Candidate Voting on the Rise? Attitudinal Stability and Change During an Election Campaign

Transcription

1 Candidate Voting on the Rise? Attitudinal Stability and Change During an Election Campaign Elena Wiegand and Hans Rattinger, University of Mannheim, Germany Paper presented at the 23rd World Congress of Political Science of the International Political Science Association (IPSA), July 19 to 24, 2014, Montréal Abstract: The aim of this paper is to examine the intra-individual processes behind shifts in the weight of candidate evaluations on voting. Previous research has not paid sufficient attention to alternative explanations for the so called priming effects. As some priming effects may result from learning and from attitudinal formation processes over the course of an election campaign, we will consider the stability of attitudes relevant for voting and describe theoretically relevant subgroups in which the effects of candidate evaluations on vote decisions should increase over the course of an election campaign. Using data from a shortterm campaign panel collected at the occasion of the 2013 German federal election within the German Longitudinal Election Study (GLES), there is only little evidence for shifts in the weight of candidate evaluations on vote choice over the course of the election campaign. Keywords: candidate evaluations, priming effects, learning, attitudinal change, panel data First draft please do not cite without permission of the authors.

2 1. Introduction When discussing the importance of attitudes toward political leaders for voting behavior, electoral research mostly refers to two trends in party-based democracies. First, a decrease in partisanship, called dealignment, can be observed in many Western democracies like the U.S., UK as well as in Germany (Dalton, McAllister, and Wattenberg 2000: 25; Falter and Schoen 1999: ). This erosion of partisan attachments makes it more likely that voters become more volatile between elections and throughout election campaigns. If fewer people within the electorate have long-standing partisan attachments there should be more room for short-term factors to determine their vote choice. Second, a trend often called personalization of politics has also emerged. This again has two sides: On the one hand, media focus on political leaders is getting stronger. On the other hand, the style of election campaigning, i.e. the presentation of politics, has also changed over the last decades. Both for the media and campaign staff it is easier to present and promote a party s politics via popular politicians. Following this reasoning, it was argued that if there are more and more voters without long-standing party loyalties and if the media favor to personalize political issues, the impact of attitudes toward candidates on voting behavior should increase over time. However, previous research has not supported this hypothesis for Germany. A secular rise of the importance of candidate evaluations for vote decisions cannot be observed over the longer run (Brettschneider 2001). Moreover, specific constellations of particular elections play a decisive role (Kellermann, 2008). Although there is no long-term rise of candidate voting, some researchers argue that the popular hypothesis has some merit - not as a secular trend, but over the course of election campaigns (Schoen 2004; Ohr et al. 2013). They argue that campaigns can alter the importance of criteria for vote decisions. Indeed, parties try to convince voters of their political concepts and candidates by presenting their strategies for the future, highlighting their performance in the last election period, but also criticize each other in order to win votes. Hence, voters might change their attitudes toward party platforms and candidates, this could be called persuasion. Other voters might attribute more weight to party candidates as the campaign progresses and ascribe less importance to issues because the news media had been focusing mainly on candidates and less on substantive party positions. Such shifts in the weight of voters judgments can be called priming effects. Ohr, Klein and Rosar (2013) found evidence for such an increase 1

3 of the impact of candidate attitudes on vote decisions over the German election campaign of The aim of this paper is to examine the intra-individual processes behind such priming effects. Previous research has not paid sufficient attention to alternative explanations for this process. This sort of priming effects may result from learning and from attitudinal formation processes over the course of an election campaign. We will therefore consider the stability of attitudes relevant for voting and describe theoretically relevant subgroups in which the effects of candidate evaluations on vote decisions should increase over the course of an election campaign. As the title of this paper suggests, we want to investigate how strong voting preferences are affected by candidate evaluations and whether the weight of candidate evaluations increases over the course of the election campaign because of such priming of candidates. These questions will be addressed using data from a short-term campaign panel collected at the occasion of the 2013 German federal election within the German Longitudinal Election Study (GLES). 2. Theoretical Framework: Attitudes towards Chancellor Candidates and Voting in the Federal Election of 2013 Although German parties always make a big point of how strongly they differ in their positions about substantive issues, candidates for chancellor still played a leading role in the 2013 Federal election campaign. As in the previous 2009 campaign the CDU/CSU tried to capitalize on Angela Merkel s popularity by almost exclusively focusing on her and running a very strongly candidate centered campaign. She was by far the dominant figure in TV ads and election posters, stressing the good economic figures as indicators of her government s successful work. Somewhat surprisingly given the issue s general economic importance, the European banking and credit crisis was absent from the 2013 election campaign (Mader 2014, 355). The SPD nominated Peer Steinbrück as their candidate for chancellor. He had earlier been the German finance minister under chancellor Merkel in the grand coalition. During the banking and credit crisis, he had demonstrated his expertise by preventing major harm to the German economy and voters savings. He supported also the highly controversial labor market reforms initiated by the last SPD chancellor Gerhard Schröder 2

