Online Celebrity Endorsement Effectiveness
A quantitative study examining the moderating effect of the online
environment on the factors predicting celebrity endorsement effectiveness
Yannick Admiraal 11045825
Bachelor Thesis Business Administration Subject: Digital Marketing
Faculty of Economics and Business, University of Amsterdam Bachelor’s programme Business Administration
Date: 26-06-2018 Study year: 2017-2018 Supervisor: Drs. C.A.W. Troost
Statement of Originality
This document is written by Student Yannick Admiraal who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents
ABSTRACT
Marketers frequently use celebrity endorsers as credible sources to influence consumers'
attitudes. However these elusive celebrities, adored by the many, do not come cheap. As a result, companies have been looking for guidance from the academic world to avoid unnecessary marketing expenses of costly failed attempts at effective celebrity endorsement. For last 50 years researchers have tried to provide this guidance and succeeded to a certain degree. They
established models to predict and explain celebrity endorsement effectiveness. The source credibility model has been traditionally used to determine the celebrity credibility and the resulting effectiveness of the celebrity endorsement, based on three celebrity characteristics: attractiveness, trustworthiness and expertise. However, these models have been construed based on traditional media. It has not been tested whether they are still properly applicable in the online environment this study aims test how these traditional success factors of celebrity endorsement hold up in the digital world. This research investigates the moderating effect of the online environment on the traditionally construed source credibility dimensions through statistical analysis of quantitative data collected through an online questionnaire. The research concluded in the online environment, higher celebrity attractiveness leads to a greater increase in positive attitude. The study establishes that celebrity expertise has a significant effect on the attitude toward the ad but is not moderated by the type of medium. It is also determined that celebrity trustworthiness has an insignificant effect on attitude toward the ad regardless of medium type. Based on the findings of this study, marketers launching celebrity endorsement campaigns in the online environment and on social media should select a celebrity with high credibility and especially with high attractiveness to be most effective at enhancing the attitude toward the ad and consequently the ad’s effectiveness.
TABLE OF CONTENTS
Introduction………..……p. 1
Theoretical Background And Hypotheses………..…………..p.3
Methods……….……..p.14 Results………..……..….p.24 Discussion………p.32 Conclusion……….……..p.38 Bibliography………..…..p.40 Appendices Appendix A………p.1
1 INTRODUCTION
It is crucial for organizations to develop different marketing strategies that help to effectively and efficiently establish a positive image of their product, service and brand in the minds of the consumers (Erdogan, 1999). Throughout the years, celebrity endorsement has been one of the main tools for marketers to aid in the communication of a more positive image with regard to their product or service (Erdogan, 1999). Celebrity endorsement has been proven to be an effective advertisement strategy for many companies in many different industries (Ohanian, 1991). It is very important for companies, interested in involving a celebrity in their marketing communications to endorse their product or service, to heavily contemplate beforehand what kind of celebrity would be best fitted for the job. Properly selecting the celebrity in advance is crucial because it can heavily afflict the marketing budget as these elusive celebrities do not come cheap (Erdogan, 1999). Over the years firms have been investing large amounts of money to help associate their brands with the celebrities to subsequently transfer the desired image of the celebrity onto their products or services which ultimately lead to more sales (McCracken, 1989).
However, celebrities are people and people change and as such endorsements can go bad. A certain celebrity might first seem like the perfect person to endorse your brand, but later can become more detrimental than beneficial for your desired brand image. As such, celebrity
endorsement can be a two-edged sword, which can make selecting the right celebrity a costly and challenging process (Erdogan, 1999). As a result, companies have been looking for guidance from the academic world to avoid unnecessary marketing expenses of costly failed attempts at effective celebrity endorsement. For last 50 years researchers have tried to provide this guidance and succeeded to a certain degree (Kaikati, 1987; Thompson, 1978; Treece, 1972). Celebrity endorsement literature proposes that the effectiveness of the celebrity endorsement is greatly
2
determined by the credibility of the celebrity (Ohanian, 1991). It is generally concluded that the most important predicting characteristics of the celebrity credibility are celebrity attractiveness, trustworthiness and expertise (Hovland et. al., 1953). Furthermore, the alignment of the product endorsed and the celebrity has been brought forward as another prominent success factor of celebrity endorsement (Kamins & Gupta, 1994). Moreover, the discrepancy between the
consumer’s ideal self-image and the celebrity’s image has also been mentioned as another factor
predicting the effectiveness of celebrity endorsement (Choi & Rifon, 2012).
