Candidate Images in the 2005 German National Election

Candidate Images in the 2005 German National Election FRANK BRETTSCHNEIDER, KATJA NELLER and CHRISTOPHER J. ANDERSON Based on data collected for all national elections between 1961 and 2005, we examine the role of candidate images in shaping voting behaviour in German elections. Our analysis shows that Gerhard Schröder started the campaign on the defensive, but managed to run an impressive race of catch up against Angela Merkel, his main competitor. In large measure, his success in winning over voters was due to his ability to mobilise traditional SPD supporters during the course of the campaign with the help of the issue of social justice, which became a major subject of discussion. However, we also find that the 2005 elections were not primarily about the candidates for the chancellorship. Our analysis reveals that evaluations of the candidates for chancellor played only a small role in shaping voter choices. Contrary to some claims in the literature we find that the importance of candidate orientations in determining German voter choice has not changed dramatically in the last 45 years, and that such candidate effects are most powerful in shaping the choices of voters who do not identify with a political party. Thus, the 2005 election reflected the rule rather than an exception with regard to the importance of candidates in explaining voter choice in German elections. Like other parliamentary democracies, Germany does not elect its head of government directly. Yet, the Federal chancellor has long played such a pre-eminent role in German politics that commentators have frequently been tempted to describe the German political system as a chancellor democracy. 1 It therefore would not be unreasonable to assume that German elections are to a considerable extent about electing a chancellor. What is more, political conditions prior to the 2005 parliamentary (Bundestag) election only served to buttress this notion. For one, in May 2005 Gerhard Schröder, the incumbent chancellor, asked for the dissolution of the Bundestag and a new election after the Social Democrats had lost an important Land election in North Rhine- Westphalia and following a precipitous decline in support for Schröder s reform policies. By doing so, Schröder clearly sought to turn the election into a vote about his leadership and his proposals for reform. On the other side of the political spectrum, the country s major opposition party (CDU/CSU) for the first time in German history nominated a woman (Angela Merkel) as candidate for chancellor, producing much speculation about the possible effects of her candidacy on voter behaviour as well as the question of whether her East German origin would affect voters choices. For much of the pre-election period, it looked as if Schröder s gamble was going to fail, as polls showed the Christian Democrats (CDU/CSU) and the Liberal Party (FDP) German Politics, Vol.15, No.4, December 2006, pp.481 499 ISSN 0964-4008 print=1743-8993 online DOI: 10.1080=09644000601062667 # 2006 Association for the Study of German Politics

482 GERMAN POLITICS leading the incumbent coalition of Social Democrats (SPD) and Greens by a significant margin. In May, the CDU/CSU was ahead of the SPD by about 25 percentage points (see the contribution of Wüst and Roth in this volume), and Angela Merkel led Gerhard Schröder when it came to the question of which candidate German voters preferred as chancellor (see Figure 1). This stood in notable contrast to previous elections in which Schröder had been a candidate for chancellor. For example, during the months prior to the 1998 and 2002 elections, Schröder had been consistently ahead of the CDU/CSU candidates (Helmut Kohl and Edmund Stoiber respectively). These unusual conditions only changed in the second half of July. At that point, Schröder was able to gain the lead over Merkel and even expand it until the day of the election. Shortly before the election, 53 per cent of eligible voters said they would prefer Gerhard Schröder as chancellor, while only 40 per cent preferred Angela Merkel. Yet, despite Schröder s personal popularity among voters, his Social Democratic Party failed to win the election; the SPD lost its leading role in government and the Red-Green coalition was ousted. With that, Gerhard Schröder s time as chancellor of the German Federal Republic came to an end. While the SPD defeat did not come as a surprise to most, given the political dynamics of the pre-election period, the failure of the Christian Democrats to win the election outright was noteworthy, especially because the 2005 federal election had been preceded by a series of 11 FIGURE 1 PREFERRED CHANCELLOR BEFORE GERMAN NATIONAL ELECTIONS 1998, 2002 AND 2005.

