Factors predictig the effectiveness of celebrity endorsement advertisements PDF

Preview

Summary

jfghljgjfugih...

Description

The Emerald Research Register for this journal is available at www.emeraldinsight.com/researchregister

The current issue and full text archive of this journal is available at www.em eraldinsight.com/0309-0566.htm

Factors predicting the effectiveness of celebrity endorsement advertisements David H. Silvera and Benedikte Austad University of Tromsø, Tromsø, Norway

Celebrity endorsement advertisements 1509 Received February 2003 Revised September 2003

Keywords Product endorsement, Advertising, Social attitudes Abstract This research examines whether consumers infer that celebrity endorsers like the products they endorse, and presents a model using these inferences and other characteristics of the endorser to predict attitudes toward the endorsed product. Participants in two experiments examined written endorsement advertisements and were asked to infer the extent to which the endorser truly liked the advertised product and to rate the endorser’s attractiveness, similarity to themselves, and knowledge of the product. Attitudes toward the advertisement, the endorser and the product were also measured. The resulting model indicated that product attitudes were predicted by inferences about the endorser’s liking for the product and by attitudes toward the endorser.

Introduction A recent estimate indicates that approximately 25 percent of American commercials use celebrity endorsers (Shimp, 2000). In support of this practice, research indicates that celebrity endorsements can result in more favorable advertisement ratings and product evaluations (Dean and Biswas, 2001) and can have a substantial positive impact on financial returns for the companies that use them (Erdogan, 2001). One possible explanation for the effectiveness of celebrity endorsers is that consumers tend to believe that major stars are motivated by genuine affection for the product rather than by endorsement fees (Atkin and Block, 1983). Freiden (1984) concluded that celebrities are particularly effective endorsers because they are viewed as highly trustworthy, believable, persuasive, and likeable. Although these results unequivocally support the use of celebrity endorsers, other research suggests that celebrity endorsements might vary in effectiveness depending on other factors like the “fit” between the celebrity and the advertised product (Till and Shimp, 1998). The objectives of the present research were to examine factors that influence celebrity endorsement effectiveness and to develop a model predicting the effectiveness of celebrity endorsements. Attribution theory and endorsement effectiveness Although past research documents a general tendency for consumers to believe in the purity of the motives of celebrity endorsers, it is likely that this tendency varies substantially both across consumers and across endorsers. For example, Tripp et al. European Journal of Marketing Vol. 38 No. 11/12, 2004

pp. 1509-1526 We thank Doug Krull and Bas Verplanken for their comments on an earlier version of this Limited manuscript. This research was partly supported by the Norwegian Research Council, and weq Emerald Group Publishing 0309-0566 DOI 10.1108/03090560410560218 thank them for their assistance.

EJM 38,11/12

1510

(1994) showed that celebrities who endorse several products are viewed as less credible endorsers than those who endorse only a single product. Louie and Obermiller (2002) also demonstrated that celebrities who are blamed for negative events (e.g. accidents) can have detrimental effects on the products they endorse. In short, the effectiveness of a celebrity endorser is dynamic, dependent on the celebrity, the product, and perhaps even societal conditions at the time and place where the advertisement is shown. As such, it could be fruitful to abandon the use of traditional measures of the celebrity endorser’s trustworthiness or credibility in general (e.g. Freiden, 1984; Ohanian, 1991) in favor of directly measuring the degree to which individuals evaluate the celebrity as liking the endorsed product after viewing the advertisement. Such evaluations fit under a class of judgment that has been referred to as “correspondent inferences” (Gilbert and Malone, 1995). Correspondent inferences more generally refer to any judgment in which observers use an individual’s behavior (e.g. an endorser saying that she loves Cheerios cereal) to infer congruent dispositions in that individual (e.g. inferring that the endorser actually does love Cheerios cereal). We propose that correspondent inferences are a direct measure of a celebrity’s credibility in the specific context of the advertisement, and thus should predict consumers’ attitudes toward the advertised product. H1. Correspondent inferences will be positively associated with attitudes toward the advertised product. Another interesting question in this context is whether consumers will tend to make correspondent inferences about celebrity endorsers. Early social psychological work in attribution theory (e.g. Kelley, 1971) suggests not – a large endorsement fee should be viewed as a strong incentive toward endorsement behavior, and thus observers should doubt that endorsements reflect true liking for the product on the part of the endorser. However, research indicating that celebrities are especially credible and trustworthy endorsers (e.g. Freiden, 1984) suggests that consumers might believe celebrities like the product regardless of endorsement fees. Furthermore, research examining a phenomenon called “correspondence bias” suggests that observers are biased such that they tend to attribute behavior to personal characteristics of the individual performing that behavior (e.g. liking for the product) even when situational factors (e.g. endorsement fees) are sufficient to fully explain the behavior (Gilbert and Malone, 1995; Gilbert and Jones, 1986). To our knowledge, the only empirical investigation of correspondence bias related to celebrity endorsement advertisements demonstrated correspondence bias when the advertisement included several arguments for the product and used the celebrity’s name repeatedly, but failed to demonstrate correspondence bias in a more typical advertisement that included only a picture and a single presentation of the celebrity’s name (Cronley et al., 1999). The results of this study are also difficult to interpret because the study used the midpoint of the rating scale as a comparison value rather than incorporating a more appropriate comparison value such as how the “average person” views the product. Advertising in Norway In addition to the very limited quantity of research related to correspondence bias and the endorsement process, it should also be noted that this research and most of the

