Keywords

Introduction

Advertisements play a key role in communicating with consumers in today’s society. This means that promotion is often one of the most important avenues through which information about the product or brand is communicated to potential consumers. Among the variety of possible communication strategies, one of the most frequently used approaches is the use of a celebrity endorser in an advertisement. Celebrities have been shown to be used in this way for a wide variety of brands around the world (Choi, Lee, & Kim, 2005).

A considerable number of theories have been devoted to the use of celebrity endorsers. These theories generally propose that celebrities are effective because they may increase attention to the message, they may increase the memorability and likeability of the advertisements, or they are usually seen as trustworthy (e.g., Menon, Boone, & Rogers, 2001; Pornpitakpan, 2003; Pringle & Binet, 2005; Roy, 2006). These factors are hoped to ultimately alter the purchase intentions of consumers.

What factors could make a celebrity effective? Friedman and Friedman (1979) define a celebrity as “an individual who is known to the public (actor, sports figure, entertainer, etc.) for his or her achievements in areas other than that of the product class endorsed.” Considerable research has been devoted to the effects of celebrity endorsers on consumer behavior. Most of the approaches propose that celebrities are effective and point to attractiveness and credibility as possible mechanisms for their effectiveness. The selection of these two factors, however, is based on existing theory and hunches and does not preclude other factors that may make a celebrity effective.

At least one empirical study has analyzed why advertisers make use of celebrities in product advertising (Erdogan, Baker, & Tagg, 2001). There have also been dozens of studies that has tried to explain if celebrity endorsements influence consumer reactions. This research has mostly focused on credibility and physical attractiveness. It is believed that several factors related to credibility – perceived trustworthiness, expertise, familiarity, and likability – determine their effectiveness (Kamins, 1990). To date this research has examined the attractiveness of the celebrity (Kahle & Homer, 1985), liking (Kahle & Homer, 1985) and the celebrity’s expertise with the product (Ohanian, 1991). To test the effects of these possible factors, research has examined various outcomes – generally focusing on evaluating consumers’ responses to advertisements using images of these celebrities.

The two most studied reasons for celebrities being effective are their physical attractiveness and credibility. On the issue of physical attractiveness, there are good reasons to believe that physical attractiveness may be an important force. Social psychology research has shown that physical attractiveness generally enhances a person’s effectiveness as a social influence agent (Baker & Churchill, 1977; Chaiken, 1986; Dion, Berschied, & Walster, 1972). In the context of advertising, Kahle and Homer proposed that celebrities are effective endorsers because they are more attractive than noncelebrities; this is often referred to as the “attractiveness” hypothesis (Kahle & Homer, 1985). In support of the attractiveness hypothesis, Kahle and Homer (1985) showed that razor ads featuring attractive celebrities resulted in more positive attitudes toward the ad than those with less attractive celebrities. Thus, attractiveness of the celebrity was an important condition underlying attitudes toward the product and purchase intentions.

Research has probed when physical attractiveness is important. The celebrity-product match-up hypothesis suggests that attractive endorsers should be more effective when promoting products that are related to physical attractiveness such as makeup or hair products. Thus, attractiveness is believed to be more important when the product is “attractiveness-related” (Kahle & Homer, 1985; Kamins, 1990). Some research also demonstrates that physical appearance is also more effective for low-involvement products than for high-involvement products (Shavitt, Swan, Lowery, & Wanke, 1994). Consistent with the match-up hypothesis, however, celebrity expertise was shown to be more important than attractiveness for technology products (Biswas, Biswas, & Das, 2006).

As a result of these findings, theorists have proposed that it is usually import to have a fit between the attributes of the celebrity and the attributes of the advertised product. Till and Busler found that there should be a congruence between the celebrity and the product in terms of characteristics such as image, expertise, or attractiveness. According to this approach, the “fit” is high when the various characteristics are congruent. This congruity could arise from physical attractiveness (Kahle & Homer 1985; Kamins, 1990), expertise (Ohanian, 1990), or liking (Kahle & Homer, 1985). Wang, Chao, and Wang conducted an experiment to examine the effects of consumer involvement, fit, and the number of celebrities on consumer behavior in a series of brand endorsement ads. The authors found that celebrities are not always better. They found that fit is important for a high-involvement target audience, but the number of celebrities (consensus effect) is more important for the low-involvement target audience. The results of this research suggest that both high- and low-involvement audience can be targeted at the same time depending on the level of fit and the number of celebrities used in a brand endorsement.