CANDIDATE IMAGES IN THE 2005 ELECTION 483 victories in Land elections. In fact, instead of celebrating what was assumed to be certain victory, the Christian Democrats scored their second-worst election result in history, with 35.2 per cent of the vote. As importantly, instead of the CDU/CSU taking over government with the help of coalition with the centrist FDP, which managed to get 9.8 per cent of the vote, the election resulted in a lengthy political impasse and eventually led to the formation of a Grand Coalition of CDU/CSU and SPD, with Angela Merkel as chancellor. Below, we examine the question of whether and how much the chancellor candidates affected the behaviour of voters in the 2005 election, and, in comparison, how candidates for chancellor have affected voters in elections since 1961. Does the unexpectedly close result mean that Chancellor Gerhard Schröder boosted the otherwise hapless SPD? And did the nomination of Angela Merkel cost the Christian Democrats the election victory they believed was certain? Moreover, was the 2005 election typical, or was the influence of the candidates for the chancellorship particularly powerful in that contest? PERSONALISATION AND ELECTORAL POLITICS IN GERMANY: SEPARATING FACT FROM FICTION Scholarly and popular analysts have long portrayed German elections as having become Americanised meaning that politics has become increasingly lacking in substantive content, with important policy issues and debates getting lost in mere beauty contests among candidates and personalities that are designed by handlers, spin-doctors, and professional image-makers and then packaged in ways that are easily communicated via mass media. 2 As normative democratic theorists would argue, such an increasing de-politicisation, if it in fact exists, can be a danger for democracy. But while these portrayals are popular, are they correct? As we argue below, these claims about the increased focus on candidates and personalities in German elections are off the mark. While political marketing has certainly become much more sophisticated, parties messages are usually dominated by programmatic appeals and descriptions during the early months of election campaigns in order to remind voters generally, but especially core supporters, of the parties and candidates positions. Only subsequently, and usually only toward the end of the campaign, does the chancellor campaign with a focus on the question of who will be chancellor dominate the parties strategies. Taking a long historic view, it is easy to see that such election year dynamics are hardly novel, and that they are difficult to construe as evidence for an increasing depoliticisation of politics. In fact, a strategic focus on the candidates for chancellor has a long track record in German election campaigns and is part of the toolkit of virtually every national campaign in post-war German history. For example, the CDU s campaign handbook from 1965 (not the 1990s or 2005!) reminded campaign staffers that: All advertising statements should be...as strongly personalised as possible... A personalisation of the election campaign and its advertising for the CDU means ¼ a complete focus on Ludwig Erhard. 3 What is more, over the years, the CDU was certainly not alone in its effort to personalise the election. The SPD s 1972 election campaign, for example, was completely focused on the then Chancellor

484 GERMAN POLITICS Willy Brandt so much so that the election became popularly known as the Willy election [ Willy-Wahl ]. Aside from the fact that parties have long emphasised the personalities and positive traits of their chancellor candidates, this does not necessarily imply that there has been a trend toward depoliticisation in German elections. Instead, as a matter of course, campaigns over the years worked hard to connect candidates and issues in order to make them more memorable to voters. Candidates are the faces and voices of parties programmes, and parties consequently have long promoted their candidates in order to communicate substantive political messages. It s about the Chancellor [ Auf den Kanzler kommt es an ] was the slogan the Christian Democrats used in the election of 1969 when Kurt Georg Kiesinger was the candidate. At the same time, chancellor candidates can make up for parties programmatic deficits (as Gerhard Schröder managed to do in the 1998 election) or reinforce them (Rudolf Scharping in 1994). While personalisation and a focus on the chancellor candidates is thus nothing new in German elections, it is critical to distinguish between the presumed personalisation of election campaigns and the presumed personalisation of voting behaviour. To demonstrate the latter, researchers have sought to document two trends: first, an increase in the importance of candidate orientations for explaining voters decisions, along with a parallel decrease in the importance of issue orientations and attitudes toward parties (partisan attachments); and second, that voters images of candidates have become increasingly shaped by the non-political characteristics of candidates, while candidates traits such as problem solving ability or competence have decreased in importance. Unfortunately, however, the vehemence with which these claims are made stands in little relation to the actual evidence to support them. 4 We provide evidence to assess the veracity of these alleged trends with old and new data and examine changes in German voting behaviour over time with the help of a series of election studies conducted since 1961. 5 Specifically, we ask how important evaluations of the chancellor candidates are for individual voter behaviour in federal elections. How has the importance of candidate orientations developed since the 1960s, and was the federal election of 2005 an exceptional election in this regard? In addition, we analyse the underlying determinants of candidate images in the 2005 elections, and examine whether the images of Schröder and Merkel changed during the course of the 2005 election campaign. CANDIDATE ORIENTATIONS AND THE BEHAVIOUR OF INDIVIDUAL VOTERS In the well-known social psychological model of voter behaviour (also known as the Michigan model), individual vote choice is based on a combination of long-term (stable) factors, as well as short-term elements. 6 The long-term element of this model is a kind of psychological party membership, usually referred to as party identification or party attachment, which constitutes the core of the model and is highly predictive of voting behaviour. 7 Yet, the vote is not an automatic reflection of partisan attachment. Each election brings with it different contextual conditions and political dynamics, and these affect the vote as well. Candidates and issues (as well as parties) change over time, and this is reflected in the vote. Thus, aside from