research investigating correspondence bias in other contexts has been conducted in the United States. Considering that the present research was conducted in Norway, it is potentially relevant to examine cultural differences between Americans and Norwegians that might influence the effectiveness of endorsement advertisements. One important difference is that the celebrity system is primarily an American cultural enterprise. As such, it is possible that Americans identify especially strongly with celebrities and are thus more willing to accept and internalize endorsement messages (McCracken, 1989). In contrast, Norwegian cultural norms are antagonistic toward the celebrity system; celebrities might earn a certain amount of grudging respect for their accomplishments, but they are also regarded with a certain amount of suspicion and distrust because of the appearance that they view themselves as better or more important than the common man (Avant and Knutsen, 1993). Another factor that could influence the receptiveness of Norwegians to endorsement advertisements is that Norway has a less commercial culture than the United States. This non-commercial orientation is evidenced not only by Norway’s social welfare system (as opposed to the more market-driven system in the United States), but also by the tendency of Norwegian advertisers to express relatively strong concern for moral aspects of advertising (Brinkmann, 1995) as compared to advertising effectiveness, as well as Norwegian advertisers’ tendency to make arbitrary decisions based on ad hoc procedures rather than empirically grounded rational processes (Helgesen, 1992). The general lack of commercialism in Norway could have mixed effects on receptiveness to endorsement advertisements – it could produce a general distrust toward commercial ¨ve consumers who are less conscious of messages, but it could also result in more naı the size of endorsement fees and their importance in determining product endorsements. Although these considerations suggest that Norwegians might be skeptical toward celebrity endorsers and thus might not exhibit correspondence bias in the context of endorsement advertisements, correspondence bias is an extremely robust phenomenon that has been reliably demonstrated in a variety of Western cultures including Norway (Silvera et al., 2000; see also Gilbert and Malone, 1995 for a review). Based on this previous research, it was predicted that Norwegian participants would also show correspondence bias in the context of celebrity endorsements. H2. Observers will view the product endorser as liking the product better than the average person likes the product, even when the endorser receives a substantial endorsement fee. Celebrity attributes that influence endorsement effectiveness Previous research examining the effectiveness of celebrity endorsements has focused primarily on personal attributes of the celebrity that enhance his or her persuasiveness (see Erdogan, 1999, for a review). For example, a number of researchers have used models in which “source credibility”, typically viewed as a function of trustworthiness and expertise, is the primary factor determining how influential the endorser will be (Ohanian, 1991; Lafferty and Goldsmith, 1999; Lafferty et al., 2002). Trustworthiness refers to the general believability of the endorser, and is thus broader but conceptually similar to correspondent inferences about the endorser. Expertise refers to the product knowledge of the endorser and thus to the validity of his or her claims regarding the

Celebrity endorsement advertisements 1511

EJM 38,11/12

product, and is believed to be a factor that increases persuasiveness above and beyond the effects of trustworthiness. H3. Perceived product knowledge of the endorser will be positively associated with attitudes toward the advertised product.