However, those theories and models have been construed based on traditional offline media even though now celebrity endorsement is also being deployed in the online environment. Previous research showed that is it fruitful to be skeptical with regard to the applicability of traditional celebrity endorsement models in the online environment (Karson & Korgaonkar, 2001). Therefore, it is of the essence to further test how the traditional celebrity endorsement models hold up in in today’s digital age. This study aims to shed more light on this issue by answering this question: What is the effect of the online environment on celebrity endorsement effectiveness? The question will be answered by quantitatively analyzing the difference in the
perceived source credibility dimensions and their subsequent effect on the attitude toward the ad between the online and offline environment through the use of a questionnaire. The main
objective of this research is to expand current expand current theorization by incorporating the moderating effect of the online environment on the source credibility dimensions. First a theoretical background and the resulting hypotheses will be proposed. Second the quantitative methods used will be discussed. Then the results will be presented and discussed as to which hypotheses are accepted or rejected and provide explanations for those results. Finally, the research question will be answered and discussed how it contributes to the current theorization.
3
THEORETICAL BACKGROUND AND HYPOTHESES
Celebrity Endorsement Effectiveness
What the research on celebrity endorsement effectiveness has generally concluded is that a celebrity in general has a greater effect on attitude change and purchase intention than an unknown spokesperson. In traditional literature this effect is mainly explained through the halo effect (Thorndike, 1920). The Halo Effect refers to the disregard for specific characteristics of a
product as result of a generally perceived positive or negative “halo” around an individual associated with that product (Thorndike, 1920). In terms of celebrity endorsement, the general positive impression a person has of a celebrity can overshadow the possible negative
characteristics of a product endorsed or simply strengthen the already-existing positive image a brand or product has through the endorsement (Djafarova & Rushworth, 2017).
Moreover, in celebrity endorsement research the application of the source credibility model (Hovland, Jannis, & Kelley, 1953) is also strongly supported. The halo effect and the
source credibility model are fundamentally interrelated because they both evaluate the credibility
of the source based on characteristics such as attractiveness and trust (Djafarova & Rushworth, 2017). The source credibility model assumes that a greater celebrity credibility results in greater effectiveness of the endorsement when used in advertising. The model establishes 3
characteristics of the celebrity that influence and predict the credibility of the endorsement: expertise, trustworthiness and attractiveness. Expertise is the degree to which the celebrity endorser is regarded as a source of valid assertions, meaning that their statements are confidently perceived as reasonable and valid (Hovland et. al., 1953). Celebrity expertise is derived from their knowledge, experience or training on the subject of the endorsed product (Choi, Rifon, 2012). Trustworthiness is the level of the perceived objectiveness and honesty of the message the
4
endorser tries to convey (Erdogan, 1999). Attractiveness is the likeability and physical attractiveness of the celebrity (Choi & Rifon, 2012).
However, several researchers of studies following Hovland and colleagues’ research
criticized the source credibility model because it assumed that the effectiveness of celebrity endorsement is solely based the endorser’s credibility without attributing for the fact that the
product endorsed might also play an important role in the overall effectiveness (Choi & Rifon, 2012; McCracken 1989). They proposed that a proper fit between the endorsing celebrity and the advertised product is also an important determinant of the endorsement effectiveness. This relationship was defined as the Celebrity/Product Congruence and was recently further explored by Choi and Rifon (2012). In their research they confirmed that a proper fit between endorser and product characteristics indeed had a considerable impact on the consumer attitude toward the advertisement and consequently on their attitude toward the brand and the purchase intention (Choi & Rifon, 2012). So they concluded that effective celebrity endorsement not only depends on the celebrity credibility but also on the perceived fit between the celebrity and the endorsed product.
