Forthcoming IJRM Volume 32 #2 (2015)

Transcription

1 Elsevier Editorial System(tm) for International Journal of Research in Marketing Manuscript Draft Manuscript Number: IJRM-D R4 Title: The Vampire Effect: When do Celebrity Endorsers Harm Brand Recall? Article Type: Full Length Article Corresponding Author: Prof. Henrik Sattler, Corresponding Author's Institution: First Author: Carsten Erfgen, PhD Order of Authors: Carsten Erfgen, PhD; Sebastian Zenker, Assistant Professor; Henrik Sattler Abstract: Although many brand managers favor the use of celebrities in advertisements, others worry that celebrities overshadow the brand and thus impair brand recall. Practitioners refer to this overshadowing as the vampire effect, defined as a decrease in brand recall for an advertising stimulus that features a celebrity endorser versus the same stimulus with an unknown but equally attractive endorser. Because there is no agreement about whether this overshadowing really exists, this research analyzes for the first time the existence of the vampire effect and its moderators in a series of experiments with a total of 4,970 respondents. The results provide important insights into how to avoid the vampire effect by creating appropriate conditions, such as high endorser-brand congruence or a strong cognitive link between the celebrity and the brand. Surprisingly, brand familiarity does not significantly moderate the effect.

2 Title Page _final The Vampire Effect: When Do Celebrity Endorsers Harm Brand Recall? Carsten Erfgen a, Sebastian Zenker b, and Henrik Sattler c December 2014 a Carsten Erfgen is Assistant Professor at the University of Hamburg, Institute of Marketing and Media, Welckerstraße 8, Hamburg, Germany, Phone: , Fax: +49 (40) ( carsten.erfgen@googl .com). b Sebastian Zenker is Assistant Professor at the Copenhagen Business School, Department of Marketing, Solbjerg Plads 3, 2000 Frederiksberg, Denmark, Phone: , Fax: ( zenker@placebrand.eu). c Henrik Sattler (corresponding author) is Professor of Marketing and Branding, University of Hamburg, Institute of Marketing and Media, Welckerstraße 8, Hamburg, Germany, Phone: , Fax: ( henriksattler@googl .com). ========================================================== ARTICLE INFO Article history: First received in July 28, 2012 and was under review for 9½ months. Area Editor: Aric Rindfleisch ============================================================ Acknowledgements The authors thank Suzanne C. Beckmann and Karen Gedenk for their helpful comments on previous versions of this manuscript. They thank Tina Müller (Henkel KGaA), the Academic Centre for Brand Management and Marketing (ZMM), as well as research assistants Jill Folger, Julia Novikova, and Caroline Schulze for their support in collecting data and other tasks.

3 *Manuscript Click here to view linked References 1 The Vampire Effect: When do Celebrity Endorsers Harm Brand Recall? ABSTRACT Although many brand managers favor the use of celebrities in advertisements, others worry that celebrities overshadow the brand and thus impair brand recall. Practitioners refer to this overshadowing as the vampire effect, defined as a decrease in brand recall for an advertising stimulus that features a celebrity endorser versus the same stimulus with an unknown but equally attractive endorser. Because there is no agreement about whether this overshadowing really exists, this research analyzes the existence of the vampire effect and its moderators in a series of experiments with a total of 4,970 respondents. The results provide important insights into how to avoid the vampire effect by creating appropriate conditions, such as high endorser brand congruence or a strong cognitive link between the celebrity and the brand. Surprisingly, brand familiarity does not significantly moderate the effect. Keywords: celebrity endorsement; brand recall; brand management; advertising effectiveness

4 2 1 Introduction Well-known celebrities appear frequently in marketing communications to facilitate brand and ad differentiation in cluttered media environments (Atkin & Block, 1983; Erdogan & Baker, 2000; McCracken, 1989; Patel, 2009). A celebrity is a publicly known individual who uses her or his recognition in favor of a product by being depicted with it in an advertisement (McCracken, 1989). These days, this also encompasses individuals who gained their recognition for example from social media sources or reality TV shows. Approximately 20 to 25 percent of advertisements in general feature some famous person as an endorser (Sliburyte, 2009). Especially in the past decade, the use of celebrity endorsers has strongly increased (White, Goddard, & Wilbur, 2009); combined worldwide spending on corporate sponsorships and endorsements reportedly amount to $50 billion (Crutchfield, 2010). Such enormous investments benefit the company only if they translate into favorable brand recall, compared with a similar advertisement that features an unknown endorser. Many brand managers appear convinced of the effectiveness of celebrity advertisements, yet others voice concerns about the appropriate returns on investments of celebrities (e.g., Agrawal & Kamakura, 1995; Erdogan & Baker, 2000; Erdogan, Baker, & Tagg, 2001). At least in some conditions, celebrities even might overshadow the advertised brand, such that consumers remember only the celebrity, not the brand (Belch & Belch, 2008; Rossiter & Percy, 1987). Practitioners refer to this overshadowing as the vampire effect 1 of celebrity endorsers; as Evans (1988, p. 35) notes, in predicting that such overshadowing is particularly prominent in the absence of any apparent connection between the celebrity and the advertised brand, the use of celebrities, if they don t have a distinct and specific 1 This term stems from a practitioner-oriented perspective. In the literature, other terms like overshadowing effect are sometimes used interchangeably.

