2.5

SPD-vote 1983 CDU/CSU-vote 1983

No, other

party ID

Weak

SPD ID

Strong

SPD ID

No, other

party ID

Weak

CDU/CSU ID

Strong

CDU/CSU ID

Start End Start End Start End Start End Start End Start End

Preference for.Candidate of other party 6 10 43 56 89 94 5 7 36 47 87 91

No candidate 17 26 70 80 96 98 16 23 69 78 96 98

Candidate of party voted for 38 52 88 92 99 99 43 55 90 94 99 99

Effect on aggregate outcome þ1.4 þ1.2

SPD-vote 1994 CDU/CSU-vote 1994

No, other

party ID

Weak

SPD ID

Strong

SPD ID

No, other

party ID

Weak

CDU/CSU ID

Strong

CDU/CSU ID

Start End Start End Start End Start End Start End Start End

Preference for.Candidate of other party 7 9 28 34 67 73 4 7 30 40 80 87

No candidate 15 18 46 53 82 85 10 16 52 64 91 94

Candidate of party voted for 27 33 66 72 91 93 23 32 74 82 96 98

Effect on aggregate outcome þ0.8 þ2.8

SPD-vote 2002 CDU/CSU-vote 2002

No, other

party ID

Weak

SPD ID

Strong

SPD ID

No, other

party ID

Weak

CDU/CSU ID

Strong

CDU/CSU ID

Start End Start End Start End Start End Start End Start End

Preference for.Candidate of other party 3 5 15 21 49 59 8 10 30 36 69 75

No candidate 10 15 37 47 76 83 20 25 56 64 87 90

Candidate of party voted for 28 37 66 75 91 94 43 50 80 84 95 97

Effect on aggregate outcome þ3.8 þ0.6

Predicted probabilities were calculated only if coefficients for interaction terms in Table 11 were statistically significant.

The cell entries 98 and 91 for SPD vote in 1980 in the row, strong identification with party voted for mean that at the outset of the 1980 campaign 98

percent of SPD identifiers intended to vote for SPD, while at the end of that campaign 91 percent did so. When calculating the voting probabilities all

variables not mentioned in the respective tables were set to their mode and mean, respectively. To examine effects on aggregate outcomes I calcu-

lated predicted probabilities using the respective regression equation in Table 11, with the coefficient for the explanatory variable of interest first set

to its initial value and then set to its value at the end of the campaign.

of about 2 points was neutralised by the loss that re-sulted from the decline of the party effect on votechoice. In the 2002 campaign, in contrast, candidatepriming upset the partisan balance considerably.Schroder’s popularity advantage over Stoiber gave theSPD about 3 points more from candidate priming thanthe Christian Democrats. As a result, if the candidate

effect had not changed during the campaign, the SocialDemocrats would have lost the election and failed toform the government (see also Schoen, 2004).

While the evidence shows that some campaigns ren-dered candidate evaluation more powerful predictors ofvote choice, it does not support the notion of candidatepriming having steadily increased in the period from

Page 13: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

336 H. Schoen / Electoral Studies 26 (2007) 324e337

1980 to 2002. This finding lends support to the idea thatlong-term trends like partisan dealignment and the riseof television are not the most important factors whichcondition candidate priming. The results for the 1983and 1994 campaigns suggest that candidate primingwas particularly likely to occur when at least one candi-date was not well-known to the public. However, candi-date priming was also strong during the 2002 campaignwhen two very prominent politicians competed for of-fice. The latter finding rather suggests that campaignstrategies aimed at making the election a choice be-tween the candidates may be of crucial importance.As a result, the analysis thus recommends that furtherresearch is needed to better understand the factors af-fecting candidate priming.