4 in 2003, an issue that has divided the left and right wing of the SPD. Therefore, Steinbrück was less attractive for the traditional social democratic clientele, but he had the potential to attract votes from independents and CDU/CSU sympathizers (Mader 2014, 354). But only a few days after the announcement of Steinbrück as the SPD-candidate for chancellor, he dominated the headlines because of his high discretionary earnings from speaker fees he had received during the last legislative period. Steinbrück was much in demand as guest speaker on financial matters, because of his role as crisis manager during the banking and credit crisis. Therefore his credibility as an advocate of the regular voters interests was broadly discussed (Mader 2014, 354). Several personal scandals followed and consequently found broad attention in the media. Therefore, the election campaign of 2013 can be heavily called candidate-centered, and priming effects of candidate evaluations on voting decision can be expected. In a next step, we now want to discuss how this campaign could favor increases in the impact of candidate evaluations on voting decisions. In empirical electoral research the social psychologically inspired Michigan Modell (Campbell et al. 1960) has turned out to have quite high explanatory power for voting behavior in western democracies. In many decades of research three main factors became have been established as determining voters decisions: These are an emotional attachment to a political party, issue opinions and candidate evaluations. Highest explanatory power is assigned to party identification which motivates individuals to vote loyally for one party. It has often shown that a long-term party identification stabilizes the voting decision at the individual level (e.g., Rattinger/Wiegand 2014). However, vote decisions can also be influenced by issue opinions and candidate evaluations. But the assessment of candidates by voters heavily depends on media reporting about them as citizens are responsive to campaign information about candidates (Lodge et al. 1995; Miller/Krosnick 1996). This follows because scarcely any citizen knows the candidates of the parties personally. This lack of personal experience forces them to derive their images of the candidates mainly from media coverage (Brettschneider 2005). When campaigns are candidate-centered, it seems to be particularly useful to pay close attention to the activation of attitudes toward candidates and to their effect on the voting decision down the line. Research by Iyengar and Kinder (1987) showed that in their decision-making individuals rely on information that is present and available. During the 3

5 decision making process activated attitudes should be relevant for the final voting decision, with others completely being irrelevant (Iyengar/Kinder 1987, 63). Zaller s (1992) RAS-model suggests a similar mechanism by which individuals arrive at their judgments. Voters might have many opinions about parties and candidates that they could potentially use as criteria for their voting decision. But individuals only use those criteria that are most present at the moment of decision making (Zaller 1992, 1). Candidatecentered election campaigns and media coverage suggest to citizens to use the evaluation of candidates as a key criterion for their decision making when getting closer to election day. The TV debate between the two chancellor candidates in 2013 took place three weeks before election day. The frequency of TV ads with both chancellor candidates and news coverage increased just before election day. If attitudes toward candidates are more heavily activated than issue opinions and therefore are more likely to be taken into account for decision-making, a priming effect of the assessment of candidates for voting decisions can be assumed. Thus it can seem plausible that the influence of candidates on vote decisions increases during the election campaign. It is not plausible, however, that all voters are affected equally. Allowing for voter heterogeneity we would assume that the increase in the influence of candidates applies to some voters but not to others. Voters can arrive at their final vote decision based on many separate evaluations of the parties or only on one single decision criterion. For some people the future chancellor plays an essential role for their vote choice, while the parties competence in solving concrete issues has priority for others. The big question then is which criteria are applied within which groups of voters. Therefore, it is mandatory to examine the potentially influential factors for voting decisions and their development over the course of the campaign more closely. Thus, in the next step, we want to discuss for which voters an increasing candidate effect over the election campaign is more likely to emerge. We here allow for two sources of heterogeneity among voters: their amount of political knowledge and the stability of their political preferences. Both are particularly important for the impact of candidate evaluations on vote decision-making processes. Lenz (2009) was among the first to note that previous priming results are the product from learning or attitudinal change. As the campaign proceeds, even less informed citizens get information 4