5 3 relationship to the product they are advertising, tends to produce the vampire effect : they suck the lifeblood of the product dry; the audience remembers the celebrity but not the product. However, there is no agreement about whether this overshadowing effect exists and what factors might moderate it. To date, no scholarly, empirical research has investigated either the existence or the moderators of the vampire effect. Accordingly, we begin by defining the vampire effect, as the decrease in brand recall in an advertisement for an advertising stimulus with a celebrity endorser compared with the brand recall prompted by the same advertising stimulus with an unknown but equally attractive endorser. Brand recall achieved through advertising is a critical prerequisite for marketing success, because consumers brand recall constitutes a necessary precondition of brand equity (Keller, 1993). For example, greater brand recall increases the likelihood that the brand becomes part of a consumer s consideration set (Nedungadi, 1990); in low involvement conditions, it also serves as a heuristic for purchase decisions (Jacoby, Syzabillo, & Busato- Schach, 1977; Park & Lessig, 1981; Petty & Cacioppo, 1986). Recall leads to brand image generation, that is, the emergence of a set of associations with the brand that require a cognitive node in memory to which they can attach, such as the brand s name (Keller, 1993). Considering how many firms trust and invest in celebrity advertising, it is pertinent to analyze whether overshadowing effects are relevant in this context, as well as whether any conditions might help managers to avoid such a vampire effect. To date, we know virtually nothing of context-specific potential moderators of the vampire effect, such as congruence between the endorser and the brand. From a managerial point of view, it is of particular importance to know, whether the vampire effect is a real threat, and if it is, how the effect can be avoided. This research contributes to the literature in that it tests for the first time the existence of the vampire effect

6 4 and its moderators. In a series of experiments with a total of 4,970 respondents we confirm the existence of the vampire effect across different brands, celebrities, study designs, and methodologies. In addition, we show ways on how managers can decrease or eliminate the vampire effect. A potentially powerful option to avoid the effect is the selection of the endorser, which determines the brand-endorser relationship. Accordingly, we focus on two dimensions of the brand-endorser relationship, namely the celebrity-brand congruence and the cognitive link between the celebrity and the brand. From a theoretical and managerial perspective brand familiarity seems also to be a potential moderator of the vampire effect, since in celebrity advertisement two stimuli compete for consumers attention and greater familiarity should shift processing of advertising elements from less to more familiar stimuli (Craik & Lockhart, 1972). Our empirical results show that high endorser brand congruence as well as a strong cognitive link between the celebrity and the brand can indeed avoid the vampire effect. In contrast with our expectations, though, brand familiarity has no significant effect on the occurrence of the vampire effect. 2 Conceptual Background and Hypotheses A few studies in the celebrity endorsement domain explicitly focus on brand recall effects and thus have laid the basis for our investigation of the vampire effect. There is one stream of research that compares the effectiveness of celebrity endorses to non-celebrity advertising. Friedman and Friedman (1979) and Petty, Cacioppo, and Schumann (1983) compare the effectiveness of different types of endorsers (celebrities, experts, typical customers) and show that celebrities improve brand recall compared with groups of typical, average-looking customers or experts. While these studies provide important insights, they do not focus on

7 5 analyzing the attractiveness of the endorser. Because attractiveness has a strong positive impact on brand recall (Kahle & Homer, 1985), it is not clear whether the higher recall values for celebrities were due to their celebrity status or their higher attractiveness. Mehta (1994) finds evidence that consumers who view a commercial featuring a celebrity mention more endorser-related responses in a thought-listing procedure than consumers viewing a commercial with an unknown model. In addition, the use of celebrity and non-celebrity endorsers does not produce significant effects on brand attitude, commercial attitude, or buying intention (Mehta, 1994). Tanner and Maeng (2012) compare familiar and unfamiliar faces of endorsers, discovering that facial familiarity increases perceptions of trustworthiness. Using morphing software that produces combinations of famous and unfamiliar faces, they find that implicit recognition is sufficient to prompt consumers automatic valence judgments of a celebrity. Another stream of research provides important results with regard to celebrity advertising without focusing on a comparison to non-celebrities. Misra and Beatty (1990) find evidence that brand recall increases with greater fit between the celebrity and the brand, which suggests that the brand-endorser relationship could be an important moderator of the vampire effect. This result is an important indication for further investigating the role of fit as a potential moderator. Constanzo and Goodnight (2005) asked students to recall brand names by cuing them with different celebrity faces. They found that a celebrity recognized in a magazine advertisement did not increase consumer recall of the brand endorsed by the celebrity for both professional athlete celebrities and other entertainment celebrities who are not professional athletes. Overall, advertising research to date has laid an important basis, but not yet addressed the vampire effect that celebrity endorsers compared to unknown advertising models might hurt rather than help brand memory. Specifically, prior research has not yet examined

8 6 whether celebrity endorsers hurt brand memory and which factors might moderate this negative relationship, if it exists. The present research addresses these questions. 2.1 Existence of the Vampire Effect The overshadowing of brand-related information by a celebrity endorser (i.e., vampire effect) can be explained by the accessibility-diagnosticity framework (Feldman & Lynch, 1988). It suggests that whether consumers use a piece of information in their opinion making depends on (a) the accessibility and (b) the diagnosticity (i.e., usefulness) of this piece of information compared to the accessibility and diagnosticity of other pieces of information that are also available. Because a celebrity endorser is more familiar than an unknown one, celebrities should be more accessible and more diagnostic in evaluating an advertisement and, therefore, should overshadow the recall of a brand more than an unknown endorser. For instance, a well-known celebrity might activate certain associations in consumers memories, whereas an unknown endorser cannot, because he or she invokes no cognitive schema (Bruce & Young, 1986; Misra & Beatty, 1990). Likewise, perceptual processing theory argues that stimulus recall relates positively to the extent of cognitive processing and negatively to the processing of competing stimuli (Massaro, 1970). In line with this theoretical reasoning, prior empirical research has shown that familiar stimuli tend to be more accessible and get processed faster and more extensively than less accessible stimuli (Craik & Lockhart, 1972; Feldman & Lynch, 1988). For instance, in a visual search task, Devue, Van der Stigchel, Bredart, and Theeuwes (2009) find that respondents look at highly familiar persons for a longer time than they do at less familiar persons. Thus, assuming consistent endorser attractiveness (see the introduction of chapter 2 above), we predict that when an advertisement features a well-known celebrity, other stimuli receive less cognitive processing than they would in an advertisement with an unknown endorser. That is,