Additionally, a second congruency based on the consumer’s self-concept was brought up to further attempt to explain the effectiveness of celebrity endorsement. An individual’s
self-concept is defined as the aggregate of their own thoughts and feelings about themselves in relation to other individuals (Onkvisit & Shaw, 1987). An individual’s self-concept consists of
two elements: actual self and ideal self. The actual self is comprised of the way in which a person perceives themselves based on their own attributes and of the perception they believe others have of them. While the ideal self refers to how a person would to perceive themselves and how they would want to be perceived by others (Sirgy, 1987). According to Graeff (1996), consumer
5
behavior can be predictable because they act in order to maintain and try to improve their own self-concept. Consumers also are intrinsically motivated to strive for consolidation of their actual self and ideal self (Higgins, 1987). So supported by the conclusions of this previous research on self-concepts, Choi and Rifon construed the Celebrity/Consumer Congruence to further analyze the effectiveness of this self-concept theory in a celebrity endorsement environment.
The Celebrity/Consumer Congruence consists of the coherence of the consumer’s ideal self-image in relation to the concurrent celebrity’s image (Choi & Rifon, 2012, p.641). The researchers found that a consumer’s alignment with a celebrity endorser in terms of personal
aspiration and ideal self-image has a positive significant effect on their attitude toward the ad. However, the Celebrity/Consumer Congruence had an insignificant effect in regard to the attitude toward the brand and as a result brand credibility (Choi & Rifon, 2012). Celebrity/Consumer Congruence had no notable effect on the consumer’s attitude toward the brand, indicating that an endorsement by an aspired celebrity would not directly impact the consumer’s attitude toward the
brand and therefore will not affect the brand credibility. This discrepancy of personal aspirations between the actual self and the ideal self as portrayed by the celebrity was also more recently analyzed through a series of in-depth interviews by Djafarova and Rushworth (2017).
Contradictory to the earlier research by Choi and Rifon (2012), the interviews revealed that participants would be willing to trust the celebrities to whom they aspire, as they presume that celebrities value their position of power and are unlikely to abuse it, therefore their opinions regarding a particular brand or website would be considered credible even if the brand was previously unknown to them (Djafarova & Rushworth, 2017, p.5). This finding could partially be explained by the halo effect¸ where a “halo” of trustworthiness surrounding the aspired celebrity radiates onto the brand and subsequently enhances the brand credibility (Thorndike, 1920). This
6
also further supports the source credibility model, due to the trustworthiness of the celebrity directly influencing brand credibility (Hovland et al., 1953). Accordingly it can be argued that celebrity endorsements could thus enhance brand credibility if they are aspired by consumers.
Celebrity Credibility and Brand Credibility
Controversially, in recent research the direct relationship between celebrity credibility and brand credibility is strongly attested. In this research source credibility dimensions are tested for their effect on celebrity credibility and furthermore how this celebrity credibility can directly influence brand credibility (Spry, Pappu & Cornwell, 2011). It was concluded that, through endorsement, consumers will associate certain brands with the celebrity, which subsequently add source credibility dimensions of attractiveness, trustworthiness and expertise to that brand, assisting in
building brand credibility (Spry et al., 2011, p. 895). This finding however is very controversial because it heavily contradicts the traditional research regarding celebrity credibility, brand credibility and the mediating role of attitude towards the ad. The traditional view based on the theory constructed by Goldsmith, Laverty & Newell (2000) constituted that celebrity credibility directly influences the consumer’s attitude toward the ad, which subsequently determines
celebrity endorsement effectiveness.
Furthermore, the study by Goldsmith and colleagues disputed any direct relationship between celebrity credibility and brand credibility. They argued that the effect of endorser credibility on attitude toward the brand is mediated by the attitude toward the ad (Goldsmith, Laverty & Newell, 2000). This statement is also supported by most conventional research literature regarding the consumer attitude toward ads. Goldsmith refers to Dual Mediation
7
Hypothesis (MacKenzie, Lutz & Belch, 1986) and The Elaboration Likelihood Model (Petty &
Cacioppo, 1981) to support these claims. The Dual Mediation Hypothesis indicates an exclusively indirect effect of Ad Cognitions (i.e., ad characteristics such as endorsement) on attitude-toward-the-brand, not a direct one (Goldsmith et al., 2000, p.45-46). This indirect effect of ad cognitions on toward-the-brand is caused by the intervening mediation of attitude-toward-the-ad. The mediating role of attitude-toward-the-ad is explained by the argument that the perceptions of the source of a message are initially influencing the attitude towards the source, which sequentially conducts the cognitive and affective response to the content of the original message (MacKenzie et al., 1986, p. 132).