6 about candidates and party programs, so that these aspects may become increasingly important for their vote choice. Especially knowledge of chancellor candidates plays an important role, because if voters are not familiar with the candidates their vote decision will not be affected by candidate evaluations. Individuals who at the beginning of the campaign do not know that Angela Merkel is the CDU candidate for chancellor, or that Peer Steinbrück is the SPD candidate, cannot use their evaluations of Merkel and Steinbrück as decision criteria for voting either for the CDU or the SPD. If they obtain information during the campaign and learn that Merkel belongs to the CDU and Steinbrück to the SPD, they can now draw upon their candidate evaluations for their voting decision. Consequently, the impact of candidate evaluations will rise during the election campaign for those who learn the party affiliation of the chancellor candidates. This results in our first hypothesis: H1: The effects of candidate evaluations on the vote increase among those individuals who acquire new knowledge about the chancellor candidates during the course of the campaign. The most obvious mediator variable for priming effects is the Michigan model s very own partisan identification (Campbell et al. 1960). The closer one feels to a party, the less the chance that other factors determine voting preferences before or during campaigns. Nonpartisans, however, cannot use party identification as a decision criterion; thus a candidate-centered campaign should lead to a stronger effect of candidate evaluations for those individuals. In addition, citizens without issue orientations cannot use issues as a decision criterion; that is why a candidate-centered campaign should also lead to an increasing effect of candidate evaluations among them. Taken together, we thus have clear expectations about differentiated effects of candidate evaluations on vote decision: H2a: Over the course of the campaign, the effect of candidate evaluations on the vote increases among individuals without party identification. H2b: Over the course of the campaign, the effect of candidate evaluations on the vote increases among individuals who do not perceive any party capable of solving the most important political problem. 5

7 Not only people without longstanding partisan feelings might base their final vote decision on their candidate evaluations. It should be considered that party identification might be open to change as voters receive new information. While predispositions can be activated during an election campaign (see Lazarsfeld et al. 1968), information may also be supplied that can lead to changes in citizens partisan feelings and attitudes. Some events during election campaigns may convince voters to update their vote intention and simultaneously to reconsider their partisan feelings. Some voters can move from one party to another or to independence (or back) because they are dissatisfied or have been influenced by events to update their beliefs about their party and its performance. However, we are not concerned here with investigating which mechanisms might lead to such volatility of partisanship. Instead, we will examine the effects of changing partisan affiliations. Individuals who turn away from their party identification over the course of the campaign will base their vote decision on other attitudes. In a candidate-centered campaign voters will rather turn to rely on candidate evaluations as their decision criterion. Similarly, individuals who abandon their initial issue orientation reported at the beginning of the election campaign and state that no party is competent in solving relevant problems cannot use it as decision criterion and will rather rely on candidate evaluations as a decision criterion in a candidate-centered campaign. Hence it follows that: H3a: Over the course of the campaign, the effect of candidate evaluations on the vote increases among individuals who turn away from their earlier party identification. H3b: Over the course of the campaign, the effect of candidate evaluations on the vote increases among individuals who, at the end of the campaign, do not perceive any party capable of solving the most important political problem. Besides these election campaigns might also have effects called activation and reinforcement. Election campaigns might arouse political interest and an increase in media exposure (Lazarsfeld et al. 1968, 75ff.). As citizens get informed about political candidates and party manifestos they might over the course of the campaign develop attitudes relevant for the election. Some non-partisans may receive new information which transforms them into partisans. Individuals who generate a party identification over the course of the campaign will rather rely on this decision criterion than on candidate evaluations. Furthermore, election campaigns can give rise to attitudes about issue competence. As citizens develop issue orientations over the course of the campaign they 6