9 7 H 1 : Recall of the brand is lower when the advertisement contains a celebrity endorser than when it contains an equally attractive but unknown endorser. In addition to demonstrating a vampire effect, we aim to provide insights into managerially relevant moderators, from both brand and brand-endorser relationship perspectives. Thus, we focus on brand-related characteristics (i.e., brand familiarity) and on the link between the brand and the endorser (i.e., congruence and cognitive link) and provide advice on how to design a celebrity campaign. With congruence and cognitive link, we analyze two distinct facets of the connection between the brand and the endorser. High congruence can be achieved through the selection of a fitting celebrity, whereas a strong cognitive link can exist even for partners with low congruence (e.g., long-running campaigns with the same endorser). 2.2 Brand Related Moderators of the Vampire Effect The first moderating effect relates to brand familiarity, obtained directly or indirectly through brand experiences (Campbell, Keller, Mick, & Hoyer, 2003), which influences attention to advertising (Hanssens and Weitz 1980). Following again the accessibilitydiagnosticity framework (Feldman & Lynch, 1988), a more familiar brand should be more accessible and more diagnostic in evaluating an advertisement. Thus, when a celebrity endorses a more familiar rather than a less familiar brand the former should suffer less from the vampire effect, because the viewer continues to process the familiar brand because of its accessibility and diagnosticity. Accordingly, Craik & Lockhart (1972) demonstrated empirically that a more familiar brand tends to enjoy more attention than a less familiar brand, and greater familiarity shifts processing of advertising elements from less to more familiar stimuli (Craik & Lockhart, 1972). Similar, Massaro (1970) found that more extensive processing leads to greater recall and Kent and Allen (1994) reveal that competitive

10 8 advertising elements only dilute the recall of the brand claim for less familiar brands. Combining theoretical thoughts and empirical evidences we conclude, that when a celebrity appears as a competitive advertising element, greater brand familiarity might weaken the vampire effect: H 2 : The negative effect of a celebrity endorser on brand recall is greater (lesser) in conditions of low (high) brand familiarity. 2.3 Brand-Endorser Relationship Related Moderators We argue that the vampire effect may increase when consumers do not perceive an overt connection between the celebrity and the brand. In general, high celebrity brand congruence should have a favorable effect for the advertised brand (Kamins, 1990; Kirmani & Shiv, 1998). Misra and Beatty (1990) find that highly congruent celebrity endorsers generally increase brand recall more than less congruent ones. Similarly, Meyers-Levy (1991) shows that recall of ad claims increases when congruent, as opposed to less congruent, stimuli are presented. Congruent information are easier to jointly remember (and would not compete against each other) and therefore, higher congruence should strengthen the tie between the celebrity and the brand and weaken the vampire effect. Thus, H 3a : The negative effect of a celebrity endorser on brand recall is greater (lesser) in conditions of low (high) perceived congruence between the endorser and the brand. According to associative learning theory (Till, 1998), brands and celebrities represent nodes in memory, linked to other nodes through previous experiences (Till & Shimp, 1998). Although a brand and celebrity initially may be unconnected, they can be linked over time through the endorsement process; consumers might learn that the brand and celebrity belong together (Till & Shimp, 1998), as in the cases of Nespresso and George Clooney in Europe or Haines underwear and Michael Jordan in the United States. If a tight connection gets

11 9 established, any activation of the celebrity automatically activates the brand-related information. Again this would lead to a lower competition between both stimuli in the advertisement. Thus, we expect the vampire effect to be less relevant when the cognitive link between the celebrity and the brand is strong, such that the advertised brand is a top-of-mind association with the celebrity. The repeated co-presentation of both stimuli in an advertising campaign should increase the probability of an associative learning process and enhance beliefs that the presence of one stimulus (the celebrity) predicts the presence of the other (the brand) (Till, 1998). The stronger the cognitive link, the higher the probability that the activation of one node prompts the activation of the other (Hoyer & MacInnis, 2007), such that the celebrity comes to represent a mental cue for the brand (McCracken & Macklin, 1998). Thus, we hypothesize: H 3b : The negative effect of a celebrity endorser on brand recall is lesser (greater) when the cognitive link between the endorser and the brand is strong (weak). 2 3 Empirical Studies To test our hypotheses, we conducted four empirical studies. In Study 1, we tested for the vampire effect (H 1 ) in a recall experiment, using a large-scale sample that explicitly included the brand s relevant target group and real-world, professionally designed advertising stimuli. In Study 2, we retested the vampire effect using different stimuli and another sample. The goal of Study 3 was to validate our recall findings and to assess the possible moderating impact of brand familiarity (H 2 ). Finally, in Study 4, we used a wide variety of different 2 Note that H3a and H3b address related, but different concepts: While we understand congruence as general perceived fit between both stimuli, the cognitive link comprises a stronger connection, for example as top-ofmind association of stimuli for the other. Therefore, a cognitive link would need more effort to get established (learning), while congruence could be generated by, for instance, finding the right celebrity for the brand. Thus, even though both constructs use the same reasoning, they are also distinct.

12 10 celebrities and brands to generalize our previous findings. In addition to brand familiarity, we tested congruence between the celebrity and the brand (H 3a ) and the impact of the strength of the cognitive link between the celebrity and the brand (H 3b ) on the occurrence of the vampire effect. 3.1 Study Method. For our first study, we tested the existence of the vampire effect (H 1 ). Specifically, we used a professionally designed advertisement showing Cindy Crawford as the celebrity endorser; the advertisement was under consideration for use in a German nationwide advertising campaign (and later was initiated, with a budget of several million Euros). This study also involved the brand s actual target group, namely, a sample of 992 German consumers, and additional dependent variables (i.e., recall of an umbrella brand and a product brand). The study was conducted in cooperation with a multinational company (Schwarzkopf). To find the equally attractive, unknown endorser, we measured attractiveness using five 7-point Likert items adopted from Ohanian (1991): attractive, classy, beautiful, elegant, and sexy (α =.94; 1 = completely disagree, 7 = completely agree ), and we observed no significant differences (M Crawford = 5.16, SD Crawford = 1.54; M unknown = 5.09, SD unknown = 1.27; t(960) =.80; p =.43). We list all the items used in Appendix A. In cooperation with the company, we used two undisclosed, authentic, professionally designed hair coloration product advertisements with a two-group design (celebrity versus unknown endorser), as we illustrate in Appendix B. The advertisement contained an umbrella brand name (Schwarzkopf) and the product brand name (Essential Color). To ensure the validity of the target group, we solicited a representative online sample of female participants who used hair coloration products frequently (self-assessed usage frequency) through a professional online panel provider (respondi AG) and followed existing

