Master’s Thesis
The Effect of Emotional Appeals in Political
Advertising on Party Brand Value and
Party Brand Personality Perceptions
Graduate School of Communication
Master’s Programme Communication Science
Supervisor: Dr. Jonas Lefevere
Author: Dario Pablo Morazán Irias Student ID: 11145412
Note: This thesis takes the format of a monograph, with permission of the supervisor.
Abstract
Based on theories from psychology in political communication and from political marketing research, this thesis examines whether emotional appeals in campaign posters have an effect on party brand value and brand personality perceptions among German voters. This is scien-tifically relevant, because no previous study has investigated the effect of emotional appeals on political brand perceptions and thus far very little research has examined political branding in multi-party systems. The present thesis fills this research gap. It has societal relevance, because, according to political communication research, emotions and political brand percep-tions can influence political behaviour. The thesis draws on two datasets, one collected in an online-survey and the other one gathered in an exit poll. Results indicate that enthusiasm ap-peals in campaign posters can positively affect voters’ perceptions of political party brand value and brand personality, especially among low-educated and partisan voters. Moreover, fear appeals affect party supporters’ brand perceptions positively and non-supporters’ brand perceptions negatively. Implications for voters and practitioners are discussed conclusively.
Introduction
Voters do not always make rational decisions. Emotional responses play an important role not only in their voting decisions, but also in voters’ processing of political advertising messages. Previous studies have examined whether and how emotional appeals in political advertisements affect voting decisions or attitudes towards political candidates (Benoit, Leshner, & Chattopadhyay, 2007; Brader, 2005; Chang, 2001). However, no study has inves-tigated the effect of emotional appeals in political advertisements on political party brand val-ue and party brand personality (PBP) perceptions among voters, although branding plays an important role in politics and electoral choice (Lees-Marshment, 2014; Smith & French, 2009; Winther Nielsen & Larsen, 2014).
A political brand can be defined as “political representations that are located in a pat-tern, which can be identified and differentiated from other political representations” (Winther Nielsen, 2015, p. 9). Accordingly, a political party brand serves as a useful heuristic that helps voters to quickly understand the main characteristics of a party, and it makes it easier to dis-tinguish a party from its competitors (Lees-Marshment, 2014). Moreover, it can induce politi-cally disinterested or ignorant voters to start paying attention to politics (Rutter, Hanretty, & Lettice, 2015). Parties use branding to get or maintain support, create a feeling of identity and to build a trusting relationship with the voters (Lees-Marshment, 2014). According to
(Winther Nielsen & Larsen, 2014), voters are likely to elect the party that they perceive to have the highest brand value, that is, the desirability of a political offering conditioned by its attached brand name.
From a relational perspective, political brands have the potential to establish and main-tain emotional relations between voters and parties (Winther Nielsen, 2015). Emotions are increasingly considered to be important for voter’s political behaviour (Brader, 2005, 2006; Gadarian & Albertson, 2014; Ryan, 2012). In fact, a number of studies have pointed to the importance of emotions in parties’ communications (Brader, 2005, 2006; Chang, 2001; Ridout & Searles, 2011).
Emotions can be defined as physiological responses to external objects or situations (Damasio, 2000). The intensity of an emotion depends on how relevant the object or situation is for an individual (Brader, 2005; Damasio, 2000). Emotional appeals are all types of com-munication that seek to elicit an emotional response from the receiver (Brader, 2005). Such appeals are increasingly used in political communication, because they have an impact on political behaviour (Benoit et al., 2007; Brader, 2005; Chang, 2001). Yet, it is still unclear to
what extent emotional appeals can aid parties in establishing and promoting their brand value and brand personality.
The present thesis is the first to combine the psychological concept of emotions with the political marketing related concept of political branding. More precisely, it is the first to measure the effect of emotional appeals in party advertisements on voters’ perceptions of par-ty brand values and brand personalities. Previous research on political branding has mainly focused on the US or the UK, which are considered to have a two-party system (Mair, 2002).1 However, the role of political party branding in multi-party systems has received relatively little attention by political communication scholars (Meister, 2016). Therefore, this thesis fo-cuses on Germany, which features a multi-party system.
Different results can be expected in a multi-party system, because it is more difficult for a party to be perceived as unique when there are more than two competing parties, which also often partly overlap with regard to issue positions and ownership, ideology, and market-ing strategy (Lees-Marshment, 2014; Rutter et al., 2015). As a consequence, party brands might play a more important role in a multi-party system than in a two-party system, because brands help voters to quickly understand what a party stands for and to distinguish parties from each other (Lees-Marshment, 2014).
Hence, by focusing on Germany, this thesis is one of the first studies to shed light on the role of political branding in a multi-party system. Moreover it addresses the relevance of emotional appeals in political advertising in Germany, where the campaigning environment is indeed less dramatic than in the U.S. (Plasser, 2002), but where emotions in political advertis-ing play an important role (Holtz-Bacha, Kaid, & Johnston, 1994; Kaid & Holtz-Bacha, 2006). Thus, the research question of the thesis is: What is the effect of emotional appeals in political advertisement on voters’ perceptions of political party brands in Germany?
This question has societal relevance, because the way people perceive the brand value of a political party can influence their decision whether or not to vote for that party (Winther Nielsen & Larsen, 2014). Moreover, a party brand is relevant to political interest and partici-pation, because it serves as a useful heuristic for voters to distinguish parties quickly while keeping the costs of informing themselves low (Lees-Marshment, 2014). It is also relevant for political parties, because it helps them to establish or maintain a trusting relationship with their voters (Lees-Marshment, 2014; Winther Nielsen, 2015).
11 While the U.S. has a typical two-party system, the UK has recently shifted to a three-party system due to the
rise of the right-wing UK Independence Party (UKIP). However, the previous studies have been conducted when the UK was still considered a two-party system.
The thesis is structured as follows: Firstly, relevant theoretical concepts are discussed. Secondly, the methodological approach is described. Thirdly, results are presented. Subse-quently, conclusions are drawn from the findings. The thesis ends with a discussion of the findings, the strengths and weaknesses of the study, suggestions for further research and the implications of the findings for voters and practitioners.
Theoretical Framework
In this section, firstly, the concepts of political branding in general as well as brand value and PBP in particular are discussed. Secondly, the role of emotional appeals in political advertising is outlined. Thirdly, it is discussed how emotional appeals in political advertising can affect brand value and PBP perceptions. Fourthly, the effects of pictures and slogans are compared. Based on the theoretical discussions, hypotheses are formulated.
Political Party Brand Value and Party Brand Personality (PBP)
In this thesis, two distinct aspects of brands will be examined, namely brand value and PBP. They are defined in the following after discussing the general concept of political brand-ing.
In the commercial context, the term ‘brand’ refers to a name or a symbol, or a combi-nation of them, which consumers can use as an anchor in order to identify the goods or ser-vices of a seller and distinguish them from those of competitors (Smith & French, 2009, p. 2010). Identification and distinction of brands also plays a central role in the political context, because it helps voters make political decisions. Especially in multi-party systems, brands are needed, because they enable voters to distinguish political parties from each other (Lees-Marshment, 2014). Accordingly, a political brand can be defined as “political representations that are located in a pattern, which can be identified and differentiated from other political representations” (Winther Nielsen, 2015, p. 9).
Lees-Marshment's (2014, p. 104) definition of a political brand implicates heuristics by which political brands can be identified and distinguished, namely one’s “overarching feel-ing, impression, association or image” towards, for example, a political organisation Lees-Marshment (2014, p. 104). A similar notion of a political brand as something associative and intangible is provided by Smith and French (2009, p. 212), who define it as “an associative network of interconnected political information and attitudes, held in memory and accessible when stimulated from the memory of a voter.”