CANDIDATE IMAGES IN THE 2005 ELECTION 485 party identification, attitudes toward particular issues, as well as the candidates, can exercise a short-term influence on the vote. To be sure: these are usually closely related to party identification for example, a voter who likes the SPD is likely to prefer the Social Democratic candidate and support the party s issue orientations but research has shown that they are not coterminous. Moreover, attitudes toward issues as well as the candidates, especially if they have been in office for some time, can strengthen or weaken voters existing partisan attachments. 8 Thus, for voters who strongly identify with a party, party attachment, issue-, and candidate orientations tend to move in tandem. Because such voters usually vote for their preferred party even when they are not in complete agreement with its or its candidate s current positions, the independent effect of candidate orientation on the vote is expected to be relatively small for this group of committed partisans. In contrast, however, among voters who are only weakly attached to a party, inconsistencies among party identification and attitudes toward candidates and issues can be expected to be more common. This also means that the potential for candidate orientations to play a significant role in the calculus of voters is greatest among these unattached voters. 9 Unsurprisingly, perhaps, the personalisation strategies employed by the managers of election campaigns are primarily focused on this group of unattached voters. Because of massive social changes in post-industrial societies like Germany, such as an increase in social mobility and the weakening of traditional milieus, the proportion of voters who do not identify with one of the big parties increased until the early 1990s and has been stable ever since. 10 Prior to the 2005 election, for example, a third of all respondents indicated that they did not lean toward one of the large parties (CDU/CSU and SPD). At the same time, the proportion of voters who expressed a strong affective attachment to the big parties decreased between the 1960s and 1990s. Moreover, while the programmatic differences between the big parties have not disappeared completely, voters perceptions of programmatic differences are smaller today than in the 1960s or the beginning of the 1970s. 11 EXPLAINING THE VOTE All this gives us good reason to pay attention to so-called candidate voting. Yet, as fascinating and plausible the assumption of increasing candidate centred voting behaviour may be, the empirical evidence in support of it is hard to come by in the German context. And this is not for lack of trying. But in contrast to the assumptions underlying the personalisation argument, the few existing studies that seek to examine the role of candidate images in shaping voter choices do not show an increasingly important role for candidate orientations for how people vote. Instead, the literature shows that candidate effects are quite variable: they vary by election, and appear to depend on the particular personalities and context of any one electoral contest. 12 In part, this may be a methodological and empirical problem. The extent to which candidate orientations influence voting behaviour is difficult to determine precisely because of the strong relationships among party identification, issue orientations, and attitudes toward the candidates. As a result, it is complicated at best to isolate unambiguously the effect of any one of these three factors in the calculus of voters. An example may help: since there is a strong relationship between party identification

486 GERMAN POLITICS FIGURE 2 THE EVALUATION OF SPD CHANCELLOR CANDIDATES BY VOTERS WITH DIFFERENT KINDS AND LEVELS OF PARTISAN ATTACHMENT IN WEST GERMANY, 1961 2005 (AVERAGE VALUES ON A SCALE FROM þ5 TO 5). and the evaluation of a candidate, as shown in Figures 2 and 3, chancellor candidates are evaluated most positively by those voters who are also attached to that candidate s party. And, as expected, candidates are least liked by supporters of the opposing party, with those who do not report any partisan identification located somewhere between the two camps. 13 In 2005, for instance, SPD partisans on average gave Schröder a score of þ3.2 on a scale running from 5 to þ5, while CDU/CSU partisans gave him an average score of only 0.7, with other partisans and independents evaluating him between these two groups at þ0.9. In contrast, CDU/CSU partisans, on average, gave Merkel a þ3.2, whereas SPD partisans evaluated her much more negatively, at 0.6. The pattern is straightforward: SPD partisans evaluated Schröder in the same way as CDU/CSU partisans evaluated Merkel. Moreover, Christian Democrats expressed the same negative attitudes toward Schröder that Social Democrats did toward Merkel. Aside from these stable partisan-driven differences in the evaluation of candidates, there are variations that are more candidate-specific and that can be found for all groups of voters. For example, all voters evaluated Rainer Barzel, the CDU/CSU s candidate for chancellor in 1972, more negatively than the party s candidate Kurt Georg Kiesinger three years earlier or Helmut Kohl four years later. And in both 1976 and 1980, Helmut Schmidt was popular among voters regardless of partisan identification. Moreover, generally speaking, it is worth noting that SPD candidates over the years

CANDIDATE IMAGES IN THE 2005 ELECTION 487 FIGURE 3 THE EVALUATION OF CDU/CSU CHANCELLOR CANDIDATES BY VOTERS WITH DIFFERENT KINDS AND LEVELS OF PARTISAN ATTACHMENT IN WEST GERMANY, 1961 2005 (AVERAGE VALUES ON A SCALE FROM þ5 TO 5). have been evaluated quite positively; the one segment of the electorate that has typically reported negative evaluations of SPD candidates has consisted of strong CDU/ CSU supporters. What is noticeable about the 2005 election is that support for Gerhard Schröder was even stronger among SPD voters than it had been in 1998, but weaker than in 2002. Whereas he achieved an average score of þ2.4 on the 5/þ5 scale in 1998, he improved to þ3.7 in 2002 and slightly declined to þ3.2 three years later. As importantly, Schröder s favourable ratings among voters without partisan attachments stood in contrast to the weak support these voters expressed for Angela Merkel. At the same time, it is worth noting that voters without partisan attachments were more supportive of Angela Merkel than they had been of Edmund Stoiber three years earlier. At the same time, CDU/CSU voters were even less happy with Schröder in 2005 (at 0.7) than they had been in 2002 (at 0.1) and in 1998 (at þ0.2). And Angela Merkel was more popular among Christian Democratic identifiers (at þ3.2) than Edmund Stoiber (at þ2.7) and Helmut Kohl (at þ2.8) had been in 2002 and 1998, respectively. In the end, differences in popularity were much smaller between Schröder and Merkel than between Schröder and Stoiber or Schröder and Kohl. This is also reflected in responses to the question of chancellor preference already mentioned above (see Figure 1). Viewed from this perspective, the initial conditions for a Christian Democratic victory were much better in 2005 than they had been in earlier contests.