13 11 recommendations for online sample acquisitions (Birnbaum, 2004). Ultimately, 992 participants completed the online survey (M age = 38.6 years, SD age = 12.0; 23.3% with a bachelor s degree or higher). We assigned participants randomly to one of the two experimental conditions (celebrity versus unknown endorser). In this study and the following ones, the cover story informed participants that they were participating in a test of new advertisements. The advertising stimulus was available for six seconds, somewhat longer than average viewing times for traditional print advertising (Wedel & Pieters, 2000). However, we did not explicitly instruct respondents to look at the ad for the whole six seconds. Participants next solved four mental math exercises involving basic arithmetic operations, to distract them from the advertisement, before responding to our dependent recall items: unaided recall of brand name, and then aided recall of the brand name (i.e., participants could choose among 10 answer options). Finally, we asked about brand familiarity, as well as some general demographic items. As our dependent variables, we measured unaided brand recall (both umbrella and product brand names were possible correct answers, assessed independently) and aided recall of the umbrella and product brand names, for which participants chose among 10 known brand names Results. The recall rates confirmed, as we predicted in H 1, that the use of a celebrity impairs brand recall, compared with the use of an equally attractive, unknown endorser. Similarly, unaided brand name recall was significantly lower for the celebrity group (21.24% correct) than for the unknown endorser group (26.37% correct; χ 2 (1, N = 992) = 3.60, p =.03, φ =.06), as was aided umbrella brand name recall (34.67% vs %; χ 2 (1, N = 992) = 8.70, p =.00, φ =.09) and aided product brand name recall (36.87% vs %; χ 2 (1, N = 992) = 2.93, p =.05, φ =.05). We thus found support for H 1. Thus, the results of Study 1 offered confidence about the existence of the vampire effect, using a realistic advertisement context.

14 Study Method. For our second study, we developed a two-group between-subjects design (celebrity versus unknown endorser) and employed two different brands from different categories to retest the occurrence of a vampire effect (H 1 ). In this study, we chose Heidi Klum as celebrity endorser because of her status as one of the most famous German women (Forbes, 2009). For the unknown model, we chose a professional-looking model. As in Study 1, participants rated the attractiveness of the endorser (M Klum = 4.69, SD Klum = 1.68; M unknown = 4.89, SD unknown = 1.44; t(581) = 1.57; p =.12). We selected Calvin Klein (fashion) and blend-a-med (toothpaste). The rest of the method for this study was similar to that for Study 1. We recruited 617 German participants through a professional online panel provider (respondi AG) and assigned them randomly to one of the two brands (i.e., either the Calvin Klein or blend-a-med condition). After the recall questions (unaided recall of brand name, and then aided recall of the brand name), we asked for the participant s unaided knowledge for remembering the celebrity (or unknown model), as well as some general demographic items. To explore the vampire effect in more depth, we tested this time whether our participants recognized the celebrity (or in the case of the unknown model, if they thought the model was a celebrity) using unaided recall of the endorser s name. We excluded participants who could not name the celebrity or thought the unknown model was a celebrity, resulting in a final sample of 526 participants (45.1% women, M age = 45.4 years, SD age = 13.6; 30.1% with a bachelor s degree or higher) Results. We tested our hypotheses using a series of hierarchical logistic regressions (dummy variable: 0 = unknown endorser; 1 = celebrity), with attractiveness as a control variable for unaided and aided brand name recall. As predicted in H 1, the results

15 13 indicated significantly impaired recall in the celebrity conditions, in further support of the vampire effect (unaided recall: B = -.50, p =.01; aided recall: B = -.42, p =.03; see Table 1). In addition, we did not find any significant differences across both brands (i.e., Calvin Klein or blend-a-med), so that both brands are equally affected by the vampire effect. ---Insert Table 1 about here--- We also controlled for a gender-specific occurrence of the vampire effect by introducing a control variable that measured the gender congruence between the participant and the endorser (0 = incongruence; 1 = congruence), but we did not find any such effect (Model 2; unaided recall: B =.30, p =.41; aided recall: B =.51, p =.19). Thus, Study 2 further confirms the existence of the vampire effect under various conditions. 3.3 Study Method. In this study, we developed a 2 (celebrity versus unknown endorser) 2 (more versus less familiar brand) full-factorial between-subjects design to assess the occurrence of a vampire effect (H 1 ) and the moderating impact of brand familiarity (H 2 ). We chose Stefan Raab as a celebrity endorser because of his status as one of the most famous television presenters in Germany. For the unknown endorser condition, we chose an unknown, professional-looking model (M Raab = 2.55, SD Raab = 1.57; M unknown = 3.72, SD unknown = 1.57; t(609) = 9.22; p =.00; note that we controlled within the subsequent analyses for attractiveness by including an attractiveness control variable). Participants rated the attractiveness of the endorser on five 7-point Likert items like in the previous studies. For the product category, we chose sunglasses, which have inherent social and psychological risks and thus are especially suitable for celebrity endorsements (Friedman & Friedman, 1979). We selected Calvin Klein and Kenneth Cole as more and less familiar fashion brands. The brand familiarity measure included three 7-point Likert items adopted from Kent and