The mechanism of memory-based information processing, as referred to by Smith and French (2009), is closely related to Zaller's (1992) ‘receive-accept-sample’ (RAS) model:
Voters pick up information about a political party brand, form associations based on this in-formation and store them in their long-term memory. This is comparable to filling up a buck-et; the oldest associations are close to the bottom of the bucket, whereas the most recent ciations are at the top. When they need to make a political decision, voters aggregate all asso-ciations they can recall to one heuristic (e.g., a brand), on which they can base their decision (Winther Nielsen & Larsen, 2014; Zaller, 1992). The concept of priming, which is relevant for the effect of emotional appeals on brand perceptions (see p. 7), refers to a similar causal mechanism as Zaller's (1992) RAS model.
Such associations can help voters make political decisions, because every association has a subjective value to its holder. Hence, brand value in the political context is “the value added to (or subtracted from) a political entity by associating it with a certain brand name” (Nielsen & Larsen, 2014, p. 155).2 Accordingly, attaching a brand name to a political offering can make the offering more desirable or undesirable for voters, which, in turn, affects voting decisions (French & Smith, 2010; Winther Nielsen & Larsen, 2014).
The brand value concept is closely related to the consumer-based brand equity concept (Keller, 1993), which suggests that the consumer’s brand knowledge determines how the con-sumer responds to the brand. Brand knowledge can be regarded as an associative network memory model consisting of two components, namely brand awareness (i.e., brand recogni-tion and recall) and brand image, that is, a set of associarecogni-tions related to the brand (Keller, 1993). If a consumer is aware of the brand and stores favourable, strong, and unique brand associations in her memory, this indicates positive brand equity or, in other words, a high brand value.
Compared to the brand value concept, the PBP concept focuses less on perceived va-lence of brands, but more on their perceived characteristics. Brand personality is a component of a brand’s image. It can be defined as a “set of human characteristics associated with a brand” (Aaker, 1997, p. 347). Aaker (1997) defined five brand personality traits for the com-mercial context, namely sincerity, competence, ruggedness, sophistication, and excitement. They were renamed and supplemented by Smith (2009) to honesty, spiritedness, image, lead-ership, toughness and uniqueness in order to fit them to the British political context. Accord-ing to Smith (2009), PBP perceptions can be influenced by events, party representatives, brand endorsers and most importantly by political advertising (see p. 7). Meister (2016) modi-fied Smith's (2009) conceptualisation of PBP in order to make it applicable to the German
2 Brand value is composed of three components, namely the strength (i.e., the number of named associations),
uniqueness (i.e., the diversity of association types) and favourability (i.e., the valence of associations) (see
political context. Since this thesis focuses on German voters, it draws on Meister's (2016) conceptualisation. The next section focuses on the independent variable of this study, namely emotional appeals in political advertising.
Emotional Appeals in Political Advertising
Emotions are responses of individuals to external stimuli (e.g. campaign posters) (Brader, 2005; Damasio, 2000). The nature and intensity of such responses is conditioned by the relevance that the external stimuli hold for the individual (Brader, 2005; Damasio, 2000). Following this, emotional appeals can be defined as “communications intended to elicit an emotional response from some or all who receive them” (Brader, 2005, p. 390).
Previous studies have shown that campaign advertisements can be used to elicit emo-tions among voters and thus affect their attitudes and behaviour (Brader, 2005, 2006; Chang, 2001; Yoo & MacInnis, 2005). Enthusiasm and fear are two commonly used emotions in U.S. political advertising (Brader, 2005). According to affective intelligence theory (AIT), enthusi-asm is a response to external stimuli that are beneficial or have positive implications for one’s goals (Marcus, Neuman, & MacKuen, 2000). If a given goal is perceived to be achieved, en-thusiasm emerges, which, in turn, reinforces commitment to the goal and reliance on predis-positions such as party preference (Brader, 2005; Marcus et al., 2000). If the goal is not achieved, people are disappointed and become less motivated to pursue it further (Brader, 2005). Fear is a response to external stimuli that are threatening (Öhman, 2008). It activates the surveillance system and thus motivates people to seek more information and to reconsider how to make a political decision that matches their political preferences (Brader, 2005; Gadarian & Albertson, 2014; Marcus et al., 2000).
Fear can be distinguished from anxiety. While the latter occurs in anticipation (pre-stimulus) of an imaginary or real threatening stimulus (e.g. future danger or harm) and comes with a feeling of dysphoria and physiological tension, the former occurs after exposure to an identifiable fear stimulus (post-stimulus) and is accompanied by the need to defend oneself (Öhman, 2008). Since emotional reactions to campaign posters are post-stimulus, this thesis focuses on fear instead of anxiety.
So far, very few studies have examined the role of emotions in German political ad-vertising. Those studies that focused on Germany examined television advertising instead of campaign posters and they focused primarily on the ads’ content rather than on their effects (Holtz-Bacha et al., 1994; Kaid & Holtz-Bacha, 2006; Kaid, 1996). Regarding German televi-sion ads it was found that – contrary to French and similar to U.S. ads – they contain
predom-inantly emotional rather than logical appeals3 (Holtz-Bacha et al., 1994). Hence, emotions play an important role in German political campaigning. Moreover, political posters are one of the key campaign channels in German political advertising (Geise & Brettschneider, 2010; Geise, 2014; Holtz-Bacha & Kaid, 2006). Hence, it is scientifically relevant to examine the effects of emotional appeals in German campaign posters.
Emotional Appeals in Political Advertising and their Effects on Political Party Brand Perceptions
The main argument of this section is that emotional appeals affect voters’ brand per-ceptions. One reason to assume this is that the individual voter and the political brand stand in an emotional relationship comparable to either a dear friendship or a casual acquaintance, depending on the voter’s emotional attachment to the party (Winther Nielsen, 2015). As a consequence, political party managers seek to appeal to voters’ emotions in order to establish and maintain a trusted and emotionally binding political party brand (Rutter et al., 2015; Winther Nielsen, 2015). If they succeed, the perceived party brand value increases. For ex-ample, in 2005, the British New Labour party was perceived as a party that does not listen to the public, hence it had a low brand value (Scammell, 2007). Consequently, the party was rebranded as “passionate, friendly and inclusive for the benefit of all” using inter alia emo-tionally appealing advertising (Scammell, 2007, p. 185). As a result, it managed to increase its brand value and win the general election. Thus, party brands are in part based on emotional attachments, and parties actively try to strengthen these emotional attachments through adver-tising.
Another reason to expect that emotional appeals in political advertising affect brand value perceptions is derived from the concept of priming. Priming refers to the “changes in the standards that people use to make political evaluations” (Iyengar & Kinder, 2010, p. 63). It is a memory-based model of information processing, which suggests that individuals form attitudes and make decisions based on those considerations or associations that are most easi-ly accessible to them at the moment of decision-making (Hastie & Park, 1986; Tversky & Kahneman, 1973). Priming research is mainly focused on news media suggesting that news content conveys to news audiences, which issues to use as benchmarks for political evalua-tions (Scheufele & Tewksbury, 2007). However, priming is also used in political campaigns and advertising, because political actors (e.g., parties) seek to prime those issues on which they have a comparative advantage to other actors (Iyengar & Simon, 2000; Schleuder,
3 Logical appeals can be defined as communications (e.g., advertising) using hard facts to convince the viewer
that the evidence (e.g., statistics, logical arguments, examples) suggests a certain position (Holtz-Bacha et al., 1994).