So in terms of advertisements, the consumer’s initial perceptions of the advertisement firstly influence their attitude toward the advertisement, which in turn dictates their cognitive and affective response to the content of the advertised message. These cognitive and affective
responses (i.e., ad cognitions) consequently determine the consumer’s predisposition to accept the statements made in the advertisement on behalf of the brand. The consumer’s willingness to
accept and internalize these claims made in the advertisement on behalf of the brand (i.e., brand cognitions) determine their attitude-toward-the-brand. What makes the Dual Mediation
Hypothesis a dual mediation is the fact that attitude-toward-the-brand is not only influenced
through brand cognitions but also directly by attitude-toward-the-ad (MacKenzie et al., 1986).
The previously mentioned The Elaboration Likelihood Model was also used in support of Goldsmith and colleagues’ claims that there is no direct effect of celebrity credibility on
attitude-toward-the-brand and consequently brand credibility. The Elaboration Likelihood Model, through integration of many different research findings, forms a general framework for organizing,
8
communications (Petty & Cacioppo, 1986). The framework establishes two relatively distinct paths to achieve persuasion communication. The main path to persuasion is the central route, which results from “a person’s careful and thoughtful consideration of the true merits of the information presented in support of an advocacy” (Petty & Cacioppo, 1986, p.125). This central
route engages in a high degree of message elaboration where the cognitions of the person receiving the message are strongly involved in determining the nature of the argument in the message’s content. The resulting attitude change of this elaborate use of cognitions will be
relatively long-lasting, resistant of change, and predictive of behavior (Cacioppo & Petty, 1984). The second path to persuasion is called the peripheral route, which occurred “as a result of some
simple cue in the persuasion context (e.g., an attractive source) that induced change without necessitating scrutiny of the true merits of the information presented” (Petty & Cacioppo, 1986, p.125). The positive or negative cues in the stimulus of persuasion received by the individual using the peripheral route are usually established irrespective of the logical quality of the stimulus as opposed to the central route. These simple cues in message content consist of elements like the attractiveness or credibility of the sources of the message, or the perceived production quality of the message (Cacioppo & Petty, 1984). As a consequence of the heuristic simplicity of the attitude changing cues the effective persuasion lasts relatively short and is more easily repressed (Petty & Cacioppo, 1986). Though this relative discrepancy between the
efficiency of persuasion cues of the peripheral route compared to the central route is reduced when the consumer regards the product or service to be of low involvement (Petty, Cacioppo & Schumann, 1983).
In terms of celebrity endorsement The Elaboration Likelihood Model assumes that the celebrity endorser effects take an indirect peripheral path toward the persuasion and attitude
9
change of an individual, because the celebrity endorser is seen as the source of the message and their characteristics comprise of factors such as attractiveness and credibility which enables peripheral processing of the message’s content (Goldsmith et. al., 2000). Therefore it can be argued that this peripheral processing also further implicates an indirect effect of celebrity endorsement on attitude change toward the brand and its credibility.
Moreover, if celebrity endorsement is considered to be contributing to attitude change under the peripheral route, it can be asserted that under the Elaboration Likelihood Model the persuasive effects on attitude change of endorsements are rather short-lived and easily suppressed and therefore can considered to be relatively inefficient (Petty & Cacioppo, 1986).
However, MacKenzie and colleagues, the researchers behind the Dual Mediation Hypothesis, addressed this assumption of relative ineffectiveness of celebrity endorsement and
subsequently disputed it. They agreed that indeed the plenary peripheral route might be less efficient, but they argued that it shouldn’t mean that celebrity endorsement is also automatically
deemed to be less efficient. According to the MacKenzie and colleagues, the consumer’s response to simplistic persuasion cue’s like celebrity endorsement might first be peripheral but
can possibly still contribute to the central route of persuasion through fostering acceptance of the endorsed message. They further state that with the Dual Mediation Hypothesis it can be argued that the central and peripheral routes are intertwined rather than substitutes (MacKenzie et. al., 1986, p. 132).
10 Online Celebrity Endorsement Effectiveness
Most of this previously mentioned research on the effectiveness of celebrity endorsement based on celebrity credibility and attitude toward the ad has been done with traditional offline media. Recently though, the focus of celebrity endorsement research has shifted toward the exploration of endorsements in the digital world, mostly on the different social media (Djafarova &
Rushworth, 2017). Although in this recent online-focused endorsement research the source credibility model and its dimensions are often mentioned, the effect of the digital environment on
the source credibility dimensions has not yet been explicitly tested.