16 14 Allen (1994): familiar, experienced, and knowledgeable (α =.97, same scale endpoints). All other elements of the advertising stimuli (e.g., claim, slogan) remained constant across the four advertisements. We recruited 611 German participants through a professional online panel provider (respondi AG) and assigned participants randomly to the four experimental conditions and used the same procedure like in the studies before. We likewise excluded participants who could not name the celebrity or thought the unknown model was a celebrity (82 participants). Thus, we had a final sample of 529 participants (46.3% women, M age = 45.1 years, SD age = 13.2; 21.6% with a bachelor s degree or higher) Results. We tested our hypotheses using a series of hierarchical logistic regressions (see Table 2). The metric variables were mean-centered before we entered them into the regression models. We first regressed an indicator-coded dummy variable (0 = unknown endorser; 1 = celebrity) for the endorser and attractiveness as a control variable on unaided and aided brand name recall. The results showed significantly impaired recall in the celebrity conditions, in support of the vampire effect (H 1 ; unaided recall: B = -.50, p =.01; aided recall: B = -.48, p =.02; see Table 2). ---Insert Table 2 about here--- To test our expectation that the vampire effect would be stronger for less familiar brands (H 2 ), we compared the effects for Calvin Klein versus Kenneth Cole, after a manipulation check confirmed the intended brand familiarity manipulation. That is, Calvin Klein was perceived as significantly more familiar (M = 4.84, SD = 1.29) than Kenneth Cole (M = 1.53, SD = 1.20, t(527) = , p <.001). In Model 2, we added brand familiarity and its interaction with endorsement. In contrast with our expectations, the product term was not significant for either unaided or aided brand name recall (unaided recall: B =.28, p =.52; aided recall: B =.05, p =.90). Thus we must reject H 2 and assert that both familiar and less

17 15 familiar brands suffer from the vampire effect. We also controlled for a gender-specific occurrence of the vampire effect by introducing a control variable that measured the gender congruence between the participant and the endorser (0 = incongruence; 1 = congruence), but we did not find any such effect (unaided recall: B =.11, p =.78; aided recall: B = -.02, p =.95). Overall, in Study 3 we find further support for the existence of the vampire effect, which occurs for familiar as well as for less familiar brands. 3.4 Study Method. In this study, we aimed to address the generalizability of our findings and test for the existence of the vampire effect (H 1 ) across 16 endorsers and eight brands, using a large-scale empirical study. In addition, we retested H 2 (brand familiarity) and introduced tests for the moderating impact of congruence (H 3a ) and cognitive link (H 3b ). Through several pretests, we ensured substantial variance in brand familiarity, the levels of congruence between the endorser and the brand, and the strength of the cognitive link between the celebrity and the brand. In a first pretest, a convenience sample of 72 participants rated 33 celebrities in terms of their familiarity ( I am familiar with [Celebrity Y] ) on seven-point Likert scales. Additionally, we asked them to list all brands they could recall as currently endorsed by the 33 celebrities. From these results, we selected 19 celebrities who were well-known (4.35 M 6.28; 1.15 SD 2.30). In a second pretest, another convenience sample of 63 respondents reviewed brand celebrity combinations that represented either a strong or a weak cognitive link. To implement the strong cognitive links, we paired 10 celebrities with their most frequently mentioned brands from the first pretest. For the weak cognitive link, we created fictitious pairings of 9 celebrities and brands; the fictitious pairings ensured that our

18 16 respondents did not perceive the celebrities as having endorsed the brands previously. Next, participants rated the cognitive links in all 19 pairings, using three items ( When I think of [Celebrity Y], I remember he/she endorses [Brand X], [Celebrity Y] has been endorsing [Brand X] for a long time, and When I think of [Brand X], one of my first associations is [Celebrity Y] ). We averaged the responses to the three items (α =.87) and selected the 8 celebrity brand pairings in Table 3 that differed significantly on the strength of their cognitive links. ---Insert Table 3 about here--- In addition, we created sixteen advertisements: eight featuring the celebrities shown in Table 4 and eight similar ads featuring unknown endorsers. To ensure equal attractiveness of the celebrities and the unknown endorsers, we conducted a third pretest with a convenience sample of 51 participants and asked them to rate the attractiveness of all endorsers on the five-item scales we used previously (α =.94; Ohanian, 1991). In none of the cases attractiveness was significantly different. For the main study, we recruited 2,923 German participants through the same professional online panel provider (respondi AG; 51.9% female, M age = 43.7 years, SD age = 13.7, 28.9% with a bachelor s degree or higher) and assigned them randomly to the sixteen groups (i.e., eight celebrity brand pairings in Table 3 and eight unknown endorsers with the same brands from Table 3). The advertising stimulus appeared for six seconds. Participants solved four mental math exercises to distract them from the advertisement, and then provided unaided and aided recalls of brand-related information. Subsequently, we measured attractiveness (α =.95) with the same scales from the pretests. Finally, we assessed congruence between the endorser and the brand with the scale from our previous studies (α =.93).

19 Results. The series of hierarchical logistic regressions showed significantly impaired recall in the celebrity conditions, in further support of the vampire effect (H 1 ; see Table 4). ---Insert Table 4 about here--- In Model 2, we added the cognitive link between the celebrity and the brand (using a dummy-coded approach, 0=weak cognitive link and 1=strong cognitive link) and found a moderating relationship with endorsement (B unaided =.59, p <.001; B aided =.69, p <.001), in support of H 3b. As predicted, the negative celebrity effect on recall decreased with a stronger cognitive link (Table 4). In Model 3, we also included congruence and its moderation, which was significant for unaided and aided brand name recall (B unaided =.23, p <.001; B aided =.20, p <.1), in support of H 3a and our prior results. Finally, in Model 4 we added brand familiarity and its interaction with endorsement. In contrast with our expectations, the product term was not significant for either unaided or aided brand name recall (B unaided =.06, n.s.; B aided =.11, n.s.). Therefore, we must reject H 2 in all our studies. Brand familiarity did not serve as an influential moderator, that is, the vampire effect occurs for familiar as well as for less familiar brands. 4 Discussion Brand managers often regard celebrity endorsement as a particularly effective advertising strategy, which has led to the widespread use of celebrities. However, besides all proven positive effects of celebrity endorsement, celebrities might also overshadow the brand and thus impair brand recall the so-called vampire effect. With this article, we analyze its existence and its moderators empirically for the first time.