488 GERMAN POLITICS As we mentioned above, the exact extent of so-called candidate voting is difficult to quantify precisely because of the strong relationships among candidate orientations, issue attitudes, and partisan attachments. With these difficulties in mind, we can estimate how much candidate orientations are related to voting behaviour with the help of a couple of analytic strategies. 14 As a first, and admittedly simple, step, we can try to explain vote choice exclusively with the help of candidate orientations. These results are shown as the top line in Figure 4. 15 They suggest that the explanatory power of candidate orientations for voting behaviour varies between 32 per cent (1969) and 59 per cent (1980). Thus, compared to other elections, the candidate effects in the 2005 election were about average compared with previous years, and similar to those in 2002. Using this approach for estimating candidate effects, it is difficult to speak of an increased personalisation of German elections in the sense of a heightened importance of candidate orientations for voting behaviour. However, this estimation procedure undoubtedly overstates the actual candidate effects significantly. Instead, we would interpret these estimates as an upper bound of candidate effects. To get a better handle on how much candidates really matter, we therefore estimated a model of vote choice in German elections that controlled for party identification and issue orientations. First, we estimated how well we can FIGURE 4 THE EXPLANATORY POWER OF CANDIDATE ORIENTATIONS FOR VOTER BEHAVIOUR, WITH AND WITHOUT CONTROLS FOR PARTY IDENTIFICATION AND ISSUE ORIENTATIONS IN WEST GERMANY, 1961 2005 (D ADJ. R 2 AND ADJ. R 2 100).

CANDIDATE IMAGES IN THE 2005 ELECTION 489 explain vote choice with the help of party identification and issue orientations. We then calculated by how much the model s explanatory power can be increased by adding candidate orientations (DR 2, see lower line in Figure 4). Our estimations show that, after controlling for party identification, candidate orientations explain only between 1 and 9 per cent of the remaining variation in vote choice (results not shown). When issue orientations are included as well, the independent influence of candidate orientations, cleansed of the influence of partisan attachment and issue orientations, becomes negligible (see Figure 4). It hovers between 0 and 4 per cent, and thus is considerably weaker than in the U.S., for example, where it has been estimated to vary between 12 and 22 per cent. 16 Again, when we trace the effect of candidate orientations on vote choice since 1961, there is no evidence of an increased influence of candidate orientations over time and thus no evidence of increased candidate-centred voting in German elections. 17 In the 2005 Bundestag election, party identification was by far the most important determinant of voting behaviour in East and West Germany, followed by voters perceptions of parties issue competence. Candidate orientations took third place. These results speak to the importance of political parties in Germany compared to other political systems, especially the U.S., where parties tend to be less important and candidate effects comparably stronger. In the late 1980s, for example, 92 per cent of American voters agreed with the statement I always vote for the person who I think is best, regardless of what party they belong to. 18 When the Forschungsgruppe Wahlen asked a similar question prior to the 2005 election ( What is more important: Which parties form the government or who will be the Federal Chancellor? ), 72 per cent said parties (compared to 64 per cent in 2002) and only 19 per cent said Federal Chancellor (compared to 27 per cent in 2002). While the evidence so far thus does not support the contention that the 2005 election brought any more, or any less for that matter, personalisation into German elections, it is worth considering how changes in partisanship have affected the role candidate images play. Specifically, in the debate over whether there has been personalisation in German voting behaviour, a number of commentators quite rightly have pointed to the increased number of voters who lack partisan attachments. 19 For our purposes, this means that it would be particularly important to take a closer look at this group of voters and the role candidate orientations play in their vote choice. Their behaviour may well matter in the aggregate: if this group of voters is increasing in the population, and if candidate orientations are more important for them than for others, then it is likely that candidate orientations do play an increasingly important role for electoral outcomes in the electorate as a whole. Figure 5 shows the influence of candidate orientations on vote choice in the form of unstandardised regression coefficients, controlling for party identification and issue orientations, for voters with and without partisan attachments. As expected, candidate orientations exert a stronger influence on the vote among voters without partisan attachments than those who report being partisan identifiers. Moreover, among non-identifiers, fluctuations in how much candidates matter are of greater magnitude. Thus, it is reasonable to conclude that the power of candidates to shape the vote is stronger for unattached voters than those who identify with a party. Among the latter, party identification thus may act as a buffer for candidate effects.