1989).4 The question is whether emotional appeals in political advertising have an influence on the accessibility of voters’ associations. If so, this would support the assumption that emo-tional appeals affect association-based party brand value and PBP perceptions.
In this regard, Nabi (2003) found that different types of emotions (i.e., fear and anger) make different types of information, that is, different causes and solutions, accessible. For example, those exposed to fear frames were likely to have societal-level causes (e.g., limited options for safe rides home as a cause for drunk driving) and protection-related solutions (e.g., breath analysers in cars) at the top of their mind, that is, accessible (Nabi, 2003).
Based on Nabi's (2003) findings, it can be assumed that different emotional appeals can prime different sets of associations, that is, make them more accessible. In other words, it might depend on the type of emotional appeal, which associations are recalled primarily in response to an emotional appeal. Brand value perceptions are based inter alia on ideology-, issue- and person-related associations(see Method) (Winther Nielsen, 2016). Ideological as-sociations relate to the “historical or symbolical background of the party, such as socialism, conservative ideology, nationalism and labour unions” (Winther Nielsen, 2016, p. 95). Issue-related associations allude to concrete policies, problems or solutions (e.g., increase of child benefit) and person-related associations relate to party leaders or other party representatives (Winther Nielsen, 2016).
Since political ideology is a predisposition (Brader, 2005; Caprara & Zimbardo, 2004), ideology-related associations can be regarded as expressions of this predisposition. According to affective intelligence theory, enthusiasm encourages reliance on predispositions (Marcus et al., 2000). Therefore, it can be assumed that those who are confronted with an en-thusiasm appeal will have ideology-related associations at the top of their mind. In contrast, fear makes individuals more attentive of their current environment and encourages them to break out of their routine and consider alternative behaviour and/or attitudes (Brader, 2005). Issues and persons can be regarded as short-term considerations related to the current envi-ronment (Brader, 2005). Hence, it can be assumed that those who are confronted with a fear appeal will have issue- and/or person-related associations at the top of their mind.
The question is whether all types of associations can be equally accessible at the same time. Based on memory research, it can be assumed that this is not the case. According to Miller (1956), people can remember around seven chunks of information. Cowan (2000)
4 This is related to salience theory, which suggests that parties compete by emphasising those issues, in which
they have a relative advantage with regard to expertise, experience, popularity and reputation (Robertson, 1976). According to this approach, not party position matters, but the extent to which parties emphasise and thus ‘make salient’ selected issues in their campaign (Libbrecht, Maddens, Swenden, & Fabre, 2009).
vised this argument suggesting a mean memory capacity for adults of only three to five chunks. A chunk is “a collection of concepts that have strong associations to one another and much weaker associations to other chunks concurrently in use” (Cowan, 2000, p. 89). A party association can be regarded as a chunk, because it is based on a collection of related pieces of information picked up from different sources such as party advertisements (Winther Nielsen, 2016). Following Miller (1956) and Cowan (2000), it can be assumed that the priming of cer-tain associations (e.g. ideology-related associations) comes at the expense of others (e.g. is-sue-related associations), because human memory has a strongly limited capacity. Hence, due to the limited memory capacity, different types of associations cannot be equally accessible at the same time.
As a result, it can be assumed that those who are confronted with an enthusiasm appeal will have ideology-related associations at the top of their mind while issue- and person-related associations shift to the bottom and thus become less accessible. On the other hand, those who are confronted with a fear appeal will have issue- or person-related rather than ideology-related associations at the top of their mind. Hence, the hypotheses are as follows:
H1: Enthusiasm appeals increase the weight of ideological associations for brand val-ue perceptions.
H2: Fear appeals increase the weight of issue- and person-related associations for brand value perceptions.
Brand value comprises not only the weight of associations, but also the self-reported valence (i.e., positive, neutral, negative) of associations. Hence, the question is whether emo-tional appeals also affect the evaluations of associations. In this regard, it can be assumed that individuals appraise a given association according to the valence of the emotional appeal that evokes the respective association (Brader, 2005). Since enthusiasm is a positive emotion im-plicating that things are going well (Brader, 2005; Colombetti, 2005), it can be expected that enthusiasm appeals evoke positive associations. In contrast, fear is a negative emotion
(Colombetti, 2005; Öhman, 2008), which encourages negative evaluations (Lerner & Keltner, 2000). Therefore, it can be expected that fear appeals evoke more negative associations; hence, the following hypotheses will be tested:
H3: Enthusiasm appeals increase the weight of positive brand associations. H4: Fear appeals increase the weight of negative brand associations.
Party brand personality (PBP). Emotional appeals are expected to not only affect
brand value, but also PBP perceptions. This expectation is based on several studies from so-cial psychology which suggest that there is a correlation between personality traits and
emo-tions (Gorbaniuk, Kusak, Kogut, & Kustos, 2015; Harker & Keltner, 2001; Izard, Libero, Putnam, & Haynes, 1993; Sweeney & Brandon, 2006; Wilson & Gullone, 1999). More pre-cisely, it is argued that fear relates to neuroticism, that is, the incapability of controlling one’s emotions (Caprara & Zimbardo, 2004; Gerber, Huber, Doherty, & Dowling, 2011; Izard et al., 1993; Wilson & Gullone, 1999). On the other hand, positive emotionality, such as enthusi-asm, relates to extraversion, that is, one’s “level of activity, vigor and assertiveness” (Caprara & Zimbardo, 2004, p. 585), and is inversely related to neuroticism (Izard et al., 1993; Wilson & Gullone, 1999). The definition of extraversion contains components that relate to the brand personalities of spiritedness (activity), leadership (assertiveness and vigor) and competitive-ness (assertivecompetitive-ness and vigor) as defined by Smith (2009) and Meister (2016). Positive emo-tional expression also relates to trait competence (i.e., being organized, mentally focused, achievement-oriented) as well as trait cheerfulness (Harker & Keltner, 2001). Competence is highly similar to Smith's (2009) and Meister's (2016) PBP of leadership and cheerfulness is one component of spiritedness (Smith, 2009).
Given this, it can be expected that fear appeals will induce voters to perceive the party as neurotic (i.e., emotionally instable). Since previous studies have shown that neuroticism and toughness are negatively correlated (Campbell-Sills, Cohan, & Stein, 2006; Horsburgh, Schermer, Veselka, & Vernon, 2009; Maddi et al., 2002), it can be expected that fear appeals will induce neuroticism perceptions, which, in turn, prompt voters to perceive the party as less tough. Moreover, fear appeals might induce voters to perceive the party as less spirited, weak in leadership and less competitive, because negative emotion is inversely related to positive traits such as extraversion, competence and cheerfulness (Izard et al., 1993; Wilson & Gullone, 1999). In contrast, enthusiasm appeals encourage voters to perceive the party as ex-traverted, competent, and cheerful (Harker & Keltner, 2001; Izard et al., 1993; Wilson & Gullone, 1999); hence, they are expected to have a positive effect on associations of spirited-ness, leadership and competitiveness. In addition, it is expected that enthusiasm appeals prompt voters to perceive the party as tough, because toughness correlates positively with extraversion (Horsburgh et al., 2009). This leads to the following hypotheses:
H5: Voters exposed to fear appeals will rank the party low on toughness, spiritedness, leadership, and competitiveness, whereas voters exposed to enthusiasm appeals will rank the party high on these traits.