In sum, evidence suggests that attractiveness should be more important for attractiveness-related products; however, it is not clear if attractiveness will still be effective in the case of products that are unrelated to “attractiveness.” Two pieces of evidence point to its across-the-board effects. First, Kahle and Homer (1985) found that in the case of attractiveness-related products, the use of physically attractive celebrities increased message recall and purchase intention across a variety of products. Similarly, Praxmarer (2011) demonstrated that attractive models are effective for even advertising non-attractiveness-related product – in this case, a food supplement. This leads to the following hypothesis.

H1

Physical attractiveness will increase the effectiveness of a celebrity.

Although considerable research has discussed the importance of “credibility” as a determinant of message effectiveness, the question arises as to where celebrity credibility arises. One possible factor behind the power of celebrities is people’s sense of identification with them. The idea of “identification” can be traced back to the 1950s when two clinical psychologists, Horton and Wohl (1956), began noticing that patients felt a sense of intimacy with television personalities. They referred to the imaginary relationship between a television viewer and television personalities as a parasocial relationship.

Research on identification process began to discover that repeated exposure to media figures through the mass media creates a sense of friendship or intimacy in media users (Levy, 1979). Audience members commonly look to media personalities as “friends” and those with whom they feel “comfortable.” This line of research suggests that parasocial identification can occur through repeated media exposure. Popular books have called this being “star struck” (Fowles, 1992).

Reviewing Burke (1950), Kelman (1961), and Bandura’s (1977) theories of social influence, Basil (1996) proposed that one reason that celebrities may be effective endorsers is that people identify them as friends. This approach suggests that a spokesperson that the audience feels to be a friend will be more effective in creating attitude or behavior change (Basil, 1996). As a result, identification, or the feeling of friendship with a celebrity, may also explain the effectiveness of celebrity endorsers because friends are more powerful influences that unknown others (Brown, Basil, & Bocarnea, 2003a). One of the limitations of Kamins’s (1990) study was that the limitation to two celebrities resulted in attractiveness and likeability being confounded. Subsequent research has shown that identification is important in the context of health messages (Brown & Basil, 1995; Brown, Basil, & Bocarnea, 2003b). The importance of likeability was shown to be important factor to advertisers also (Erdogan et al., 2001). However, the importance of identification has generally not been tested in the context of consumer goods. There is one recent study suggesting that identification may be relevant in consumer advertising (Basil, 2012). In sum, we wanted to test the importance of identification with a celebrity endorser.

H2

Identification will increase people’s interest in a product.

Finally, which of the two factors – attractiveness and identification – are most important in determining the power of a celebrity? Do these factors have some multiplicative effect so that ads are more effective when they use a combination of these forces?

RQ1

Is physical attractiveness or identification more important in shaping interest in the ad?

RQ2

Is there a nonlinear interaction between attractiveness and identification so that the most effective messages will use a combination of physical attractiveness and identification?

Methods

To investigate the importance of attractiveness and identification, two pretests and an experiment were conducted. All of these studies were conducted online (hosted on the Qualtrics platform).

Pretest 1

The first pretest was used to select well-known celebrities. A survey listed 50 male and female celebrities from five broad categories – acting, food, music, sports, and television. Respondents were asked to pick celebrities that they were most familiar with. The ten celebrities that resulted were Jennifer Aniston, David Beckham, Justin Bieber, Ellen DeGeneres, Robert Downey Jr., Jimmy Fallon, Heidi Klum, Gordon Ramsay, Taylor Swift, and Serena Williams.