8 can use issues as a decision criterion for voting. Even if campaign is candidate-centered, it stimulates individuals to think about issues so that the new developed attitude may become relevant for voting. Therefore, individuals who develop issue orientations over the course of the campaign will rather rely on this decision criterion than on candidate evaluations. H4a: The effect of candidate evaluations on the vote declines among those individuals who acquire a party identification over the course of the election campaign. H4b: The effect of candidate evaluations on the vote declines among those individuals who begin to ascribe issue competence to a party over the course of the campaign. Before going on to test our hypotheses, we will now describe the data at hand and the operational definitions for the theoretical concepts. Because we will analyze the effect of candidate evaluations for different voter groups, we need to clarify these classifications in the following. 3. Data and measurement To examine the change of criteria during decision-making processes we require information about the same individuals at multiple points in time. Therefore we will analyse a campaign panel from the German federal election 2013 (which took place on September 22). This panel study of the German Longitudinal Election Study (GLES) was conducted online from June 20, 2013, to October 4, Six times before and one time after the election one and the same persons were asked to participate in a survey about their political attitudes and their voting behavior individuals took part in at least four and in all seven interviews (for a more detailed description of the data see Rattinger et al. 2014). Because the study was conducted online, the participants are, on average, very web-affine and not necessarily representative of the general population in Germany. However, since our goal is to study intra-individual processes of change, and not to describe correct distributions for the entire population, this should not be a disadvantage for our analysis. The evidence offered in studies is based on cross-sections. While these data seem appropriate for detecting aggregate shifts over an election campaign, they are much less useful for investigating changes in the causal impact of 7

9 candidate evaluations on voting behavior. With panel studies, we can monitor individual changes of political attitudes and vote intention over the course of an election campaign. The indicators for our key concept candidate evaluations are the feeling thermometers for the two chancellor candidates Merkel and Steinbrück. Respondents were asked to state on an eleven-point scale what they think about these leading politicians. Their answers were rescaled from minus one I have a very negative view of this politician to plus one I have a very positive view of this politician. As outlined above, other studies claim that candidates gain in importance because there is rising media coverage during the campaign. We assume, however, that priming effects are not plausible for all voters. Therefore we want to evaluate the priming hypothesis for different voter groups. Based on measurements of political knowledge we can study the learning processes during the election campaign and the development of the impact of candidate evaluations by political sophistication. To measure knowledge about political actors respondents were asked to match politicians portrait pictures with their respective parties. The question was simply: Which party do the following politicians belong to? This question was part of the first, third and seventh panel waves. Some voters already knew the politicians at the beginning of the election campaign and allocated them correctly to their respective party, every time they were asked to do so. Other respondents did not yet know the correct party in June, but later classified the politicians correctly in September. While these individuals obviously received new information during the election campaign to perform the task correctly, others never learned the correct answer. And still other voters forgot the correct party affiliation of both candidates from June to September. Before going on to describe the occurrence of these different voter types, we have to mention how the explanatory variables for the subsequent regression analyses of vote choice were measured. As has been argued above, the strength of party identification and attitudes about issues should strongly affect the vote decision. Therefore we have to investigate the effects of candidate evaluations while controlling these two factors. We constructed an indicator for respondents strength of party identification by rescaling responses into an eleven-point scale, ranging from minus one to plus one. Minus one stands for a very strong identification with a party someone does not vote for, while zero 8

10 indicates no party identification at all, and plus one a very strong identification with the party someone also wants to vote for (for question wording and coding details see the Appendix). Furthermore, we have rescaled the answers to the question about issue competence so that the variable represents the consistency of this attitude and ranges from minus one if someone mentions a party as competent he does not want to vote for via zero for indifferent assignments of competence to plus one if a respondent mentioned the same party he also wanted to vote for. 4. Results First we should point out that the sample this analysis is based on is altogether highly interested in politics, and well-informed about the political affiliations of chancellor Merkel and her opponent Steinbrück. 95 percent of the respondents were able to correctly associate Angela Merkel with the CDU when shown her picture (see Table 1). While five percent of the respondents did not know that chancellor Merkel belongs to the CDU at the beginning of the election campaign, 3.5 percent acquired that knowledge over the course of the election campaign. In contrast, only 81 percent of the respondents successfully matched Peer Steinbrück with the SPD. During the election campaign, about 12 percent learned his party affiliation. Table 1: Knowledge of candidates for chancellor in the campaign panel 2013 Wave 1 Wave 7 Merkel Steinbrück Voter Groups June September N % N % Knew before Yes Yes 3, , Learned No Yes Forgot Yes No Never Learned No No Note: Entries are column percentages. As outlined above, we are also interested in the development of party identification and attitudes on political issues. Both may change during the 2013 election campaign. The German standard indicator for party identification was contained in panel waves one, three, five and seven. Based on these measurements, we can investigate the impact of candidate evaluations in relation to the stability and change of party identification during 9