6. Conclusions

This paper examined the effect of election cam-paigns on voters’ attitudes toward candidates in Ger-many. Building on previous research, I proposed threehypotheses and tested them using pooled survey datafrom the 1980 to 2002 campaigns. The analysis showedthat during campaigns a candidate was increasingly per-ceived as the opponent of his rival and his rival’s party.Moreover, in the course of a campaign voters increas-ingly brought attitudes toward candidates into linewith attitudes toward parties. Though significant, thiseffect turned out to be rather small; if political new-comers without prior party commitments ran for office,it is likely a stronger would be found. Notwithstandingthis qualification, the findings lend support to the con-clusion that election campaigns render voters’ attitudestoward candidates and parties more consistent. Thus, itis difficult to draw valid conclusions about attitude con-sistency in off-election times from findings about atti-tude structure immediately before or after an election.

Moreover, this analysis showed that in Germany, as inother countries (e.g. Mendelsohn, 1996; Gidengil et al.,2002; Druckman, 2004), campaigns play a considerablerole in influencing how strongly political attitudes affectvote choice. While in no federal election was the effectof party attachments found to increase, in three of sevencampaigns candidate evaluations became considerablymore powerful predictors of vote choice. In 2002, candi-date priming even changed the election outcome. As thisanalysis indicates, if campaigning can make a differenceat the polls it also may encourage campaigners to con-sider carefully which strategies to utilize. A party witha popular candidate, for example, has a strong incentiveto turn an election into a contest between candidates. Aswas the case with polarization and party politicization

effects, priming effects varied considerably across elec-tions. Models of voting behaviour that assume campaigneffects to be constant across societal and political condi-tions thus cannot add much to our understanding ofvoters’ decision-making.

While the analysis found variation in campaign ef-fects it was suggestive rather than conclusive in deter-mining the factors that caused changes in campaigneffects. Future research may overcome this shortcomingby including more than seven elections and by substitut-ing quantitative campaign information for the rather an-ecdotal evidence used in this analysis. This design mayalso be apt for examining campaign effects in differentpolitical systems. Finally, future research must addressthe question whether campaigns affect the criteriavoters use to evaluate candidates.

References

Arzheimer, K., 2006. ‘Dead men walking?’ party identification in

Germany, 1977e2002. Electoral Studies 25, 791e807.

Bean, C., Mughan, A., 1989. Leadership effects in parliamentary

elections in Australia and Britain. American Political Science Re-

view 83, 1165e1179.

Berelson, B., Lazarsfeld, P.F., McPhee, W.N., 1954. Voting. A Study

of Opinion Formation in a Presidential Campaign. University of

Chicago Press, Chicago.

Bowler, S., Broughton, D., Donovan, T., Snipp, J., 1992. The in-

formed electorate? Voter responsiveness to campaigns in Britain

and Germany. In: Shaun, Bowler, David, M. Farrell (Eds.), Elec-

toral Strategies and Political Marketing. Macmillan, Basingstoke,

pp. 82e104.

Brettschneider, F., 2001. Candidate voting. Die Bedeutung von

Spitzenkandidaten fur das Wahlerverhalten in Deutschland,

Großbritannien und den USA von 1960 bis 1998. In: Hans-

Dieter, Klingemann, Max, Kaase (Eds.), Wahlen und Wahler.

Analysen aus Anlass der Bundestagswahl 1998. Westdeutscher

Verlag, Wiesbaden, pp. 351e400.

Brettschneider, F., 2002. Spitzenkandidaten und Wahlerfolg. Person-

alizierung e Kompetenz e Parteien. Ein internationaler Vergleich.

Westdeutscher Verlag, Wiesbaden.

Campbell, A., Converse, P.E., Miller, W.E., Stokes, D.E., 1960. The

American Voter. Wiley, New York.

Druckman, J.N., 2004. Priming the vote. Campaign effects in a US

senate election. Political Psychology 25, 577e594.

Druckman, J.N., Jacobs, L.R., Ostermeier, E., 2004. Candidate strategies

to prime issues and images. The Journal of Politics 66, 1180e1202.

Falter, J.W., Schoen, H., Caballero, C., 2000. Dreißig Jahre danach:

Zur Validierung des Konzepts ,Parteiidentifikation’ in der Bun-

desrepublik. In: Markus, Klein, Wolfgang, Jagodzinski,

Ekkehard, Mochmann, Dieter, Ohr (Eds.), 50 Jahre Empirische

Wahlforschung in Deutschland. Entwicklung, Befunde, Perspek-

tiven, Daten. Westdeutscher Verlag, Wiesbaden, pp. 235e271.