490 GERMAN POLITICS FIGURE 5 THE EXPLANATORY POWER OF CANDIDATE ORIENTATIONS FOR VOTE CHOICE AMONG VOTERS WITH AND WITHOUT PARTY IDENTIFICATION, 1961 2005. Among the more interesting results, our estimations of candidate effects on vote choice among partisan identifiers and non-identifiers since the 1960s suggest that candidate orientations actually had a stronger influence on the vote among respondents who reported that they were attached to a political party in 1998. This shows that, at that time, Gerhard Schröder s positive and Helmut Kohl s negative evaluations had a particularly strong impact among voters with partisan attachments and may, in fact, have produced very strong differences in the vote among these groups. As the results for 2002 and for 2005 show, however, this was not the beginning of a trend. Instead, the 2002 and 2005 elections brought a return to well-documented patterns in German politics. Voters choices in these elections were neither particularly strongly nor particularly weakly influenced by people s attitudes toward the candidates. The incumbent chancellor was clearly more popular than his own party and his challengers. Yet, when voters were asked to make a choice, partisan attachments and issue orientations mattered to a much greater degree. Thus, in 2002 and 2005, the influence of candidate images on the vote remained stable among voters who identify with a party, while the size of candidate effects returned to the 1998 level among voters without partisan attachments. But because voters evaluations of the candidates were much closer in 2005 than they had been in 1998, the parties did not benefit significantly from these differences. CANDIDATE IMAGES IN GERMAN POLITICS Until now, we have focused on the overall evaluations of candidates by voters. What, then, explains whether voters viewed a candidate positively or negatively? And is there evidence for the second personalisation hypothesis, according to which non-political

CANDIDATE IMAGES IN THE 2005 ELECTION 491 characteristics of candidates should have been important for how voters viewed candidates in 2005? To summarise briefly, the scholarly literature on candidate images has found that a number of seemingly disparate candidate characteristics can be summarised along four trait dimensions: problem solving ability, leadership qualities, integrity, and non-political or personal characteristics. 20 Studies of the ideal chancellor provide some evidence as to what voters expect from their candidates. 21 First, perhaps unsurprisingly, the ideal chancellor should combine all the positive traits and exhibit none of the negative ones. Moreover, the ideal chancellor should be credible, energetic, and a strong leader. Third, issue competence is almost as important as leadership qualities and integrity. This does not mean that voters expect the candidates to be experts on arcane policy details. Instead, this aspect of candidate characteristics is directional: does the candidate stand for social justice? Can they be trusted to maintain internal and external peace? Do they support environmental protection? etc. While German voters seem to agree that problem solving competence is a critical trait for the ideal chancellor, there is partisan disagreement about the issues that should be at the centre of a candidate s competence. Fourth, German voters agree that personal characteristics, such as age, appearance, or charisma, are the least important traits when they imagine the ideal chancellor. While voters do pay attention to them they talk about the candidates appearance or families with their neighbours and friends these characteristics appear secondary when it comes to exercising the duties of Federal chancellor. And fifth, and importantly for the purposes of this analysis, these attitudes have not changed markedly since the 1970s. While there is a sizable literature on candidate evaluations in the U.S., and in particular on how these orientations have changed over the years, there is relatively little existing research in the context of German politics. What is more, because of differences in question wording and other kinds of variability in representative surveys, it is difficult to say with much certainty whether and how voters views of candidates have changed over time. 22 With that caveat in mind, the relatively few empirical studies on candidate images reveal results that stand in contrast to the personalisation thesis. 23 Specifically, they show that the overall evaluation of candidates by voters appears to be based less on non-political characteristics or the integrity dimension. Instead, leadership qualities and issue competence are at the centre of candidate evaluations. In the 1969 election, for example, voters evaluated Willy Brandt and Kurt Georg Kiesinger primarily with an eye toward their leadership qualities. At the time, about half of all responses to an open-ended questionnaire item regarding the good and bad qualities of the candidates mentioned traits such as decisiveness, vigour, and the ability to get things done. In contrast, issues and the candidates integrity played only a secondary role. But about one out of five mentions of candidate traits dealt with non-political characteristics. While voters in the case of Brandt emphasised qualities such as youthfulness, charm, and his life during National Socialism (as they did in 1961), they focused on Kiesinger s age and appearance. In fact, because of his elegant looks, he was frequently referred to as the TV-Beau. 24 The only other time that a similarly sizable proportion of references was made to candidates non-political characteristics was in 1976, 25 after which they became essentially meaningless. Thus, during the 1998 election, for example, only a very small number of respondents