Pretest 2

The second pretest was used to select celebrities that varied in physical attractiveness and identification. In this pretest, respondents were shown a picture of the celebrity to rank the celebrity on attractiveness and identification. The questions were [celebrity name] is (1) attractive/beautiful/handsome, (2) is likeable, (3) is someone whom I could see as a friend, and (4) is someone whom I admire. Each was asked on a 1–10 scale (1 (strongly disagree), 5 (neutral), 10 (strongly agree)). A factor analysis examined dimensionality. Two factors emerged: attractiveness (single item) and identification (Cronbach’s alpha = 0.935). These results were used to select celebrities that were more or less attractive and for whom respondents felt close (identified and liked) or not. Seven celebrities fit well into the fully crossed design – David Beckham, Justin Bieber, Ellen DeGeneres, Jimmy Fallon, Heidi Klum, Gwyneth Paltrow, and Taylor Swift. The overall evaluations and categorizations are shown in Table 1. One cell remained empty – an attractive male to whom respondents did not identify. We brainstormed with a small sample of students who would fit this cell and identified Benedict Cumberbatch.

Table 1 Pretested celebrities
Full size table

Experiment

Subjects

A total of 235 respondents were used in this study. These respondents were recruited from the students at the University of Lethbridge. To further broaden the demographics, friends and family of students were also recruited. Of the 245 that started the study, 239 completed, a 98% response rate. After excluding 2 participants who did not indicate their gender and 2 whose responses were incomplete, the final sample consisted of 104 males (44%) and 131 females (55%). Respondents age ranged from 18–69 (M = 23.9, median = 23).

Procedure

Using an online platform, subjects were shown a total of eight advertisements and asked to rate the attractiveness and likeability of each celebrity and their interest in finding out more about the product. Subjects were shown one advertisement at a time and then asked a series of seven questions for each ad. Multiple advertisements were used in each condition to increase generalizability of the findings (Jackson, 1992).

Stimuli

Of the eight advertisements that were selected for the final study, four used a female celebrity and four used a male celebrity. The final ads used are shown in the Fig. 1.

Fig. 1
figure 1

Experimental stimuli

Full size image

Measures

Four questions were intended to measure attributes of the celebrity: [celebrity X] (1) is someone whom I can see as a friend, (2) is likeable, (3) is attractive, (4) is good-looking. Three questions were intended to measure advertising outcomes: (5) how interested are you in [this brand or store], (6) how interested are you in learning more about the product, and (7) how likely are you to consider purchasing [brand or from this store]?

Results

Manipulation Checks

The mean attractiveness rating for celebrities selected to be attractive was 7.48, while those believed less attractive was 5.99 [out of ten]. The selection of attractiveness was deemed successful (t = 11.2, p < 0.001). The measures of identification also showed that the identification level was also successfully selected (6.38 vs 5.18, t = 11.6, p < 0.001).

Scale Construction

Before proceeding to the analyses, we examined the dimensionality of our variables. We intended two attractiveness items – [celebrity X] is attractive, and [celebrity X] is good-looking. Factor analysis and reliability measures indicated that this is a single unidimensional construct (Cronbach’s alpha = 0.829 across the eight celebrities). We intended two likeability items – [celebrity X] is someone whom I can see as a friend, and [celebrity X] is likeable. Factor analysis and reliability measures also suggested a single unidimensional construct (Cronbach’s alpha = 0.887 across the eight celebrities).

The three dependent items were (1) how interested are you in the brand, (2) would you like to learn more, and (3) based on the advertisement, how likely are you to consider purchasing [product]? Factor analyses and correlation and reliability tests suggested that these measures were also unidimensional (Cronbach’s alpha = 0.907). As a result, these three items were averaged into a single outcome measure – ad interest.

Analyses

A regression analysis examined the relationships between attractiveness and identification on interest in the product. For the entire sample, attractiveness showed a significant relationship with interest supporting H1 (beta = 0.540, t = 7.86, p < 0.001). Contrary to H2, identification was not significant (beta = 0.102, t = 1.49, p = 0.138). The overall R 2 for the equation was 0.374.

Next, we broke the analysis down by gender of respondent. For male respondents, both attractiveness (beta = 0.429, t = 3.97, p < 0.001) and identification (beta = 0.227, t = 2.11, p = 0.038) were related to interest. This supports H1 and H2. The overall R 2 for the equation was 0.367. For female respondents, only attractiveness (beta = 0.583, t = 6.17, p < 0.01) predicted interest, supporting H1. Contrary to H2, identification was not a significant predictor (beta = 0.050, t = 0.56, p = 0.600) of interest for women. The overall R 2 for the equation was 0.382.