1512

Other researchers have emphasized the importance of source attractiveness in determining liking for the endorser and thereby increasing endorsement effectiveness (e.g. Friedman and Friedman, 1979). To the extent that attractiveness is an important determinant of endorsement effectiveness, research based on the matchup hypothesis (Kahle and Homer, 1985; see also Till and Busler, 1998, for a discussion of the matchup hypothesis in relation to endorser expertise) suggests that its importance is limited by the degree to which attractiveness “fits” well with the advertised product (e.g. Kamins, 1990). Thus, for example, physical attractiveness might be useful when selling cosmetics but not when selling computers. Furthermore, although source attractiveness research has focused primarily on physical attractiveness, attractiveness can also be viewed more generally as a positive attitude toward the endorser. Such positive attitudes might result from admiration or perceived similarity (McGuire, 1985), although effective advertising is more likely to rely on the admiration component because the influential power of celebrities is closely connected to their status as role models (Raven et al., 1998). H4. Positive attitudes toward the endorser will be associated with positive attitudes toward the advertised product. H5. Admiration and elements of physical attractiveness associated with the product will be associated with positive attitudes toward the endorser, and will contribute to predicting attitudes toward the advertised product. H6. Perceived similarity and elements of physical attractiveness that are unrelated to the product will be associated with positive attitudes toward the endorser, but will not contribute to predicting attitudes toward the product. Experiment 1: research methodology Participants The participants were 66 students (41 females, 25 males) at the University of Tromsø who participated on a voluntary basis. The mean age of the participants was 24.59 years (SD ¼ 1:91 years). Materials Baseline measure. Prior to examining any of the other experimental materials, participants were asked to indicate what they believed people in general think about the Omega watch brand on a scale from 1 ¼ “not good at all” to 7 ¼ “very good”. Endorsement advertisement. Participants were given a black-and-white copy of a printed advertisement in which Cindy Crawford endorsed the luxury, diamond-set watch “Constellation Carre `” from Omega. The advertisement had a close-up picture of Cindy Crawford with the watch in front of her. “Cindy Crawford’s choice” was printed above the picture, and “My choice” was written in cursive next to the Omega logo at the bottom of the advertisement.

Evaluation questionnaire. The evaluation questionnaire included measures of correspondent inferences, attitude toward the advertisement, attitude toward the product, and attitude toward the celebrity endorser. Correspondent inferences were measured with three items asking participants to indicate their level of belief that Cindy Crawford: (1) likes the Omega brand; (2) frequently uses the Omega brand; and (3) views the Omega brand as a good product. Each of these items was answered on a seven-point scale ranging from 1 ¼ “Not at all likely” to 7 ¼ “Extremely likely”. The correspondent inference items showed satisfactory internal reliability a( ¼ 0:83). All attitude items used semantic differential scales with a value of one associated with the more negative word and a value of seven associated with the more positive word. Attitude toward the advertisement was measured using the items pleasant/unpleasant, likeable/not likeable, interesting/uninteresting, and good/bad; attitude toward the product was measured using the items desirable/not desirable, pleasant/unpleasant, likeable/not likeable, and good/bad; attitude toward the endorser was measured using the items interesting/uninteresting, pleasant/unpleasant, likeable/not likeable, and good/bad. Each of the three attitude measures showed satisfactory internal reliability (attitude toward advertisement:a ¼ 0:85; attitude toward product: a ¼ 0:82; attitude toward endorser:a ¼ 0:87). Participants were also asked to rate how much they admired Cindy Crawford as well as Cindy Crawford’s physical attractiveness, sense of style (attractiveness relevant to the product domain), similarity to the participant, and knowledge of the product domain. All of these ratings were made on seven-point scales where 1 indicated low values (e.g. “not at all attractive”) and 7 indicated high values (e.g. “very attractive”). Procedure Small groups of participants received an experimental pack that was completed in the presence of the experimenter and returned immediately upon completion. The first part of the experimental pack contained a brief description of the study followed by the baseline measure. After completing the baseline measure, participants were asked to read a set of instructions that contained the experimental manipulation. Participants were randomly assigned either to the “unpaid” condition, in which they were told that Cindy Crawford endorsed Omega watches on a voluntary basis without receiving any payment, or the “paid” condition, in which they were told that Cindy Crawford was paid 50 million Norwegian kroner (approximately US$6 million) for endorsing Omega watches. This amount was described as Cindy Crawford’s standard endorsement fee. After reading the instructions, participants viewed the endorsement advertisement and completed the evaluation questionnaire. Analyses In addition to preliminary correlational analyses, the primary results will be presented using a structural equations model developed in Amos 4.0 (Arbuckle and Wothke, 1999). In addition to the chi-square measure of absolute fit, the comparative fit index