Farrell, D.M., 2001. Electoral Systems. A Comparative Introduction.

Palgrave, Houndmills.

Finkel, S.E., 1989. The effects of the 1980 and 1984 campaigns on

mass ideological orientations: testing the salience hypothesis.

The Western Political Quarterly 42, 325e346.

Page 14: Campaigns, candidate evaluations, and vote choice: Evidence from German federal election campaigns, 1980–2002

337H. Schoen / Electoral Studies 26 (2007) 324e337

Finkel, S.E., Schrott, P.R., 1995. Campaign effects on voter choice in

the German election of 1990. British Journal of Political Science

25, 349e377.

Fishbein, M., Ajzen, I., 1981. Attitudes and Voting Behaviour: an appli-

cation of the theory of reasoned action. In: Stephenson, Geoffrey

M., Davis, James H. (Eds.), Progress in Applied Social Psychology.

John Wiley & Sons, Chichester, pp. 253e313.

Fishbein, M., Coombs, F.S., 1974. Basis for decision: an attitudinal

analysis of voting behaviour. Journal of Applied Social Psychol-

ogy 4, 95e124.

Funk, C., 1999. Bringing the candidate into models of candidate eval-

uation. The Journal of Politics 61, 700e720.

Gelman, A., King, G., 1993. Why are American election polls so var-

iable when votes are so predictable? British Journal of Political

Science 23, 409e451.

Gidengil, E., Blais, A., Nevitte, N., Nadeau, R., 2002. Priming and

campaign context. Evidence from recent Canadian elections. In:

David, M. Farrell, Rudiger, Schmitt-Beck (Eds.), Do Political

Campaigns Matter? Routledge, London, pp. 76e91.

Granberg, D., Holmberg, S., 1988. The Political System Matters.

Cambridge University Press, Cambridge.

Green, D., Palmquist, B., Schickler, E., 2002. Partisan Hearts and

Minds. Political Parties and the Social Identities of Voters. Yale

University Press, New Haven.

Hetterich, V., 2000. Von Adenauer zu Schroder e Der Kampf um

Stimmen. Eine Langsschnittanalyse der Wahlkampagnen von

CDU und SPD bei den Bundestagswahlen 1949 bis 1998. Leske

und Budrich, Opladen.

Holtz-Bacha, C., Kaid, L.L., 1995. A comparative perspective on po-

litical advertising. In: Lynda, Lee Kaid, Christina, Holtz-Bacha

(Eds.), Political Advertising in Western Democracies. Parties

and Candidates on Television. Sage, Thousand Oaks, USA,

pp. 8e18.

Iyengar, S., Kinder, D.R., 1987. News that Matters. University of

Chicago Press, Chicago/London.

Jaccard, J., 2001. Interaction Effects in Logistic Regression. Sage,

Thousand Oaks, USA.

Jacobs, L.R., Shapiro, R.Y., 1994. Issues, candidate image and prim-

ing. American Political Science Review 88, 527e540.

Johnston, R., Blais, A., Brady, H.E., Crete, J., 1992. In: Letting the

People Decide: Dynamics of a Canadian Election. McGill

Queen’s University Press, Montreal.

Kinder, D., 1986. Presidential character revisited. In: Richard, Lau,

David, Sears (Eds.), Political Cognition. Lawrence Erlbaum As-

sociates, Hillsdale, NJ, pp. 233e256.

Krosnick, J.A., Kinder, D.R., 1990. Altering the foundations of sup-

port for the president through priming. American Political Sci-

ence Review 84, 497e512.

Lau, R.R., 1989. Construct accessibility and electoral choice. Politi-

cal Behaviour 11, 5e32.

Lazarsfeld, P.F., Berelson, B.R., Gaudet, H., 1944. The People’s

Choice. How the Voter Makes up his Mind in a Presidential Cam-

paign. Duell, Sloan and Pearce, New York.