20 18 In Table 5 we summarize the results across all of our studies. The data clearly indicate the existence of a vampire effect in various conditions: across different celebrities (i.e., varying gender and attractiveness), brands (i.e., product categories, brand familiarity), and samples. Our empirical results further show that high endorser brand congruence as well as a strong cognitive link between the celebrity and the brand can avoid the vampire effect. ---Insert Table 5 about here--- Our findings have a number of important implications. First, our results clearly indicate that the vampire effect is a real threat: Managers have good reason to be concerned about the overshadowing effect of celebrities in brand advertisements. The vampire effect is a phenomenon that exists for various kinds of celebrities, brands and consumers. In particular, it is striking that it occurs for familiar as well as for less familiar brands. However, even though celebrity-endorsement ads may not contribute to enhancing the value of a familiar ad, it is conceivable that the familiar brand would still enter into a consumer's consideration set when the consumer decides which product to purchase. Therefore, it may be important to investigate the effects of the vampire effect on actual purchase and brand attitudes in future research. More generally, beyond the vampire effect, other positive outcomes of celebrity endorsers might justify their use (Mehta, 1994; Ohanian, 1991), such as positive emotional conditioning for the brand (Till, Stanley, & Priluck, 2008) or identification and aspiration effects, which further emphasis the need to measure the vampire effects not only on brand recall, but also on brand attitude, purchase or related constructs. These other positive outcomes of celebrity endorsers might also justify their use over equally attractive noncelebrities, who are presumably less expensive and do not suffer from the vampire effect. Nevertheless, the vampire effect has general relevance for the effectiveness of celebrity campaigns, because recall constitutes a critical prerequisite of marketing communication success and is a necessary precondition for brand equity (Keller, 1993).

21 19 Second, our study shows ways on how managers can decrease or eliminate the vampire effect and thus can further improve the effectiveness of celebrity endorsement ads, in addition to other positive outcomes of celebrity endorsers. According to our results, a powerful tool to avoid the effect is the selection of the endorser, which determines the brand-endorser relationship. We focus on two dimensions of the brand-endorser relationship, namely the celebrity-brand congruence and the cognitive link between the celebrity and the brand. Our results indicate that the vampire effect does not occur when there is (1) high congruence between the endorser and the brand or (2) a strong cognitive link between the celebrity and the brand. However, these conditions are not easy to achieve. A strong cognitive link between the celebrity and the brand, for example, likely requires extensive exposures to celebrity advertisements. It seems that extensive marketing efforts are necessary to avoid the vampire effect, such as long-term celebrity campaigns with many and consistent exposures. Our findings have also implications for designing advertising campaigns that can avoid an overshadowing effect of celebrities. Designers need to consider the brand logo as a competitive element in an advertisement. It should be placed in a prominent perceptual area, in accordance with the mandate that the product, not the personality, has got to be the star (Cooper, 1984, p. 120). Beyond the tools to avoid the vampire effect tested in this study, there are other potential moderators that should be analyzed by future research, such as the number of exposures to a celebrity advertisement (to test whether multiple exposures can build a sufficiently strong link between the brand and the celebrity to avoid the vampire effect) or consumers involvement with the product category (when a consumer is highly involved with the product category, advertising stimuli receive more detailed processing and should therefore avoid the vampire effect).

22 20 Third, and more generally, the vampire effect can be seen as a special case of an overshadow effect. Such overshadowing effects have been investigated in domains such as sex or humor, that is, whether advertising containing humor or sex might impair recall of an advertised brand. Several studies observe a negative effect (Alexander & Judd, 1978; Soley & Kurzbard, 1986; Sutherland & Middleton, 1983), whereas others fail to confirm this effect, find no effects of humorous compared with non-humorous messages (Eisend, 2009), or even reveal positive effects (Stewart & Furse, 2000). Thus, the findings of overshadowing effects in the domain of sex and humor are quite mixed, whereas we could consistently demonstrate across four studies the existence of overshadowing effects for celebrities. One explanation could be that advertising tools such as sex or humor are very different from celebrity effects. In particular, sex and humor represent general emotion-related appeals, whereas celebrities are personalities whose specific characteristics might fit more or less well with the advertised brand s personality. Because of their socio-biological nature, human faces represent particularly meaningful stimuli (Ro, Russell, & Lavie, 2001). Celebrities trigger more than one single emotion and thus offer more comprehensive information than single emotions such as humor and sex appeals. Thus, overshadowing effects seem to be highly dependent on the marketing domain, which emphasizes the need for a separate analysis within the celebrity domain. While our findings regarding the existence and the moderators of the vampire effect are very robust within the tested conditions, our studies have, in addition to the issues already discussed, certain limitations that suggest directions for further research. First, effect sizes were rather small. Thus, it would be very interesting to compare the negative effects, such as the vampire effect, with potential positive effects of celebrity endorsements, to see if neteffects are positive or negative. Second, in conducting our studies, the fixed advertisement viewing times did not allow for self-pacing by the participants, and the distraction task (i.e.,

23 21 mental math exercises, with a short time span before the recall measurement) represented a simplification of reality. Additionally, because all the studies were laboratory experiments, their external validity is limited. Moreover, we focused only on print advertising. Third, by presenting two recall measurements (unaided and aided) separately, we might have created a problem of over-concluding effects. Aided recall is the prerequisite for the formation of unaided recall, which implies that both measures might exhibit notable interrelations. Therefore, a finding of the vampire effect in one dependent variable renders its occurrence in the other dependent variable more likely. However, both aided and unaided recall are important measures, often used separately in advertising research and practice, so we considered it important to report both values. Finally, our study design followed a strict order. Because the aim of our studies was to measure brand recall, all brand-related questions appeared after all brand recall questions. Accordingly, we cannot rule out the influence of certain order effects, especially on the familiarity measures. Nevertheless, this article provides clear evidence that celebrities might overshadow the brand and thus impair brand recall the so-called vampire effect. We also offer several solutions for fighting this effect, including establishing a high endorser brand congruence or a strong cognitive link between the celebrity and the brand.