492 GERMAN POLITICS mentioned Gerhard Schröder s private life and multiple marriages. Aside from the occasional tabloid story, this topic was completely irrelevant in the minds of voters. In contrast, Konrad Adenauer s push to make his challenger s (Brandt s) out of wedlock birth an issue in the 1961 campaign created much more excitement. 26 Moreover, Brandt was frequently compared to Jack Kennedy because of his youthful appearance and charm in order to contrast the youthful and dynamic challenger Brandt to the then 85-year-old Adenauer. Although we have little systematic evidence to make this case, non-political traits thus appear to have played a larger role in the 1960s. Given that they subsequently lost whatever influence they may have had leaves us unable to support the personalisation thesis. While there certainly is little evidence of a depoliticisation of candidate evaluations, it is also the case that, in contrast to the U.S., the candidates leadership qualities are somewhat more important than issue competence. In a parliamentary democracy with strong and disciplined political parties, the latter appears to be more strongly connected with the political parties than in a system such as the U.S., where candidates and issue attitudes are more closely connected. The idea that chancellor candidates are also evaluated on the basis of issues in the Federal Republic can be supported both anecdotally and empirically. For example, while the new policy of détente toward the East was connected with the SPD in the 1972 election, the policy was primarily Willy Brandt s creation. After all, German commentators usually refer to the policy of détente as the Brandt s Ostpolitik, not the policy of the SPD. The same is true for the notion of modernisation, which was connected with Brandt s famous expression of risking more democracy [ mehr Demokratie wagen ]. Analogously, in the 1990s, Helmut Kohl was closely linked with the issue (re-)unification of Germany and European integration. In other election campaigns during the 1970s and 1980s, candidates for chancellor were perceived predominantly in light of their leadership qualities. 27 In this context, a particularly important element of candidates images has to do with voters perceptions of the candidates ability to win support for positions within the party and then to fashion the necessary level of intra-party agreement. In addition, more general managerial abilities such as decisiveness and persistence are part of this dimension. These traits marked the image of Chancellor Helmut Schmidt, for example. In both the 1976 and the 1980 campaigns, he was presented as the doer [ Macher ] and perceived this way by voters. Especially in 1980, the SPD campaign characterised the Christian Democratic candidate Franz Josef Strauß as unpredictable and impulsive, while Helmut Schmidt was portrayed as the rational and careful statesman in pursuit of a reliable set of policies. 28 However, it is not just the presence of leadership qualities that can mark a candidate s image, but also the absence of such qualities. In the 1994 election, the Social Democratic challenger to incumbent Chancellor Helmut Kohl, Rudolf Scharping, seemed to have promise in the eyes of voters, and the election outcome appeared open. Yet, as Election Day approached, Scharping made several important mistakes, including tactical blunders with regard to the election of the Federal President. These severely damaged Scharping s chances and helped put Kohl in an even more positive light. 29 The integrity of candidates, which was already a topic of public debate in the 1980 election, became even more important in the 1990 contest. Even in an election that had Germany s unification as a central theme, the perceived integrity of Helmut Kohl and

CANDIDATE IMAGES IN THE 2005 ELECTION 493 his Social Democrats challenger Oskar Lafontaine played a major role. 30 Credibility and a sense of duty were also at the centre of voters perceptions of the candidates in 1994. 31 And in 1998 more than a quarter of all mentions of candidates traits dealt with integrity. In particular, voters mentioned doubts about Kohl s credibility. This historical narrative and the few existing studies of candidate images in Germany lead to the following conclusion: It appears that candidates non-political characteristics were at least as important in the 1960s as in the 1990s. While candidates were viewed in terms of their leadership qualities more frequently in the second half of the 1970s and into the 1980s, it appears that the candidates integrity has become more important in the 1990s. Yet, based on the available information, there is little sign of a personalisation of candidate perceptions in German politics. COMPOSITION AND DYNAMICS OF CANDIDATE IMAGES IN THE 2005 ELECTION When we look at the 2005 contest in the early part of the campaign (June), the picture of candidate images is mixed. For example, Angela Merkel outpolled Gerhard Schröder in credibility, and she was perceived as more energetic. But voters attributed greater leadership qualities to Gerhard Schröder than to Angela Merkel. With regard to problem solving competence, voters perceived different strengths and weaknesses: Angela Merkel was considered more competent in generating new jobs, and voters thought she had a better plan for Germany s future. In contrast, Gerhard Schröder was said to be much more interested in the issue of social justice, and voters perceived him as more capable of representing German interests abroad. Additionally, Schröder was ahead of Merkel on a presumably apolitical dimension: likeability (see Table 1). By election day, this pattern had changed considerably. Immediately prior to the election, Schröder either led or pulled even with Merkel on almost all candidate TABLE 1 VOTERS PERCEPTIONS OF CANDIDATE TRAITS PRIOR TO THE 2005 FEDERAL ELECTION (%) June July/ Aug. Sept. I Sept. II Stands for social justice Schröder 36 45 51 57 Merkel 32 27 28 23 Creating jobs Schröder 20 27 29 31 Merkel 43 38 45 39 Representing German interests internationally Schröder 57 59 64 69 Merkel 24 24 24 20 Solving future problems Schröder 16 20 22 30 Merkel 33 30 39 33 More energetic Schröder 29 41 35 42 Merkel 44 35 40 35 More of a leader Schröder 47 57 57 63 Merkel 36 28 32 25 More credible Schröder 26 31 32 35 Merkel 43 34 41 38 More likeable Schröder 50 56 53 55 Merkel 31 25 30 29 Question wording: If you now compare Gerhard Schröder and Angela Merkel. Who of these two is (better in)... ; Source: Infratest dimap.