However, it can be questioned that this mechanism applies to all voters in the same way, re-gardless of their party preference (Jerit & Barabas, 2012; Taber & Lodge, 2006).
Party preference. The concept of party preference entails liking and support for as
well as identification with a political party (Walgrave, Lefevere, & Tresch, 2014). Identifica-tion is an intensive form of party preference expressing a strong attachment to a given party, which can induce voters to consistently focus on what is favourable to their partisan leaning and to reject or ignore what is unfavourable (Campbell, Converse, Miller, & Stokes, 1960). This thesis focuses on party preference, as that is the broader concept, which also entails party identification.
Party preference can be affected by several dimensions such as (policy) issues, ideolo-gy (left-right) and voter personality traits (Vecchione, González Castro, & Caprara, 2011; Walczak & Van der Brug, 2012). Voters assess the degree to which they match with the party with regard to these dimensions (Lupu, 2013; Turner, 1999). Hence, the stronger the match with the party, the stronger the party preference (Lupu, 2013; Turner, 1999). Following this, party preference is conceptualised as a matter of degree, whereas party identification is rather two-sided, that is, one either identifies with a party (or several parties) or not.
The question is whether party preference moderates the effect of emotional appeals on brand value and PBP perceptions. One reason to expect this is that partisans treat political information according to their party preference (Jerit & Barabas, 2012; Lefevere, Tresch, & Walgrave, 2015; Taber & Lodge, 2006). In this regard, Lefevere et al. (2015) found that pay-ing attention to a party’s campaign is conditioned by the degree of support for the party. The greater the support, the more attention is paid to the party’s campaign (Iyengar, Hahn, Krosnick, & Walker, 2008; Lefevere et al., 2015). Related to this is the finding that partisan citizens are more likely to learn political facts that are congruent and less likely to learn facts that are incongruent with their political orientation (Chaiken & Stangor, 1987; Jerit &
Barabas, 2012; Lefevere et al., 2015). Such perceptual bias occurs due to psychological pro-cesses, which make learning political facts that have negative implications for one’s preferred party more difficult than learning facts that have positive implications (Jerit & Barabas, 2012).
While the above-mentioned studies focus on partisan attention and learning, Taber and Lodge (2006) examined partisan evaluation of political information. They found that people tend to evaluate arguments about political issues according to their prior political attitudes, even when they are repeatedly encouraged to evaluate the arguments objectively; the stronger the prior attitude, the higher this tendency (Taber & Lodge, 2006; Zaller, 1992). This is in line with the finding of Walgrave et al. (2014) that partisans tend to not accept messages from parties they dislike. These findings furnish evidence for the disconfirmation bias, that is, the
counter-arguing of information that is incongruent with prior attitudes (Jerit & Barabas, 2012; Taber & Lodge, 2006). Moreover, they support the concept of confirmation bias, in which people seek out information that confirms their prior attitudes (Jerit & Barabas, 2012; Taber & Lodge, 2006). Both types of bias can lead to more extreme attitudes after exposure to polit-ical messages, especially among those with strong prior attitudes (Taber & Lodge, 2006).
Following the partisan bias literature and considering that political brands carry politi-cal messages and attitudes (Smith & French, 2009), it can be assumed that partisan citizens update their perceptions of party brands according to their political leaning. In fact, studies on party branding argue that voters with a preference for a given party are receptive to positive (congruent) information related to the party brand, whereas negative (dissonant) information is likely to be ignored or rejected (Lupu, 2013; Smith, 2009). This is in line with the above-mentioned concepts of disconfirmation and confirmation bias.
Hence, party preference can be expected to moderate how emotional appeals are used to update brand value perceptions. More precisely, an enthusiasm appeal is expected to prime more positive associations among party supporters than among non-supporters5. Consequent-ly, brand value scores will be higher among supporters than among non-supporters after expo-sure to an enthusiasm appeal. Due to the negativity it implies, the fear appeal is expected to prime negative associations among non-supporters, because it confirms their prior attitude. These negative associations will, in turn, lead to a relatively low brand value score amongst non-supporters. In contrast, supporters will tend to ignore or reject the fear appeal, because they will regard it as dissonant information. Hence, even after exposure to the fear appeal, they will have positive associations at the top of their mind, which leads to a relatively high brand value score. In sum, it is expected that the difference in brand value scores between those exposed to enthusiasm and those exposed to fear appeals will be smaller amongst sup-porters compared to non-supsup-porters. This leads to the following hypothesis:
H6: The effect of emotional appeals on brand value perceptions is moderated by party preference: Compared to non-supporters, fear and enthusiasm appeals lead to smaller differences in brand value perceptions amongst supporters.
With regard to PBP, it is also expected that party preference moderates how emotional appeals are used to update PBP perceptions. Smith (2009, p. 218) argues: “The more partisan a voter, the more positive their perception of the personality of the party should be”. As with brand value, the reason for this is partisan bias in information processing: Supporters accept
5 Note that, throughout the thesis, the term ‘non-supporters’ refers to unlikely CDU supporters (AfD, SPD,
GRÜNE, PIRATEN) and very unlikely CDU supporters (NPD, LINKE), whereas ‘supporters’ refers to CDU supporters and likely CDU supporters (FDP, CSU).
positive and filter out negative information about their preferred party (Jerit & Barabas, 2012; Lupu, 2013; Smith, 2009; Taber & Lodge, 2006). This, in turn, leads to high brand personali-ty ratings (Smith, 2009). Hence, since an enthusiasm appeal implies positive information6, it is expected to lead to higher PBP ratings among supporters than among non-supporters. Since fear appeals convey negative information, supporters are expected to filter them out and give relatively high PBP ratings, whereas non-supporters might see their negative attitude toward the party brand confirmed and therefore they are expected to give relatively low PBP ratings. This leads to the following hypothesis:
H7: The effect of emotional appeals on party brand personality perceptions is moder-ated by party preference: Compared to non-supporters, party supporters rate the party higher on toughness, spiritedness, leadership and competitiveness after exposure to fear or enthusiasm appeals.
Superiority Effect of Pictures compared to Text
Emotional appeals in campaign posters can be conveyed through picture and text (e.g. a slogan). In this regard, previous studies have clearly established a so-called picture superior-ity effect, which indicates that pictures induce more thorough processing and are better re-membered than words (Geise & Brettschneider, 2010; Hockley, 2008; Nelson, Reed, & McEvoy, 1977; Nelson, Reed, & Walling, 1976; Paivio & Csapo, 1973; Pieters & Wedel, 2004; J. G. Snodgrass & Burns, 1978; J. G. Snodgrass, Wasser, Finkelstein, & Goldberg, 1974; J. Snodgrass, Volvovitz, & Walfish, 1972). One explanation for this is that people tend to process the meaning of pictures more extensively than the meaning of words (Nelson et al., 1977; Weldon, Roediger, & Challis, 1989). A second explanation is that pictures are more memorable than text, because they offer more distinctive visual characteristics than words (Nelson, 1979). Thus, they are more likely to be encoded uniquely in memory (Nelson, 1979). Accordingly, dual-encoding theory suggests that pictures are more likely than text to be en-coded both as verbal and as picture traces, thus increasing the likelihood of later recall (Paivio & Csapo, 1973).