11 the election campaign. This analysis deals only with the development from the first wave in June to last wave in September and ignores any switches in between. Not surprisingly, most people have stable party attachments during the election campaign. Furthermore, about 20 percent of our panel participants did not feel close to any party (see Table 2). Eight percent of the respondents developed a party identification or turned away from it over the course of the campaign. The same share of respondents (eight percent) switched their party identification from the first to the last panel wave. Individuals with no party identification and those who develop or lose a party attachment are particularly interesting for our subsequent analyses. Table 2: Partisan stability across campaign panel 2013 Wave 1 Wave 7 Voter Groups June September N % Stable Party ID Same Party 2, Changed Party ID Different Parties No Party ID No No Lost Party ID Yes No Acquired Party ID No Yes Note: Entries are column percentages. As some citizens base their vote decision on the solutions the parties offer for the prevailing major societal problems, we also want to control the impact of the stability of these evaluations. During the election campaign some voters might change their consideration of a particular party as being competent to solve these problems in Germany. These measurements enable us to build groups of voters with different developments of these evaluations. While almost 40 percent reported the same party as most competent for fixing urgent problems in Germany, every time they were asked for, 12 changed their issue competence evaluation from June to September (see Table 3). About one fourth of the respondents have no positive attitudes about the issue competence of the parties at all. The latter three voter groups are particularly relevant for our subsequent analyses: About 15 percent of the respondents developed attitudes on issue competence whereas approximately ten percent have abandoned such attitudes over the course of the campaign. 10

12 Table 3: Stability of issue competence across campaign panel 2013 Wave 1 Wave 7 Voter Groups June September N % Stable attitudes on issue competence Same party 1, Changed attitudes on issue competence Different parties No party competent No party No party Lost attitudes on issue competence Any party No party Crystallized attitudes on issue competence No party Any party Note: Entries are column percentages. In the next step the connection between the feeling thermometers of politicians and the second votes are analysed in order to answer our question whether the impact of candidates on the vote change over the course of the election campaign. Consistent with recent work on priming effects on candidate voting, we ran multivariate logistic regression models (see Bartels 2006: 85). We defined voting intention (and behavior) as the dependent variables as dummy variables, indicating a vote choice for the CDU or SPD as one, and all other responses as zero. The analyses comprise only those with a vote intention; respondents who wanted to abstain or have abstained in the federal election are not considered. We only report the regression coefficients representing the effects of candidate evaluations on vote intentions and compare these coefficients over time, but the effects of partisanship and issue preferences are also included in the models. An increase of those regression coefficients is interpreted as evidence for a growing importance of candidate evaluations for vote choice. We estimate Michigan type models in total and separately for the different voter groups for the first and seventh panel wave and then compare the logit coefficients and the relative additional explanatory power within the seven voter groups between the two points in time. Voter groups who learned the political affiliations of candidates for chancellor and those who have no party identification or no attitudes on issue competence as well as those who develop or lose such attachments or attitudes are particularly interesting for our subsequent analyses. Not surprisingly, all these three determinants together in the aggregate do a good job in explaining variations in voting patterns. The more an individual is attached to the CDU or SPD, the more likely he or she is to vote for these parties. Respondents naming the Conservatives or the Social Democrats as the most competent party for solving urgent 11

13 problems of the country are also more likely to vote for this party. Similarly, candidate evaluations have a significant influence on vote choice in each of the two panel wave. However, we here are interested not in whether candidate preferences shape vote choices but in the stability of the impact of candidate evaluations on voting over the course of the election campaign. In the aggregate of all respondents, it appears that there is no rising effect of candidate evaluations over time (see first row of Table 4). Considering only the regression coefficients for all voters, our hypothesis of a priming effect of campaigns cannot be confirmed. However, separating the voters into the groups described above, increases of the candidate effect over time can be detected. The second row of table 4 shows the difference between the estimated effects of candidate evaluations on voting for those who learned about the party affiliation of Merkel or Steinbrück. If a respondent, e.g., does not know at the beginning of the election campaign that Angela Merkel is the chancellor (candidate) of the CDU, his or her vote choice can hardly be based on sympathies for Merkel. Voters who learn about the party affiliation of Merkel or Steinbrück during the election campaign use their sympathy rating as a key criterion for their vote choice in September. Therefore, our first hypothesis can be confirmed. This result also indicates that an increased effect of candidates can occur, not only because the media raised attention to them, but also because they inform voters about the candidates. Such an increased effect of candidate evaluations can also be observed for SPD voters without a party identification. They rely more strongly on candidate evaluations as the campaign proceeds. In contrast to our expectations we find that the effect of candidate evaluations on voting did not increase over time for CDU voters without a party identification: The regression coefficient of candidate evaluations is a little lower in September than in June. Because the difference of these two regression coefficients is not significant, there might be no increasing candidate effect among CDU-voters without party identification. Therefore, the hypothesis 2a cannot be confirmed completely. An increase of the impact of candidate evaluations is only the case for SPD-voters without party identification. Consistent with our third hypotheses, we see rising effects of candidate evaluations on vote choice among those who lost their party identification. In addition, the estimated 12