494 GERMAN POLITICS traits. Thus, the chancellor s challenger lost support on almost all dimensions and was perceived as slightly more competent than the incumbent only with regard to the (admittedly important) issue of creating new jobs. In contrast, Schröder managed to improve in all areas. Especially when it came to leadership qualities and sympathy, he was clearly ahead of Merkel in September, and with respect to credibility, he was able to pull even with Merkel. And in the field of issue competence, he was consistently ahead of Merkel on the issue of social justice that dominated the last weeks of the election campaign. 32 The trajectory of the issue of social justice during the 2005 campaign deserves more detailed consideration. The issue initially rose to prominence in the field of tax policy. The Christian Democrats made tax policy one of their central campaign topics, and they connected their novel proposals in this area with a new face: Angela Merkel appointed Paul Kirchhof, a former judge at the German Constitutional Court, as a member of her team of competence (Kompetenzteam) a type of shadow cabinet. At first, Kirchhof and his career switch to policymaker started out on a very hopeful note for the CDU/ CSU, and he enjoyed high and favourable attention by the media. There were detailed and positive reports about him, the topic of tax policy, and the CDU/CSU. But this initially successful strategy eventually turned into a debacle for the Christian Democrats. It began with some inconsistencies and clumsiness in Kirchhof s public persona and appearance. For example, he was critical of his own party s plans to increase the Value Added Tax (VAT) and demanded a flat income tax that was inconsistent with the CDU/CSU party programme. To achieve a flat tax, he suggested massive cuts in existing tax benefits. He then made some suggestions in the field of retirement politics that gave the Social Democrats an opening. At this point at the latest, Gerhard Schröder saw his chance. During the SPD party convention in late August in Berlin, he massively attacked the professor from Heidelberg as being radically unsocial ( radikal unsozial ), wanting to turn German voters into guinea pigs for his policies: When I hear this professor from Heidelberg as he goes on about pensions, who thinks and you can read up on this that one could organise old age insurance the way we organise automobile insurance, this reveals an attitude about human beings that at least we (as Social Democrats) have to fight with all our might. Human beings are not things and have to be treated differently from things. 33 From this point on, Kirchhof was described as a symbol of radical reforms with incalculable effects and a symbol of a cold social climate. In TV news broadcasts of that week, there were fewer but almost exclusively negative reports about Kirchhof. After a televised debate between Merkel and Schröder that was used by the incumbent to launch further attacks on Kirchhof, press coverage of Kirchhof increased and the negative tenor persisted. Most critically for the CDU/CSU, the subject of tax policy changed from an economic issue into a social policy issue. This allowed Schröder and the SPD to avoid talking about Schröder s reform project (Agenda 2010); instead, the SPD s major theme became the claim that it was the party that could save the welfare state from massive unsocial radical reforms. Thus, the Christian Democrats had given their opponents the munitions to attack it since there is hardly a better way to mobilise the social democratic base and to push the undecided voters of the political centre toward the Social Democrats. Thus, as in earlier elections, discussions about issues were communicated with the help of personalities and candidates.

CANDIDATE IMAGES IN THE 2005 ELECTION 495 WHAT IS LIKEABILITY? The research literature in political science for the most part considers candidates likeability to be apolitical. Moreover, given the importance of likeability during the election campaign, it appears to be further evidence of a personalisation in German politics. To examine whether these claims are justified, voters were asked about their general feelings about Gerhard Schröder and Angela Merkel in early 2005. Subsequently, they were asked to provide reasons for their evaluation. The answers to this open question were categorised under one of the four dimensions of candidate traits mentioned above: problem solving ability, leadership qualities, integrity, and non-political or personal characteristics. As expected, most entries account for non-political characteristics. Some people characterise Gerhard Schröder as good-looking, self-confident, and charming, while others call him arrogant and slippery. Additionally, for some people his several marriages and divorces and the fact that he adopted a child affect his likeability. Some describe Angela Merkel as someone with a pinched expression, unattractive, and not very feminine, and as someone with a bad haircut and style of dress. Others see her as calm, serene, and composed. Furthermore, voters made it a point to mention that she hails from East Germany. However, this proves to be neither an advantage nor a disadvantage. Non-political characteristics thus play an important role for likeability ratings and they are incidentally more important in the West than in East Germany. Indeed they make up slightly less than a half of all survey entries. In addition, about a quarter of all mentions (and about a third in the East) concern the problem solving abilities of the candidates, while 10 per cent refer to leadership qualities, and 15 per cent of all mentions refer to integrity. This leads to an important conclusion: Likeability is by no means apolitical. Moreover, it cannot be interpreted to constitute a measure of a single non-political candidate trait. Roughly half of the likeability ratings are rooted in candidate characteristics that are closely connected with their role as a political actor. Taken together, this implies that, in the German context, likeability is a kind of personally coloured summary evaluation of the candidates. This conclusion is further buttressed by the fact that likeability is strongly correlated with the general evaluations of the candidates. DISCUSSION AND CONCLUSION While popular notions of personalisation and the Americanisation of German electoral politics persist, the 2005 election were not primarily about the candidates for chancellor. Speaking generally, the historical evidence we present suggests that German elections have not turned into mere beauty contests between persons and personalities. Our analysis of data from 1961 to 2005 and of data collected throughout the 2005 election year reveals that evaluations of the candidates for chancellor play only a small role in shaping the behaviour of voters in German federal elections. Moreover, we find that such candidate effects are most powerful in shaping the choices of voters who do not identify with a political party. As importantly, our results show very clearly that the importance of candidate orientations in determining who Germans vote for has