Hence, a picture is more extensively processed, better recalled and therefore more ac-cessible in memory than text (Weldon et al., 1989). It can be assumed that the higher accessi-bility of pictures in memory does not only have an effect on recall, but also on attitudes, be-cause priming suggests that individuals tend to base their attitudes on information that is most easily accessible to them (Hastie & Park, 1986; Iyengar & Kinder, 2010; Scheufele &
6 Enthusiasm implies positive information, because it is a positive emotion (Brader, 2005; Colombetti, 2005). In
Tewksbury, 2007; Tversky & Kahneman, 1973). Given that pictures have a superior effect on attitudes due to the priming mechanism, it can be assumed that the effect of the emotional appeal in the picture on brand value and PBP perceptions is ‘superior’ (in terms of strength) to the effect of the emotional appeal in the slogan. This leads to the following hypotheses:
H8: The emotional appeal in the picture has a stronger effect on brand value percep-tions than the emotional appeal in the slogan.
H9: The emotional appeal in the picture has a stronger effect on party brand personali-ty perceptions than the emotional appeal in the slogan.
These two hypotheses apply to all previous hypotheses, that is, the picture superiority effect is expected to also hold for hypotheses H1 to H7.
Method Research Design and Measurement
To answer the research question, this thesis used two sources of data: an exit poll and an online survey-experiment. The online survey-experiment was conducted from April, 29th to May, 13th, 2016 using the survey software Qualtrics. The exit poll was conducted in Neu-wied, Germany, during the state elections of Rheinland-Pfalz on March 13th, 2016. The dual approach was used, because both methods complement each other in terms of strengths and limitations: The exit poll aims to be representative because participants do not self-select, but the presence of the experimenter can induce voters to give socially desirable answers (exper-imenter/ social desirability effect, see Discussion) (Keeter, 2008; Levy, 1983). On the other hand, the online survey-experiment is less representative, but experimenter and social desira-bility effects are less likely to occur (Barabas & Jerit, 2010; Berinsky, Huber, & Lenz, 2012; Duffy, Smith, Terhanian, & Bremer, 2005; Reips, 2000). The convergence of both methods contributes to internal and external validity (Gosling, Vazire, Srivastava, & John, 2004).
Online survey-experiment. The population under study consisted of
German-speaking people, who are eligible to vote at the German parliamentary elections for the Bun-destag. The survey was distributed via email and on Facebook using a snowball sampling method. Survey completion was incentivised by offering participation in a raffle for five 15€ Amazon vouchers. A total of 315 surveys were completed. 47 surveys were excluded for dif-ferent reasons: Some respondents indicated in the survey that they had been informed about the purpose of the study beforehand, or they gave invalid answers, or they did not complete the survey. This resulted in a final sample of N = 268. Due to the snowball sampling method, a response rate could not be determined.
One limitation of the snowball sampling method is self-selection bias (Berinsky et al., 2012; Reips, 2000). People who are interested in politics might be more motivated to partici-pate in the survey than others (Keeter, 2008). In fact, Duffy et al. (2005) found that online respondents tend to be more politically active than face-to-face interviewees. Hence, the snowball sampling can be a limitation to external validity (Berinsky et al., 2012; Reips, 2000). However, although they are not as representative, online samples are more diverse than tradi-tional samples, because they can be accessed by a larger amount of people (Gosling et al., 2004; Reips, 2000).
Exit poll. Regarding the exit poll, the population under study consisted of
German-speaking people who are eligible to vote at the 2016 state elections of Rheinland-Pfalz. Re-spondents were randomly sampled in front of the polling station at the town hall in Neuwied. Randomization was attained by asking every third person leaving the polling station to partic-ipate in the poll. After an interview was completed, the interviewer again waited for the third person and invited him or her to participate. A total of N = 54 voters (48.1% women; mean age M = 51.8 years) participated in the exit poll and 32 refused participation. This equals a response rate of 63%.
One limitation of the exit poll is that participants were surveyed face-to-face. This can cause social desirability and experimenter effects (Duffy et al., 2005; Keeter, 2008; Reips, 2000). Several precautions were taken to minimize the likelihood of such effects: Firstly, re-spondents were informed beforehand that their participation was voluntary, that their answers were anonymous, that they were treated confidentially and that they could opt out at any point in time without negative consequences. Secondly, the experimenter used a standardised ex-perimental procedure by strictly adhering to the script of the survey. Thirdly, the question wording of the brand value item allowed respondents to freely name any association (and as many as) they could think of. Fourthly, respondents were given the opportunity to ask addi-tional questions and to comment on the survey after participation. None of the questions and comments that came up gave reason to suspect social desirability or experimenter effects. A strength of the exit poll is that it is based on a probability sample, which is more representa-tive than the online-survey’s convenience sample (Keeter, 2008; Levy, 1983).
As great care was taken to make sure the two methods were comparable, in the re-mainder of the methods, the various aspects of the data collection – stimuli, key variables and measurements and manipulation checks – will be discussed for both methods at the same time. Where applicable, differences between the exit poll and online experiment will be dis-cussed.
Stimuli. Participants were randomly exposed to one of the following four campaign
posters of the Christian Democratic Union (CDU) (see Table 1, p. 16): A picture of a young girl expressing enthusiasm with a slogan expressing enthusiasm (PeSe), a picture of a young
girl expressing enthusiasm with a slogan expressing fear (PeSf), a picture of a young boy
ex-pressing fear with a slogan exex-pressing enthusiasm (PfSe), a picture of a young boy expressing
fear with a slogan expressing fear (PfSf). This 2x2 factorial design with randomized
distribu-tion of participants to the four different experimental condidistribu-tions allowed for experimental control and contributed to internal validity (Barabas & Jerit, 2010; Reips, 2000). Moreover, it allowed for testing the picture superiority effect (H8 and H9).
Table 1
2x2 Factorial Design
Slogan Enthusiasm
(Slogan: “With more enthusiasm towards the future. Vote for CDU.“)
Fear
(Slogan: “The future of your children is at stake. Vote for CDU.“) Pict u re Enthusiasm Fear
All campaign posters were created for the purpose of this study, taking into account the design of current real CDU campaign posters. By portraying children and addressing their future in the slogans, the campaign posters refer to the issue of family-policy. This issue was chosen, because it is emotionally loaded in as much as it is concerned with the regulation of individuals’ closest and most intimate social relationships, which have a highly emotional
quality (be it positive or negative) (Daly, 2015; McCarthy, 2012; Melton, 1987). Hence, fami-ly-policy has high potential to evoke strong emotions among the survey participants.
The CDU was chosen as the posters’ sponsor, because it is one of the two mainstream governing parties in Germany, next to the Social Democratic Party (SPD). Hence, due to its popularity and its high visibility in the media (Semetko & Schoenbach, 2003), it is likely that most German voters hold brand associations of the CDU. A second reason is the CDU’s posi-tion on the political left-right spectrum. As a mainstream centre-right party (Boswell & Hough, 2008; Marg, 2014), the CDU is less likely than extreme right- or left-leaning parties (e.g., NPD or DIE LINKE) to provoke radical associations and opinions (be it from supporters or their opponents), which are less representative of mainstream public opinion in Germany. A third reason is that the CDU traditionally dominates the issue of family-policy within the German political arena (Green, 2013; Leitner, 2007; Marg, 2014; Tjarks, 2011). As such it is unlikely that a poster featuring this issue would strike respondents as odd.
Analyses H8 and H9 were tested using the poster’s picture and the slogan as two sepa-rate independent variables (picture: 1 = fear, 2 = enthusiasm; slogan: 1 = fear, 2 = enthusi-asm). However, results do not confirm the picture superiority effect (see p. 21); hence, the other hypotheses (except for H6) were tested using poster as the key independent variable, which allows incorporating all four stimuli in a single independent variable. Poster has four categories, thus including all experimental conditions: PeSe = enthusiasm picture/ enthusiasm
slogan, PeSf = enthusiasm picture/ fear slogan, PfSe = fear picture/ enthusiasm slogan, PfSf =
fear picture/ fear slogan. H6 was tested using pure poster as the independent variable. It in-cludes only the ‘pure’ conditions PeSe and PfSf. This allowed for a clearer view on the
differ-ences in brand value perceptions between enthusiasm and fear appeals.