14 impact of candidate evaluations on voting increased between June and September among those who lost attitudes on issue competence. But the differences between the coefficients in June and September are small, so that it does not suggest a substantial priming effect. Even if the campaign primes the candidates for chancellor, their weights do not increase for people who have lost their political attachments or attitudes as well. Table 4: Effects of candidate orientations on the CDU/CSU or SPD vote in June and September 2013 CDU/CSU SPD Voter Groups June Sept. Diff. June Sept. Diff. All 4.04*** 4.12*** *** 2.79***.16 Learned Merkel/Steinbrück * 2.67* 1.33* 3.35*** 2.02 No Party ID 6.60*** 5.85*** * 4.67*** 3.01* No party competent 5.49*** 4.57*** *** 3.14***.48 Lost Party ID 3.39*** 5.27*** ** 4.35*** 2.03 Lost attitudes on issue competence 3.73** 4.05** * 4.02*** 2.00 Acquired Party ID 4.32*** 3.57** *** 3.30*** -.61 Crystallized attitudes on issue competence 4.09*** 2.85*** ** 2.61***.35 Note: Entries are unstandardized logit coefficients. Dependent variables are vote intention for CDU or SPD (1 respectively) versus other party (0). As control variables party identification and issue attitudes are included. Significance: * p < 0.05; ** p < 0.01; *** p < The results are also consistent with our hypothesis 4a. Individuals who mentioned no party identification in June but acquire one over the course of the election campaign show also a little bit smaller candidate impact on vote choice for the CDU as well as the SPD. But the impact of candidate evaluations did not change significantly over time. The weight people attached to candidates in deriving their vote decisions does not decrease simultaneously to the development of partisan feelings. The election campaign might have activated citizens to acquire party identifications but this does not consequently lead to use it as key criterion for voting. Furthermore, the reasoning above suggested that the impact of candidate evaluations should diminish among those who develop attitudes on issue competence over the course of the campaign (H4b). As the campaign proceeds candidate 13

15 evaluations became, however, not less important among those. The candidate effects remain constant from June to September. While the logit coefficients of candidate evaluations provide some first evidence for an increased influence over time, we also want to take a look at the explanatory power of candidates in our models (Mc Fadden s Pseudo R²). In these multivariate regression models the explanatory power of candidate evaluations is not very high, because as it is well known party identification has a high explanatory power over vote choice. Still they are quite important for the explanation of vote choice in that the variance explained significantly differs between models with and without candidate evaluations. It should be mentioned that the explanatory power of all models decreases toward the end of the election campaign. We would argue that not all voting behavior is the outcome of the interplay between party identification, evaluations of chancellor candidates and of issue competence. Quite some variables that have not been included in our models, like, e.g., strategic calculations desirable and possible coalitions, might impact vote choices. Therefore it seems reasonable not only to consider the absolute explanatory power of candidate evaluations but also the relative one. Accordingly, we have calculated the additional explanatory power of the candidates as a function of the overall explanatory power of the respective model (Table 5). Based on this measure, we see that candidate evaluations gain relative additional explanatory power over the course of the election campaign in the theoretically relevant subgroups. In the aggregate of the whole electorate, however, the hypothesis of a priming effect cannot be confirmed. Again it proves more useful to focus on specific voter groups. If our hypotheses of priming holds true, there should be larger additional explanatory power of candidate evaluations over time. And indeed, the differences of the relative explanatory power between June and September point in the right direction, but most of the differences are too small to be meaningful. The second row of table 5 shows, e.g., the relative explanatory power of candidate evaluations on vote decisions and its difference between June and September for those who learned about the party affiliation of Merkel or Steinbrück. While the variation in voting explained by the evaluations of Merkel only increases about 14 percent points, the relative explanatory power of attitudes towards Steinbrück rises around.3 among the learners. The least increase of relative explanatory power is shown for attitudes towards Merkel among those without party identification or 14