Celebrity endorsement advertisements 1513

EJM 38,11/12

1514

(CFI) and the root mean square error of approximation (RMSEA; Browne and Cudek, 1993) were used as measures of relative model fit. Even for models with poor absolute fit, recent research suggests that CFI values of 0.95 or higher (Hu and Bentler, 1999) and RMSEA values of 0.08 or lower indicate reasonably good overall model fit (Browne and Cudek, 1993). Results Preliminary analyses Gender was originally included as a covariate in all analyses, and was not significantly associated with any of the primary variables in the study. Thus, gender was not included in the analyses reported below and will not be discussed further. Pearson correlations between the variables in Experiment 1 are shown in Table I. These correlations are consistent with all hypotheses except H3 – perceived product knowledge of the endorser was not significantly correlated with attitudes toward the advertised product. Model predicting attitude toward the endorsed product A theoretical model was first developed based on the hypotheses that attitudes toward the product were predicted by attitudes toward the endorser, perceived product knowledge of the endorser, and correspondent inferences about the endorser; and that attitudes toward the endorser were predicted by admiration of the endorser, the endorser’s sense of style, perceived similarity to the endorser, and physical attractiveness of the endorser. This theoretical model had both poor absolute fit ( x 2 (21, N ¼ 66Þ ¼ 85:04, p , 0:001) and poor relative fit (CFI ¼ 0:94, RMSEA ¼ 0:22). In the next step, perceived similarity and physical attractiveness were removed from the model in accordance with the prediction that they would not contribute to predicting attitude toward the product (H6). Although this revised model still had poor absolute fit (x 2 (10, N ¼ 66Þ ¼ 25:09, p , 0:01) and somewhat poor relative fit (CFI ¼ 0:98, RMSEA ¼ 0:15), it was a significant improvement over the original model (x2change ð11Þ ¼ 59:85; p , 0:001), supporting H6. This model was then refined to improve goodness of fit by making two further modifications: (1) sense of style predicted correspondent inferences in addition to attitude toward the endorser; and (2) product knowledge predicted correspondent inferences instead of attitude toward the product. This final model (see Figure 1) had good absolute fit x( 2 (9, N ¼ 66Þ ¼ 11:14, p . 0:25) and relative fit (CFI ¼ 1:00, RMSEA ¼ 0:06), and explained 27.01 percent of the variance in attitude toward the advertised product. All paths shown in Figure 1 are significant at p , 0:05. The association between correspondent inferences and attitude toward the product supports H1, and the paths from admiration and sense of style through attitude toward the endorser to attitude toward the product support H4 and H5. H3 is also partly supported by the model; although there is no direct path between product knowledge and attitude toward the product, these two variables are indirectly connected via correspondent inferences.

Att: endorser Corr infs Att: ad Att: product Admiration Attractiveness Similarity Style Knowledge

Attitude: endorser

Correspondent inferences

Attitude: advertisement

Attitude: product

0.18 0.69*** 0.36** 0.52*** 0.35** 0.34** 0.59*** 20.08

0.30* 0.50*** 0.09 0.05 0.28* 0.40** 0.21

0.34** 0.31* 0.34** 0.19 0.45*** 0.03

0.27* 0.22 0.17 0.45*** 20.03

Physical Admiration attractiveness

0.46*** 0.41** 0.30* 20.02

0.57*** 0.15 20.22

Similarity

0.33** 0.05

Sense of style

...