Key variables and measurement. The key variables of the present study are brand
value and PBP (see conceptual model in Appendix A). These variables were captured after exposing respondents to one of the four campaign posters. In addition, demographics (i.e., age, origin, education, gender), party preference and issue salience were captured before ex-posure to the stimuli. To keep the exit poll sufficiently short, the measures of PBP and issue salience were only included in the online-survey.
The survey questions and scales adopted from Meister (2016), Smith (2009) and Winther Nielsen (2016) were slightly adjusted to fit the present thesis’ research goal and the political context of Germany. Moreover, they were translated into German and pretested on ten German-speaking people, who are not experts in the field. To make sure that the questions
apply well to the German political context, they were checked and discussed with political communication scholars who are familiar with German politics and political party branding.
Brand value was measured using the following question: “When people think of polit-ical parties, they think of all kinds of things. What associations come to your mind when you think of the CDU? Please name all associations that come to your mind.” On average, exit poll respondents named fewer associations (M = 3.02) than online-survey respondents (M = 5.91) (see Discussion). Each distinct association was coded into one of the following catego-ries: 1 = Ideology: associations referring to the historical or symbolical background of a party; 2 = Issue: policy-, problem-, or solution-related associations; 3 = Person: associations about the party leader or other representatives; 4 = Affect: associations relating to emotions, values, and feelings a party can elicit; 5 = Hoopla: associations about media management, strategy decisions, or evaluations of party performances; 6 = Poster-related: that is, associations that are explicitly linked to the poster; 0 = Other: associations that do not fit into any of the above-mentioned categories (Winther Nielsen, 2016).7 A second coder (political communication student) double coded 103 associations (6.6% of all associations) of 15 randomly selected online-survey respondents. Krippendorff’s alpha (Kalpha) for the association coding was .84, which is satisfactory (Hayes & Krippendorff, 2007).
After coding, brand value (PBV) was calculated by the following three measures: 1) Strength (S): The number of associations that respondents provided, 2) Uniqueness (U): The amount of different categories covered by the associations (e.g., person, issue, ideology), and 3) Favourability (F): The self-reported valence of each association (Winther Nielsen, 2016). The latter was measured by asking respondents to rate each of their associations as positive (+1), neutral (0) or negative (-1) (Winther Nielsen, 2016). A respondent’s favourability score was calculated by adding up the values of each named association. To give an example: A respondent, who named two positive and three negative and associations received a favoura-bility score of 2 – 3 = –1 (Winther Nielsen, 2016). Finally, these three components were inte-grated in a composite measure, which expresses the brand value of a party a (PBVa) by
multi-plying strength (Sa), uniqueness (Ua) and favourability (Fa): PBVa = Sa × Ua × Fa.
Party brand personality (PBP) was measured by asking respondents to think of Ger-man parties as if they had a personality like huGer-mans and to indicate on a five-point Likert
7 Note that the share of affect and hoopla associations was relatively high in both, the online-survey and the exit
scale the extent to which 15 different PBP indicators8 as conceptualized by Smith (2009) and Meister (2016) describe the CDU (1 = “not at all descriptive” and 5 = “extremely descrip-tive”). The 15 items are indicators of the four PBPs that the present thesis focuses on, name-ly toughness, spiritedness, leadership, and competitiveness (see H5). To reduce the 15 PBP items down to their four underlying dimensions, a principal component factor analysis was conducted using oblique rotation (direct oblimin) and extraction based on eigenvalues above 1. The initial analysis indicated that the PBP item masculine correlated very weakly with all other items including tough, r(189) = .07, p = .184. Moreover, these two toughness items did not load on the same factor. In fact, masculine loaded on spiritedness and tough loaded on leadership. Hence, due to its ambiguity, the items measuring toughness were excluded from the analysis.
A second factor analysis without the toughness items provided satisfactory results. The Kaiser-Meyer-Olkin (KMO) measure verified the sampling adequacy for the analysis, KMO = .862 (Hutcheson & Sofroniou, 1999). Furthermore, all KMO values were equal to or higher than .81, which is well above the acceptable limit of .5 (Field, 2013). Eigenvalues were obtained for each factor in the data. Three factors had eigenvalues over Kaiser’s criterion of 1 and combined explained 60.5% of the variance. Due to the convergence of the scree plot and Kaiser’s criterion on that value, three factors were retained. Appendix D shows the factor loadings after oblique rotation. Factor 1 seems to represent leadership (i.e., successful, hard-working, secure, intelligent), factor 2 represents spiritedness (i.e., spirited, imaginative, up-to-date), and factor 3 represents competitiveness (i.e., risk-taking, determined, assertive). Some items (i.e., secure, up-to-date, intelligent, determined, assertive) load on more than one factor. One reason for this is that personality traits usually strongly correlate (Hendriks, Hofstee, & Raad, 1999). Furthermore, three items (i.e., confident, competitive, ambitious) do not load on the factors they were expected to load on.
Reliability tests using Cronbach’s alpha (α) were conducted to see whether the PBP subscales consistently reflect the constructs they were supposed to measure. Spiritedness (α = .76), leadership (α = .80) and competitiveness (α = .72) all had reliabilities above the mini-mum acceptable value of α = .70 (Field, 2013).
8 Smith's (2009) and Meister's (2016) brand personality models both comprise 34 trait indicators. Based on the
theoretical framework, 15 of the 34 traits are relevant for the present study. See Appendix C for a list of the 15 traits.
The moderator, party preference, was measured by asking participants what party they would vote for if elections were held today (Belknap & Campbell, 1951; Perry, 1979).9 Based on the Chapel Hill expert survey, parties’ mean position on the left-right was calculated to determine how ‘close’ other parties were to the CDU (Bakker et al., 2015). As a result, re-spondents were categorized into four groups: 1 = CDU supporter, 2 = likely supporter (FDP, CSU), 3 = unlikely supporter (AfD, SPD, GRÜNE, PIRATEN), 4 = very unlikely supporter (NPD, LINKE).
The control variable, issue salience, was measured by asking participants to choose those three issues from a list of 24 issues (including three “Other” options) that are most important to their party vote choice (1 = “most important”; 3 = “third most important”). It was coded into a binary variable (0 = other, 1 = family salience). With the exception of age, the other control variables origin, education and gender were coded into binary variables so that they could be used as covariates in the analyses.10 Manipulation checks were included at the end of the survey to ensure that the stimuli evoked the emotions they were supposed to elicit.
Manipulation Checks
After being exposed to the campaign poster, respondents of the exit poll and of the online-survey experiment were subject to a manipulation check. They were asked to report the extent to which the campaign posters made them feel enthusiastic, hopeful and fearful on a five-point Likert scale (1 = “strongly agree”; 5 = “strongly disagree”) (Brader, 2006). En-thusiasm was measured by combining the scales for enEn-thusiasm and hope (Brader, 2006). The t-tests showed that the manipulation by the pictures was successful in both methods, whereas manipulation by the slogans was not.