Markus, G.B., 1982. Political attitudes during an election year: a re-

port on the 1980 NES panel study. American Political Science

Review 76, 538e560.

Mendelsohn, M., 1996. The media and interpersonal communica-

tions: the priming of issues, leaders, and party identification.

The Journal of Politics 58, 112e125.

Miller, W.E., Shanks, J.M., 1982. Policy directions and presidential

leadership: alternative interpretations of the 1980 presidential

election. British Journal of Political Science 12, 299e356.

Miller, W.E., Shanks, J.M., 1996. The New American Voter. Harvard

University Press, Cambridge, London.

Miller, A., Wattenberg, M.P., 1985. Throwing the rascals out: policy

and performance evaluations of presidential candidates, 1952e1980. American Political Science Review 79, 359e372.

Norris, P., 1997. The rise of postmodern political communication. In:

Pippa, Norris (Ed.), Politics and Press. The News Media and their

Influences. Rienner, Boulder, pp. 1e17.

Pancer, S.M., Brown, S.D., Barr, C.W., 1999. Forming impressions of

political leaders: a cross-national comparison. Political Psychol-

ogy 20, 345e368.

Peterson, D.A.M., 2005. Heterogeneity and certainty in candidate

evaluations. Political Behaviour 27, 1e24.

Popescu, M., Toka, G., 2002. Campaign effects and media monopoly.

In: David, M. Farrell, Rudiger, Schmitt-Beck (Eds.), Do Political

Campaigns Matter? Routledge, London, pp. 58e75.

Schoen, H., 2004. Winning by priming? Campaign strategies, chang-

ing determinants of voting intention, and the outcome of the 2002

German federal election. German Politics and Society 22, 64e82.

Schoen, H., Weins, C., 2005. Der sozialpsychologische Ansatz zur

Erklarung vonWahlverhalten. In: Jurgen, W. Falter,

Harald, Schoen (Eds.), Handbuch Wahlforschung. VS Verlag

fur Sozialwissenschaften, Wiesbaden, pp. 187e242.

Sears, D.O., Chaffee, S.H., 1979. Uses and effects of the 1976 de-

bates: an overview of empirical studies. In: Sidney, Kraus

(Ed.), The Great Debates: Carter vs. Ford. Indiana University

Press, Bloomington, pp. 223e261.

Shaw, D.R., 1999. The impact of news media favorability and candi-

date events in presidential campaigns. Political Communication

16, 183e202.

Swanson, David L., Mancini, Paolo (Eds.), 1996. Politics, Media, and

Modern Democracy. An International Study of Innovations in Elec-

toral Campaigning and their Consequences. Praeger, Westport.

Valentino, N.A., Hutchings, V.L., White, I.K., 2002. Cues that mat-

ter: how political ads prime racial attitudes during campaigns.

American Political Science Review 96, 75e90.

Vetter, A., Gabriel, O.W., 1998. Candidate evaluations and party

choice in Germany, 1972e94: do candidates matter? In:

Christopher, J. Anderson, Carsten, Zelle (Eds.), Stability and

Change in German Elections. How Electorates Merge, Converge,

or Collide. Praeger, Westport, pp. 71e98.

Wattenberg, M.P., 1991. The Rise of Candidate-Centered Politics:

Presidential Elections of the 1980’s. Harvard University Press,

Cambridge.

Weßels, B., 2000. Kanzler-oder Politikwechsel? Bestimmungsgrunde

des Wahlerfolgs der SPD bei der Bundestagswahl 1998. In: Jan

van, Deth, Hans, Rattinger, Edeltraud, Roller (Eds.), Die Repub-

lik auf dem Weg zur Normalitat? Leske und Budrich, Opladen,

pp. 35e66.

Zaller, J.R., 1992. The Nature and Origins of Mass Opinion. Cam-

bridge University Press, Cambridge.

Zelle, C., 1998. A third face of dealignment? An update on party

identification in Germany, 1971e1994. In: Christopher, J. Ander-

son, Carsten, Zelle (Eds.), Stability and Change in German Elec-

tions. How Electorates Merge, Converge, or Collide. Praeger,

Westport, CT, pp. 55e70.