16 reporting that no party is competent. While the explanatory power of candidate evaluations for non-partisans is already high in June and cannot increase even more, the explanatory power of attitudes towards Merkel remains constant among individuals who do not perceive any party capable of solving the most important political problem at a medium explanatory level. In contrast, the relative explanatory power of attitudes towards Steinbrück differs sufficiently between the first and the last panel wave among those without or lost party identification. More than half of the explained variance of voting is based on candidate evaluations. The apparent magnitude of candidate priming is 48 percent points. Table 5: Relative explanatory power of attitudes towards Merkel or Steinbrück on vote decisions CDU/CSU SPD Voter Groups June Sept. Diff. June Sept. Diff. All Learned Merkel/Steinbrück No Party ID No party competent Lost Party ID Lost attitudes on issue competence Acquired Party ID Crystallized attitudes on issue competence Note: Entries are the relation of additional explanatory power and overall explanatory power. Consistent with our third hypotheses, the effect of candidate evaluations increases over time among individuals who lost their party identification or attitudes on issues. For those voter groups, candidate evaluations became quite important for voting over the course of the campaign. At least, it should be mentioned that the relative explanatory power of candidates decreases a little, as individuals develop partisan feelings or begin to regard the 15

17 CDU/CSU as competent to deal with important issues (H4). However, the explanatory power of attitudes towards Steinbrück remains constant among individuals who develop attitudes on issues during the election campaign. 5. Conclusion The aim of this paper was to compare and investigate the impact of candidate evaluations on voting over the German election campaign of According to the Michigan model, party identification, candidate evaluations and issues are relevant for voting decisions. Previous studies assume an increase of the effect of candidate evaluations over the election campaign due to the personalization of campaigning outlined above. Consequently, the question was raised, if candidate evaluations become more and more important for voting as the campaign proceeds. It is, however, not reasonable to assume that all voters are affected equally. That is why we have examined which voters characteristics favor a stronger effect of candidate evaluations over the course of an election campaign. We have tested several hypotheses for different voter groups and find little evidence for an increased candidate impact. Candidate evaluations play a major role in shaping voting behavior, but its effect on voting does not strongly vary between the beginning and the end of the campaign. While the differences of the regression coefficients between June and September point in the right direction, most of the differences are too small to be meaningful. We see that the interplay between party identification, evaluations of chancellor candidates and attitudes towards issue competence in determining vote choice is moveable during the 2013 election campaign but only for some voter groups. Attitudes towards Steinbrück become increasingly important for their vote choice among those who learned his party affiliation, have no party identification or lost their partisan feelings. For others, the weight of candidate evaluations remains more or less the same. Especially when the party identification or issue orientations ultimately determine the formation of preferences for voting, there is no rise of the impact of candidate evaluations. This is why campaigns may stimulate not only thinking about candidates but also other issues being relevant for voting. 16

18 The short-term panel data at hand enable us to gain insights in the intra-individual decision-making processes of voters. This paper is just a first step to analyze learning processes and individual change of attitudes during the German election campaign of Further research might also model the effect of changed candidate evaluations over time. References Bartels, Larry M Priming and Persuasion in Presidential Campaigns. In: Henry Brady/Richard Johnston (ed.). Capturing Campaign Effects. Ann Arbor: The University of Michigan Press: Brettschneider, Frank Candidate-Voting. Die Bedeutung von Spitzenkandidaten für das Wählerverhalten in Deutschland, Großbritannien und den USA von 1960 bis 1998 In: Hans-Dieter Klingemann/Max Kaase (ed.). Wahlen und Wähler. Analysen aus Anlass der Bundestagswahl Wiesbaden: Westdeutscher Verlag: Brettschneider, Frank Massenmedien und Wählerverhalten In: Jürgen Falter/Harald Schoen (ed.). Handbuch Wahlforschung. Wiesbaden: Westdeutscher Verlag: Campbell Angus/Philip E. Converse/Warren E. Miller/ Donald E. Stokes (ed.) The American Voter. Chicago: University of Chicago Press. Dalton, Russel J./Ian McAllister/Martin P. Wattenberg The Consequences of Partisan Dealignment In: Russel J. Dalton/Martin P. Wattenberg (ed.) Parties without partisans. Political change in advanced industrial democracies. Oxford: Oxford University Press: Falter, Jürgen W./Harald Schoen Wahlen und Wählerverhalten In: Thomas Ellwein/Everhard Holtmann (ed.). 50 Jahre Bundesrepublik Deutschland. Rahmenbedingungen Entwicklungen Perspektiven. Opladen: Westdeutscher Verlag: Iyengar, Shanto/Donald R. Kinder News That Matters: Television and American Opinion. Chicago: University of Chicago Press. Kellermann, Charlotte Trends and Constellations. Klassische Bestimmungsfaktoren des Wahlverhaltens bei den Bundestagswahlen Baden-Baden: Nomos. 17