As for the online-survey experiment, on average respondents exposed to the enthusi-asm picture reported that they felt more enthusiastic and hopeful (M = 2.33, SE = 0.10) than respondents exposed to the fear picture (M = 1.43, SE = 0.06). This difference, -0.91, BCa 95% CI [-1.14, -0.67], was significant, t(217) = -7.60, p = .000. Moreover, on average spondents exposed to the enthusiasm picture felt less fearful (M = 2.08, SE = 0.09) than re-spondents exposed to the fear picture (M = 3.25, SE = 0.01). This difference, 1.17, BCa 95% CI [0.91, 1.44], was also significant t(262) = 8.62, p = .000.
9 The exit poll contained a slightly different party preference measure consisting of one item capturing what
party the participants have voted for on the polling day.
10 Note that analysis of covariance (ANCOVA) cannot be conducted using categorical variables as covariates.
The controls were therefore coded as follows: origin: 0 = not German, 1 = German; education: 0 = low-educated (high school degree, apprenticeship, or the like), 1 = highly educated (University degree, doctorate or the like).
In contrast, the t-test results of the slogans are insignificant. On average, online-survey respondents exposed to the enthusiasm slogan reported that they felt slightly less enthusiastic and hopeful (M = 1.82, SE = 0.08) than respondents exposed to the fear slogan (M = 1.92, SE = 0.10). However, this difference, 0.1, BCa 95% CI [-1.16, -0.36], was insignificant t(262) = 0.75, p = .453. Moreover, on average respondents exposed to the enthusiasm slogan felt slightly less fearful (M = 2.57, SE = 0.11) than respondents exposed to the fear slogan (M = 2.78, SE = 0.11). Again, this difference, 0.21, BCa 95% CI [-0.09, 0.51], was insignificant t(262) = 1.37, p = .173.
As for the exit poll, on average respondents exposed to the enthusiasm picture reported that they felt more enthusiastic and hopeful (M = 4.29, SE = 0.10) than respondents exposed to the fear picture (M = 1.58, SE = 0.10).11 This difference, -2.71, BCa 95% CI [-3.00, -2.41], was significant t (49) = -18.47, p = .000. Moreover, on average respondents exposed to the enthusiasm picture felt less fearful (M = 1.65, SE = 0.12) than respondents exposed to the fear picture (M = 4.16, SE = 0.21). This difference, 2.51, BCa 95% CI [2.03, 2.98], was significant t (49) = 10.55, p = .000.
Again, the t-test results of the slogans are insignificant. On average respondents ex-posed to the enthusiasm slogan reported that they felt more enthusiastic and hopeful (M = 3.24, SE = 0.29) than respondents exposed to the fear slogan (M = 2.69, SE = 0.29). This dif-ference, -0.55, BCa 95% CI [-1.36, 0.27], was not significant t(49) = -1,35 p = .184. Moreo-ver, on average respondents exposed to the enthusiasm slogan felt slightly less fearful (M = 2.76, SE = 0.31) than respondents exposed to the fear slogan (M = 3.0, SE = 0.29). This dif-ference, 0.24, BCa 95% CI [-0.62, 1.10], was not significant t(49) = 0.56, p = .578.
In sum, the results of the online-survey and the exit poll manipulation checks suggest that manipulation of the pictures did work as intended, but that it did not work for the slogans.
Results
This section is structured according to the hypotheses. It starts with the results of the picture superiority effect hypotheses H8 and H9, followed by the results of hypotheses H1 to H7 in the respective order.
Superiority Effect of Pictures compared to Text (H8, H9)
H8 stated that the emotional appeal in the picture has a stronger effect on brand value perceptions than the emotional appeal in the slogan. To test this hypothesis a two-way
11 Mean differences for the exit poll are higher than for the online survey. This might be due to experimenter
effects: In the presence of the experimenter, respondents are more likely to give scientifically desirable answers (Duffy et al., 2005; Keeter, 2008; Reips, 2000) (see also Discussion).
al analysis of variance (ANOVA) was used, because it allows for comparing group means of two (or more) experimental treatment groups by involving two (or more) independent varia-bles (Field, 2013). The ANOVA was conducted using the independent variavaria-bles slogan and picture (both coded 1 = fear, 2 = enthusiasm) and the dependent variable overall brand value score was conducted (independent factorial design).
Regarding the online-survey, results do not confirm H8. Firstly, the insignificant result of the Levene’s test suggested that variances in the different conditions were equal, F (3, 254) = 0.59, p = 0.621. Note that the Levene’s tests of the subsequent analyses were mostly insig-nificant; hence they will be reported hereafter only if they were significant. Regarding the effect sizes, the results indicate that the picture had a slightly stronger main effect (ω2 = .002) on brand value perceptions than the slogan (ω2
= .0008). However, these main effects were both insignificant (picture: F(1, 254) = 2.25, p = .135; slogan: F(1, 254) = 2.25, p = .135). The interaction effect of the picture and the slogan on brand value perceptions was also insig-nificant, F (1, 254) = 0.22, p = .636, ω2 = -.002, which indicates that pictures and slogans did not affect brand value perceptions differently. H8 is therefore rejected for the online-survey.
Regarding the exit poll, results are similar. They indicated that the picture had a stronger effect (ω2
= .43) on brand value perceptions than the slogan (ω2 = -.43). However, these main effects were both insignificant (picture: F(1, 50) = 1.96, p = .168; slogan: F(1, 50) = 0.01, p = .918). The interaction effect of the poster and the slogan on brand value percep-tions was also insignificant, F (1, 50) = 3.16, p = .082, ω2 = 1. Hence, just as in the online survey, in the exit poll pictures and slogans did not affect brand value perceptions differently. H8 is therefore also rejected for the exit poll.
H9 stated that the emotional appeal in the picture has a stronger effect on PBP percep-tions than the emotional appeal in the slogan. To test this assumption, firstly, the mean of the PBP items that had been retained after factor analysis were calculated per personality trait (i.e., leadership, spiritedness and competitiveness). Then a two-way MANOVA was conduct-ed, because this type of analysis allows for including more than one dependent variable in the analysis (Field, 2013). Picture and slogan were used as independent variables and the PBP traits as dependent variables. H9 was tested for the online-survey only, because PBP was not measured in the exit poll (see Method).
The results do not confirm H9, because they indicated that neither pictures nor slogans had a significant effect on the PBP traits (picture effect on leadership: F(1, 231) = 2.08, p = .151, η2
= 0.01, spiritedness: F(1, 231) = 0.44, p = .509, η2 = 0, competitiveness: F(1, 231) = 0.04, p = .849, η2 = 0; slogan effect on leadership: F(1, 231) = 0.01, p = .909, η2 = 0,
spirit-edness: F(1, 231) = 0.76, p = .383, η2 = 0, competitiveness: F(1, 231) = 1.03, p = .311, η2 = 0). Moreover, there was no significant interaction effect of picture and slogan, which indicat-ed that pictures and slogans did not affect PBP perceptions differently (interaction effect on leadership: F(1, 231) = 0.08, p = .784, η2 = 0; spiritedness: F(1, 231) = 0.58, p = .448, η2 = 0; competitiveness: F(1, 231) = 0.59, p = .443, η2 = 0). H9 is therefore rejected.