24 22 References Agrawal, J., & Kamakura, W. A. (1995). The economic worth of celebrity endorsers: An event study analysis. Journal of Marketing, 59, Alexander, M. W., & Judd, B. (1978). Do nudes in ads enhance brand recall? Journal of Advertising Research, 18, Atkin, C., & Block, M. (1983). Effectiveness of celebrity endorsers. Journal of Advertising Research, 23, Belch, G. E., & Belch, M. A. (2008). Advertising and promotion: An integrated marketing communications perspective. Boston: McGraw-Hill/Irwin. Birnbaum, M. H. (2004). Human research and data collection via the internet. Annual Review of Psychology, 55, Bruce, V., & Young, A. (1986). Understanding face recognition. British Journal of Psychology, 77, Campbell, M. C., Keller, K. L., Mick, D. G., & Hoyer, W. D. (2003). Brand familiarity and advertising repetition effects. Journal of Consumer Research, 30, Constanzo, P. J., & Goodnight, J. E. (2005). Celebrity endorsements: Matching celebrity and endorsed brand in magazine advertisements. Journal of Promotion Management, 11, Cooper, M. (1984). Can celebrities really sell products? Marketing and Media Decisions, 19, Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, Crutchfield, Dean (2010). Celebrity endorsements still push product. Ad Age.

25 23 Devue, C., Van der Stigchel, S., Brédart, S., & Theeuwes, J. (2009). You do not find your own face faster; you just look at it longer. Cognition, 111, Eisend, M. (2009). A meta-analysis of humor in advertising. Journal of the Academy of Marketing Science, 37, Erdogan, B. Z., & Baker, M. J. (2000). Towards a practitioner-based model of selecting celebrity endorsers. International Journal of Advertising, 19, ,, & Tagg, S. (2001). Selecting celebrity endorsers: The practitioner s perspective. Journal of Advertising Research, 41, Evans, R. B. (1988). Production and creativity in advertising. London: Pitman. Feldman, J. M., & Lynch, J. G. (1988). Self-generated validity and other effects of measurement on belief, attitude, intention, and behavior. Journal of Applied Psychology, 73, Forbes (2009). Celebrity 100: The world's most powerful celebrities. Retrieved from Friedman, H. H., & Friedman, L. (1979). Endorser effectiveness by product type. Journal of Advertising Research, 19, Hanssens, D. M., & Weitz, B. A. (1980). The effectiveness of industrial print advertisements across product categories. Journal of Marketing Research, 17, Hoyer, W. D., & MacInnis, D. J. (2007). Consumer behavior. Boston, MA: Houghton Mifflin. Jacoby, J., Syzabillo, G. J., & Busato-Schach, J. (1977). Information acquisition behavior in brand choice situations. Journal of Consumer Research, 3, Kahle, L. R., & Homer, P. M. (1985). Physical attractiveness of the celebrity endorser: A social adaptation perspective. Journal of Consumer Research, 11,

26 24 Kamins, M. A. (1990). An investigation into the "match-up" hypothesis in celebrity advertising: When beauty may be only skin deep. Journal of Advertising, 19, Keller, K. L. (1993). Conceptualizing, measuring, and managing customer-based brand equity. Journal of Marketing, 57, Kent, R. J., & Allen, C. T. (1994). Competitive interference effects in consumer memory for advertising: The role of brand familiarity. Journal of Marketing, 58, Kirmani, A., & Shiv, B. (1998). Effects of source congruity on brand attitudes and beliefs: The moderating role of issue-relevant elaboration. Journal of Consumer Psychology, 7, Massaro, D. W. (1970). Perceptual processes and forgetting in memory tasks. Psychological Review, 77, McCracken, G. (1989). Who is the celebrity endorser? Cultural foundations of the endorsement process. Journal of Consumer Research, 16, McCracken, J. C., & Macklin, M. C. (1998). The role of brand names and visual cues in enhancing memory for consumer packaged goods. Marketing Letters, 9, Mehta, A. (1994). How advertising response modeling (ARM) can increase ad effectiveness. Journal of Advertising Research, May/June, Meyers-Levy, J. (1991). Elaborating on elaboration: The distinction between relational and item-specific elaboration. Journal of Consumer Research, 18, Misra, S., & Beatty, S. E. (1990). Celebrity spokesperson and brand congruence: An assessment of recall and affect. Journal of Business Research, 21, Nedungadi, P. (1990). Recall and consumer consideration sets: Influencing choice without altering brand evaluations. Journal of Consumer Research, 17, Ohanian, R. (1991). The impact of celebrity spokespersons perceived image on consumers intention to purchase. Journal of Advertising Research, 31,