496 GERMAN POLITICS TABLE 2 CANDIDATES LIKEABILITY, 2005 (% OF ALL ENTRIES) Schröder-sympathy Merkel-sympathy Non-political and personal characteristics (good looking, self-confident, charming, arrogant, slippery, too many marriages/divorces, adopted a child...) Non-political and personal characteristics (woman, East German origin, pinched expression, unattractive, unfavourable hair style, bad dress style, unfeminine, calm, serene, objective...) Total: 44% West: 48% East: 33% Total: 43% West: 45% East: 38% Problem solving ability (competent/incompetent, capable/incapable, has no concept, Iraq conflict...) Problem solving ability (competent/incompetent, capable/incapable, has no concept, Iraq conflict...) Total: 28% West: 23% 0East: 39% Total:26% West: 24% East: 31% Integrity (does not keep promises made during election campaign, only talks, does nothing, liar, actor, reliable/unreliable...) Integrity (only talks, does nothing, sells her own grandmother, obscure, reliable, honest) Total: 15% West: 15% East: 15% Total: 13% West: 13% East: 13% Leadership qualities ( doer, dynamic, forceful/not forceful, reformfreudig...) Leadership qualities (assertive in the context of a men s world, determined, fighter, forceful/not forceful, straightforward/not straightforward...) Total: 10% West: 10% East: 8% Total: 12% West: 12% East: 10% Don t know/refusal Total: 3% West: 3% East: 4% Total: 5% West: 5% East: 6% Base: 4115 entries for Schröder (West Germany: 2870, Germany: 1245), 3678 entries for Merkel (West Germany: 2621, East Germany: 1057). Open question asked after a closed question on sympathy evaluation. Three entries for each respondent at most. not changed dramatically in the last 45 years. Thus, in the end, the 2005 election was the rule, not the exception, with regard to the importance of candidates for explaining vote choice in German elections. Our analysis of candidate images in the 2005 election also showed that Gerhard Schröder managed to run an impressive race of catch up with Angela Merkel between the time the election was set and election day. In large measure, his success in turning voters opinions about him around was due to his ability to mobilise traditional SPD partisans as Election Day drew near. While these voters were reluctant to support the SPD and its chancellor in May, they decided to follow their party identification in the end. Schröder s success in mobilising core supporters was substantially connected the issue of social justice, which became a major subject of discussion. As we point out, numerous strategic and tactical mistakes by the Christian Democrats gave Schröder an opening that he was more than willing to take advantage of. In the end, Gerhard Schröder was not only considered more likeable and energetic than Angela Merkel, but also to be more competent with regard to the key issue in the campaign: social justice. Assuming that the results presented are not an artefact of the methodology we employed, it may be surprising to some that we find no evidence that candidates matter more than they used to. Given the transformation of German society since the

CANDIDATE IMAGES IN THE 2005 ELECTION 497 war the concomitant decline in partisanship, the rise in media democracy, as well as the ongoing transformation associated with German unification it would be natural to expect voters to rely more on their attitudes toward the chancellor candidates when deciding whom to vote for. Moreover, the continued Americanisation of German politics in the form of U.S. style debates between the two major candidates for the chancellorship a first in Germany in 2002, and continued in 2005 should focus voters attention on candidates over parties and on the horse race over substance. Yet, there is little hard evidence to support this conjecture. And there may be good reasons for why this is the case. Specifically, we would argue that the absence of an increase in the importance of candidates for the vote in the German context has to do with the nature of the political system a parliamentary democracy where parties play the central role for the functioning of the political system at all levels. In such an environment, coupled with an increased importance of television, candidates certainly help parties to communicate their appeals, but they do not drive parties campaigns. Interestingly, the parallel weakening of partisan ties, which suggest more room for candidate effects, may well be consistent with our results because voters who lack strong partisan ties also tend to be more sophisticated than they used to be. Thus, instead of being distracted by candidate personalities, such sophisticated voters should be more likely to focus on hard information, such as issues. In addition, to the extent that they do rely on candidate images, they should focus on those characteristics of candidates that provide substantial information about the issues or parties. In the end, and despite the political drama that ensued, when it came to the candidates for chancellor and their influence on voter behaviour, the 2005 elections was a very normal election, though it did produce an historically unprecedented outcome: Germany s first female chancellor. ACKNOWLEDGMENTS The data used were made available by the Zentralarchiv für empirische Sozialforschung (ZA), University of Cologne and the Forschungsgruppe Wahlen e.v., Mannheim. Neither the principal investigators nor the Zentralarchiv are responsible for any of the analyses or interpretations of the data in this study. Further information on the principal investigators, data collection, and other details can be found at: http://www.gesis.org/za/. NOTES 1. Theodor Eschenburg, Zur politischen Praxis in der Bundesrepublik. Band II: Kritische Betrachtungen 1961 1965 (München: Piper, 1966). Karlheinz Niclauß, Bestätigung der Kanzlerdemokratie? Kanzler und Regierungen zwischen Verfassung und politischen Konventionen, Aus Politik und Zeitgeschichte 20/99 (1999), pp.27 38; Angelika Vetter and Oscar W. Gabriel, Candidate Evaluations and Party Choice in Germany, 1972 94: Do Candidates Matter? in Christopher J. Anderson and Carsten Zelle (eds.), Stability and Change in German Elections: How Electorates Merge, Converge, and Collide (Westport, London: Praeger, 1998), pp.71 98. 2. Christopher J. Anderson and Frank Brettschneider, The Likeable Winner versus the Competent Loser. Candidate Images and the German Election of 2002, German Politics and Society 21 (2003), pp.95 118.