To test the research question, we investigated the possibility of an interaction between these two factors. The results for the entire sample show no significant interaction effect (beta = −0.192, t = 0.62, p < 0.539). Similar results were found for both men (beta = 0.026 t = 0.47, p = 0.962) and women (beta = −0.374, t = 0.96, p = 0.361). Further, with the interaction terms included, the effects of attractiveness and identification remained consistent with the simpler regression. For female respondents, only attractiveness (beta = 0.845) predicted interest, identification did not (beta = 0.186). For male respondents, both attractiveness (beta = 0.413) and identification (beta = 0.215) were significant.

Discussion

The results of this study support Hypothesis 1 – that attractiveness is an important predictor of interest in a product. Interest was used as a measure of advertising effectiveness. The evidence that using an attractive model is an effective strategy has been shown in a good deal of earlier work. Our findings also support the importance of attractiveness generally. This finding conflicts, however, with a study reporting what advertisers perceive to be important (Erdogan et al., 2001). Attractiveness, however, may not be equally effective for all products. As a result, these findings can reconcile two seemingly contradictory findings – (1) that attractiveness is effective and (2) that attractiveness may be more relevant and effective for products that are beauty-related (Baker & Churchill, 1977; Kahle & Homer, 1985; Kamins, 1990). In some of these cases, attractiveness may not only be a peripheral cue but can also be product-relevant where attractiveness of the product “fits” the image of the celebrity. The importance of attractiveness was not due to fit because products advertised with women were less related to physical attractiveness; in fact, only one of the four ads with female celebrities was somewhat attractiveness-related (perfume), while three of the four ads featuring male celebrities was attractiveness-related (underwear, jeans, and cologne). Therefore, attractiveness seems to have an effect for all products, although that effectiveness may be more important for attractiveness-related products and less important for non-attractiveness-related products. Therefore, this observation may help us resolve the seemingly conflicting findings from social psychology with that of the match-up hypothesis.

The results only partially support Hypothesis 2 – that identification determines product interest. In this case the finding appeared only for male respondents. Therefore, our results only offer partial support for the importance of identification. As a result we must consider that men and women may have different criteria or ways that they perceive and assess celebrities in advertising. Given the fact that most of the previous research on identification has studied male celebrities (e.g., Brown & Basil, 1995), it is possible that males are more influenced by identification with the celebrity and that this increases their interest in or the influence of an advertisement. This result is consistent with Basil’s (1996, 2012) studies which showed that identification increases the effectiveness of celebrities.

With regard to the research questions, the relative size of these effects suggests that attractiveness was more important than identification (attractiveness beta = 540, identification beta = 0.102). Finally, there was no evidence of an interaction where a combination of these factors would be especially effective.

Based on our result, consumers are generally more interested in learning about a product that is advertised with attractive people. This supports the notion that physically attractive endorsers are generally more successful (Chaiken, 1986). The effectiveness of a message also may depend on the felt identification in terms of similarity, familiarity, and liking of the endorser.

Limitations and Further Research

One limitation of this study was the reliance on a student sample. To correct this, we would suggest collecting a more representative sample of the population, such as recruiting a sample from research panels such as MTurk. Another limitation was the limited variation in the variety of advertisements used. Although we believe the study was strengthened through the use of eight advertisements to increase the generalizability and validity of the study (Jackson, 1992), it is possible that there are other types of ads and other types of executions that might result in different effects of physical attractiveness and identification. A handful of celebrity advertisements cannot accurately predict the reflection of one celebrity compared to another celebrity in the same category.

For later research, we would suggest including other factors that may shape credibility such as admiration (Basil, 2012). While we are confident that a relationship exists between attractiveness and identification on advertisement interest, there are likely other variables that may affect a respondent’s interest in the advertisement. Including all of these factors across a wider variety of messages could help determine which factors have stronger effects and under which circumstances, such as the type of product. Interest in advertisements is likely dependent on the individual’s interest in particular products, and this could also be taken into consideration in future studies.