27 25 Park, C. W., & Lessig, V. P. (1981). Familiarity and its impact on consumer biases and heuristics. Journal of Consumer Research, 8, Patel, P. C. (2009). Impact of celebrity endorsement on brand acceptance. Journal of Consumer Behavior, 4, Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion. New York: Springer.,, & Schumann, D. (1983). Central and peripheral routes to advertising effectiveness: The moderating role of involvement. Journal of Consumer Research, 10, Ro, T., Russell, C., & Lavie, N. (2001). Changing faces: A detection advantage in the flicker paradigm. Psychological Science, 12, Rossiter, J. R., & Percy, L. (1987). Advertising and promotion management. New York: McGraw-Hill. Sengupta, J., Goodstein, R. C., & Boninger, D. S. (1997). All cues are not created equal: Obtaining attitude persistence under low-involvement conditions. Journal of Consumer Research, 23, Sliburyte, L. (2009). How celebrities can be used in advertising to the best advantage? World Academy of Science, Engineering and Technology, 58, Soley, L., & Kurzbard, G. (1986). Sex in advertising: A comparison of 1964 and 1984 magazine advertisements. Journal of Advertising, 15, Stewart, D. W., & Furse, D. H. (2000). Analysis of the impact of executional factors on advertising performance. Journal of Advertising Research, 40, Sutherland, J. C., & Middleton, L. A. (1983). The effect of humor on advertising credibility and recall. Proceedings of the 1983 Convention of the American Academy of Advertising,

28 26 Tanner, R. J., & Maeng, A. (2012). A tiger and a president: Imperceptible celebrity facial cues influence trust and preference. Journal of Consumer Research, 39, Till, B. D. (1998). Using celebrity endorsers effectively: Lessons from associative learning. Journal of Product and Brand Management, 7, , Stanley, S. M., & Priluck, R. (2008). Classical conditioning and celebrity endorsers: An examination of belongingness and resistance to extinction. Psychology & Marketing, 25, , & Shimp, T. A. (1998). Endorsers in advertising: The case of negative celebrity information. Journal of Advertising, 27, Wedel, M., & Pieters, R. (2000). Eye fixations on advertisements and memory for brands: A model and findings. Marketing Science, 19, White, D. W., Goddard, L., & Wilbur, N. (2009). The effects of negative information transference in the celebrity endorsement relationship. International Journal of Retail and Distribution Management, 37,

29 27 TABLE 1 Logistic Regression Predicting Unaided (Aided) Brand Name Recall (Study 2) Model 1 Model 2 Predictor B p OR B p OR Endorser (E) UR AR Attractiveness UR AR Gender Congruence (G) UR AR E G UR AR Nagelkerke R 2 UR AR Δ LL (df) UR (2) (2).593 AR (2) (2).053 Notes: OR = odds ratio; LL = log-likelihood, UR = unaided recall, AR = aided recall.

30 28 TABLE 2 Logistic Regression Predicting Unaided (Aided) Brand Name Recall (Study 3) Model 1 Model 2 Model 3 Predictor B p OR B p OR B p OR Endorser (E) UR AR Attractiveness UR AR Familiarity (F) UR AR E F UR AR Gender Congruence (G) UR AR E G UR AR Nagelkerke R Δ LL (df) (2) (2) (2) (2) (2) (2) Notes: OR = odds ratio; LL = log-likelihood, UR = unaided recall, AR = aided recall.

31 29 TABLE 3 Celebrity Brand Pairings Resulting from the Pretests (Study 4) Weak cognitive link Strong cognitive link CEL 1 (Stefan Raab)/Brand 1 (Congstar) (CL: M=1.94, SD=1.06) CEL 2 (Lena Meyer-Landrut)/ Brand 2 (Singstar)(CL: M=2.07, SD=1.17) CEL 3 (Michael Ballack)/ Brand 5 (Gatorade) (CL: M=2.09, SD=1.08) CEL 4 (Verona Pooth)/ Brand 6 (Maybelline Jade) (CL: M=1.95, SD=1.14) CEL 5 (Mario Barth)/ Brand 3 (Media Markt) (CL: M=4.80, SD=1.61) CEL 6 (Bruce Darnell)/ Brand 4 (Fonic) (CL: M=4.14, SD=1.93) CEL 7 (Thomas Gottschalk)/ Brand 7 (Haribo) (CL: M=5.62, SD=1.32) CEL 8 (Wladimir Klitschko)/ Brand 8 (Milchschnitte) (CL: M=3.92, SD=1.69) Notes: CEL = celebrity; CL = cognitive link; M = mean; SD = standard deviation from the main sample of Study 4.

32 30 TABLE 4 Logistic Regression Predicting Unaided (Aided) Brand Name Recall (Study 4) Model 1 Model 2 Model 3 Model 4 Predictor B P OR B p OR B p OR B p OR Endorser (E) UR AR Attractiveness UR AR Cognitive link (CL) UR AR E CL UR AR Congruence (C) UR AR E C UR AR Familiarity (F) UR AR E F UR AR Nagelkerke R Δ LL (df) (2) (2) (2) (2) (2) (2) Notes: OR = odds ratio; LL = log-likelihood, UR = unaided recall, AR = aided recall (2) (2)

33 31 Study Dependent Variables Vampire Effect TABLE 5 Summary of Results: Recall Moderators Endorser/ brand Brand familiarity congruence Cognitive link celebrity/ brand 1 Unaided brand name supported Aided umbrella brand name supported Aided product brand name supported Unaided brand name supported Aided brand name supported Unaided brand name supported not supported - - Aided brand name supported not supported Unaided brand name supported not supported supported supported Aided brand name supported not supported supported supported

34 32 APPENDIX A Measures Latent Construct Items Source Attractiveness of the In my opinion, the endorser is Ohanian (1991) endorser attractive classy beautiful elegant sexy Brand familiarity Concerning brand X, I am Kent & Allen familiar (1994) experienced knowledgeable Endorser/brand congruence When I think of endorser as an endorser, brand is one of the first brands I think about. Sengupta, Goodstein, & Boninger (1997) The idea of endorser endorsing brand represents a very good fit. I think endorser is a relevant endorser for brand. I think endorser is an appropriate endorser for brand. Strength of the cognitive link When I think of celebrity, I remember he/she endorses brand. between the endorser and the brand Celebrity has been endorsing brand for a long time. When I think of brand, one of my first associations is celebrity. Notes. All measures used seven-point Likert-type scales, anchored by 1 = completely disagree and 7 = completely agree.

35 33 APPENDIX B Advertising Stimuli Featuring Cindy Crawford/Unknown Model (Study 1)