19 Lazarsfeld, Paul F./Bernard Berelson/Hazel Gaudet The People's Choice. How the Voter Makes up His Mind in a Presidential Campaign. 3.Auflg. New York: Columbia University Press. Lenz, Gabriel Learning and Opinion Change, Not Priming: Reconsidering the Priming Hypothesis American Journal of Political Science 53(4): Lodge, Milton/Marco R. Steenbergen/Shawn Brau The responsive voter: Campaign information and the dynamics of candidate evaluation. American Political Science Review 89: Mader, Matthias The German federal election, September Electoral Studies. doi: /j.electstud Miller, Joanne M./Jon A. Krosnick News Media Impact on the Ingredients of Presidential Evaluations: A Program of Research on the Priming Hypothesis. In: D. C. Mutz & P. M. Sniderman (ed.), Political Persuasion and Attitude Change (pp ). Ann Arbor, MI: University of Michigan Press. Ohr, Dieter/Markus Klein/Ulrich Rosar Bewertungen der Kanzlerkandidaten und Wahlentscheidung bei der Bundestagswahl In: Bernhard Weßels/Harald Schoen/Oscar W. Gabriel (ed.). Wahlen und Wähler. Analysen aus Anlass der Bundestagswahl Wiesbaden: VS Verlag für Sozialwissenschaften: Schoen, Harald Winning by Priming? Campaign Strategies, Changing Determinants of Voting Intention, and the Outcome of the 2002 German Federal Election. German Politics and Society, 22, Rattinger, Hans/Sigrid Roßteutscher/Rüdiger Schmitt-Beck/Bernhard Weßels/Christof Wolf Wahlkampf-Panel (GLES 2013), GESIS Datenarchiv, Köln: ZA5704 Datenfile Version 1.0.0, doi: / Rattinger, Hans/Elena Wiegand Volatility Around the Clock? Attitudinal Stability, Attitudinal Change, and Voters Volatility. In: Bernhard Weßels, Hans Rattinger, Sigrid Roßteutscher, Rüdiger Schmitt-Beck (ed.): Voters on the Move or on the Run? Information-processing and Vote Choice in a Complex World. Oxford: Oxford University Press Zaller, John R The nature and origins of mass opinion. Cambridge: Cambridge University Press. 18

20 Appendix: Question Wording and Construction of Variables Voting intention/behavior: Dummy variable for the second vote of CDU or SPD (1) vs. other parties (0). Question: You have two votes in the federal election. The first vote is for a candidate from your constituency and the second is for a party. Please tell me where you will place your checkmarks on the ballot./ Where did you place your checkmarks on the ballot? Strength of party identification: Eleven-point scale ranging from -1 (very strong party identification with a party s.o. does not vote for). 0 (no party identification) to 1 (very strong party identification with the same party s.o. wants to vote for). Question: Many people in Germany feel close to a particular political party for a longer period of time even if they occasionally vote for another party. What about you? In general terms, do you feel close to a particular political party? And if so, which one? Follow-up Question: All in all, how strongly or weakly attached are you to this party? Attitudes on issue competence: three-point scale, ranging from 1 (if issue competence is ascribed to another party s.o. does not vote for), via 0 (if indifferent, i.e., answers such as all parties are equally good, no party, or don t know ) to +1 (if issue competence is ascribed to the same party s.o. wants to vote for). Question: In in your opinion, what is the most important political issue Germany is facing? Follow-up Question: You consider... to be the most important issue. Which party do you think is best able to solve this problem? Attitudes towards candidates: Eleven-point scale ranging from -1 I have a very negative view of this politician to +1 I have a very positive view of this politician. Question: Now, please state what you think about some leading politicians. 19