In sum, the results of H8 and H9 consistently suggest that there was no picture superi-ority effect (see also Discussion). Consequently, to assess the effects of emotional appeals on brand value and PBP perceptions, instead of using the binary variables picture and slogan, the variable poster (PeSe = enthusiasm picture/ enthusiasm slogan, PeSf = enthusiasm picture/ fear
slogan, PfSe = fear picture/ enthusiasm slogan, PfSf = fear picture/ fear slogan) will be used as
the key independent variable in all subsequent analyses, except for the analysis of H6; this hypothesis will be tested using pure poster as the independent variable (see p. 28)
The Effect of Emotional Appeals on Ideology-, Issue- and Person-Related Associations (H1, H2)
H1 stated that enthusiasm appeals increase the weight of ideological associations for brand value perceptions. This hypothesis was tested using analysis of covariance (ANCOVA), because this type of analysis allows for comparing means among more than two experimental treatment groups; hence, it is well suited for the 2x2 factorial design used in this study (Field, 2013). Moreover, an allows for including covariates (i.e., variables other than the experi-mental manipulation that might affect the outcome variable in the analysis) (Field, 2013). In the analysis, poster was used as the independent variable, the weight (i.e., share) of ideology-related associations as the dependent variable and age, origin, education, gender and issue salience as covariates. Weights were obtained for each respondent by dividing all ideology-related associations per respondent by the total number of associations per respondent.
Results of the online-survey do not confirm H1: Firstly, the effect of the posters on the weight of ideology-related associations after controlling for the covariates was insignificant, F(3, 203) = 0.34, p = .796, η2 = 0.01. Secondly, on average respondents in condition PeSe (M
= 0.44, SE = 0.03, 95% CI [0.38, 0.51]) showed a similar weight of ideology-related associa-tions as respondents in condiassocia-tions PeSf (M = 0.42, SE = 0.03, 95% CI [0.36, 0.49]), PfSe (M =
0.41, SE = 0.03, 95% CI [0.35, 0.47]), and PfSf (M = 0.45, SE = 0.03, 95% CI [0.39, 0.52]).
Hence, H1 is rejected for the online-survey, because the results are insignificant and they do not point in the expected direction.12
12 Besides the hypothesis-related results, it was found that education had a significant effect on the weight of
Regarding the exit poll, the results confirm H1: The effect of the posters on the weight of ideology-related associations after controlling for age, origin, education and gender (note that issue salience was not measured in the exit poll) was significant, F(3, 46) = 3, p = .041, η2
= 0.16. More specifically, on average respondents in condition PeSe (M = 0.68, SE = 0.11,
95% CI [0.47, 0.89]) showed a higher weight of ideology-related associations than respond-ents in conditions PeSf (M = 0.31, SE = 0.11, 95% CI [0.1, 0.52]), PfSe (M = 0.36, SE = 0.11,
95% CI [0.14, 0.59]), and PfSf (M = 0.28, SE = 0.12, 95% CI [0.05, 0.51]). This pointed in the
expected direction. Hence, overall, H1 is confirmed for the exit poll, but not for the online-survey.
H2 stated that fear appeals increase the weight of issue- and person-related associa-tions for brand value percepassocia-tions. This hypothesis was tested using ANCOVA with poster as the independent variable, the combined weight of issue- and person-related associations as the dependent variable and age, origin, education, gender and issue salience as covariates. Weights were obtained for each respondent by combining all issue- and person-related associ-ations per respondent and dividing them by the total number of associassoci-ations per respondent.
Results of the online-survey do not confirm H2: Firstly, the effect of the posters on the weight of issue- and person-related associations was insignificant, F(3, 56) = 0.3, p = .825, η2 = 0.02. On average, respondents in condition PeSe (M = 0.47, SE = 0.06, 95% CI [0.35, 0.58])
showed a similar weight of issue- and person-related associations than respondents in condi-tions PfSe (M = 0.44, SE = 0.05, 95% CI [0.35, 0.53]), and PfSf (M = 0.45, SE = 0.05, 95% CI
[0.35, 0.55]). Those in condition PeSf (M = 0.51, SE = 0.05, 95% CI [0.4, 0.61]) had a higher
weight, which suggests that a combination of an enthusiasm picture and fear slogan increases the weight of issue- and person-related associations. However, given that results are insignifi-cant, this should not be over-interpreted. Hence, in sum the results are neither significant nor do they point in the expected direction. However, it could be that the stimuli had a significant effect on either issue- or person-related associations.
Therefore, as an additional test, two ANCOVAs were conducted using issue- and per-son-related associations separately as dependent variables. Firstly, results indicate that the effect of the posters on the weight of issue-related associations was insignificant after control-ling for age, origin, education, gender and issue salience, F(3, 118) = 0.51, p = .680, η2 = 0.01. The same goes for the effect of the posters on the weight of person-related associations, F(3,
ANCOVA was conducted to test whether education moderated the impact of the stimuli on the weight of ideo-logical associations. The results indicate that among low-educated respondents the stimuli had the expected effect (i.e., enthusiasm increased the weight of ideological associations), but among highly-educated respondents effects point in an unexpected direction (i.e., fear increased the weight of ideological associations more than enthusiasm) (see Appendix E). This will be discussed in the conclusion.
107) = 0.25, p = .861, η2 = 0.01. Appendix F shows the profile plots of issue- and person-related associations based on their estimated marginal means. Though the effect was insignif-icant, the results point in the expected direction: On average, respondents in the PfSf condition
had a higherweight of person- or issue-related associations, respectively, than respondents in the PeSecondition (see Appendix F). That said the results of the mixed conditions are less
straightforward.
In sum, the fear conditions did not significantly increase the weight issue- and person-related associations after controlling for the covariates, neither when both association types were combined nor when they were analysed separately. H2 is rejected for the online-survey.
Regarding the exit poll, results do not confirm H2 either: Firstly, the effect of the post-ers on the weight of issue- and ppost-erson-related associations was insignificant, F(3, 46) = 1.59, p = .206, η2 = 0.09. Secondly, on average respondents in condition PfSf (M = 0.31, SE = 0.09,
95% CI [0.13, 0.49]) and PfSe (M = 0.28, SE = 0.09, 95% CI [0.11, 0.45]) showed a higher
weight of issue- and person-related associations than respondents in conditions PeSe (M =
0.18, SE = 0.08, 95% CI [0.01, 0.34]) and PeSf (M = 0.08, SE = 0.08, 95% CI [-0.08, 0.24]).
These results point in the expected direction, however, they were insignificant.
Just as for the online-survey, two additional ANCOVAs were conducted using issue- and person-related associations separately as dependent variables. Results do not show any significant effects (issue-related: F(3, 46) = 2.2, p = .101, η2 = 0.13; person-related: F(3, 46) = 0.64, p = .594, η2 = 0.04). Though the effects were insignificant, the results point in the ex-pected direction: Just as in the online-survey, on average, exit poll respondents in the PfSf
condition had a higherweight of person- or issue-related associations, respectively, than re-spondents in the PeSecondition (see Appendix G). That said the results of the mixed
condi-tions are less straightforward. Overall, results of the additional analysis for the exit poll point in the expected direction, but they were insignificant. Hence, H2 is rejected for both, the online-survey and the exit poll.13
The Effect of Emotional Appeals on the Favourability of Associations (H3, H4)
H3 stated that enthusiasm appeals increase the weight of positive associations, which, in turn, increases perceived overall party brand value. This hypothesis was tested using AN-COVA with poster as the independent variable, the weight of positive associations as the
13 Besides the hypothesis-related results, it was found that gender had a significant effect on the weight of
per-son-related associations, F(1, 46) = 5.39, p = .025, η2 = 0.11. For this reason, another ANCOVA was conducted to test whether gender moderated the effect of the stimuli on the weight of person-related associations. The re-sults indicate that among females the posters had the expected effect on person-related associations, but not among males (see Appendix H). However, since the Levene’s test was highly significant, F(7, 46) = 9.52, p = .000, these findings should be interpreted under reserve.