Furthermore, the predictability of celebrity credibility on the consumer’s attitude toward
the ad in online endorsements has also not been notably researched, even though research has shown that there is good reason to be skeptical and to not assume that traditional frameworks and theories perfectly hold up in the digital environment. This is specifically the case with regard to celebrity endorsement and persuasion literatureas has been demonstrated by Karson &
Korgaonkar (2001). In their research they tested the applicability of the previously mentioned
Elaboration Likelihood Model in online advertising and whether its assumptions with regard to
central and peripheral pathing are still valid. The research concluded that in the Elaboration
Likelihood Model (ELM) involvement no longer moderates the effect of message argument on
attitude-toward-the-ad. Additionally, the consumer’s involvement also no longer moderates the
effect of peripheral persuasion cues on attitude-toward-the-brand when applied in an online
environment (Karson & Korgaonkar, 2001, p.70). Although these moderating effects of
involvement are not the core fundamentals on which the ELM is build, disproving even minor
parts of the model still demonstrate possible shortcomings the traditionally widely-accepted
11
revisiting traditional models in this relatively new digital age and not just taking them for
granted.
So to continue testing traditional models, and assess whether the application of the
traditional Source Credibility Model still is justifiable in online context, and to further expand the research of digital celebrity endorsement with regards to attitude toward the ad, the following question is raised: What is the effect of the online environment on celebrity endorsement effectiveness? Answering this question provides an explanation for the effect of the type of
medium used on the success of celebrity endorsed product promotion based on the relationship between celebrity credibility and attitude toward the ad. The question also tests how the offline success factors construed in traditional celebrity endorsement literature hold up in the rapidly expanding digital environment, namely being the source credibility dimensions. This question will be answered by testing 6 hypotheses based on the conceptual model presented in Figure 1.
12
Previous literature (Goldsmith et al, 2000; MacKenzie et. al., 1986) stated that celebrity credibility is an effective predictor of celebrity endorsement effectiveness through its positive influence on a consumer’s attitude toward the ad. The celebrity credibility is proven to be accurately predicted by the dimensions consisting of Attractiveness, Trustworthiness and Expertise as provide in The Source Credibility Model (Hovland et. al., 1953). So to be able to firstly analyze the factors determining celebrity endorsement effectiveness, the source credibility dimensions are used to explain the celebrity credibility and its subsequent influence on attitude toward the ad. These positive relationships between the source credibility dimensions and attitude toward the ad are summarized in the following hypotheses:
H1a: Celebrity Attractiveness has a generally positive effect on Attitude toward the Ad.
H1b: Celebrity Trustworthiness has a generally positive effect on Attitude toward the Ad.
H1c: Celebrity Expertise has a generally positive effect on Attitude toward the Ad.
However, these aforementioned positive relationships between celebrity credibility and attitude toward the ad are assumed to be identical regardless of where and how the endorsement
advertisement is presented (Djafarova & Rushworth, 2017). To assume that the type of medium of the endorsement has no effect on the validity and predictability of the traditional models used in celebrity endorsement literature is proven to be objectionable (Karson & Korgaonkar, 2001). So to further assess to what extent The Source Credibility Model is still an effective predictor of celebrity endorsement effectiveness, the moderating effect of the type of medium on the different source credibility dimensions in determining attitude toward the ad are tested by proposing the
13
H2a: Type of medium moderates the effect of Celebrity Attractiveness on Attitude toward the Ad.
H2b: Type of medium moderates the effect of Celebrity Trustworthiness on Attitude toward the Ad.
H2c: Type of medium moderates the effect of Celebrity Expertise on Attitude toward the Ad.
In summary, this study is proposed to extend the understanding of the moderating effect of the online environment on traditional celebrity endorsement literature by analyzing the difference in perceptions of the source credibility dimensions between the online and offline media through the second set of hypotheses. While empirical evidence already exists for the general positive effect of the source credibility dimensions on attitude toward the ad, the first set of hypotheses are included in Figure 1 to form a comprehensive model. Together with the second set of hypotheses this model is presumed to be fitting and able to assess the effect of the online environment on celebrity endorsement effectiveness.
14 METHODS
Research Design
The research took a deductive and cross-sectional approach to test what the effect of the online environment is on the source credibility dimensions through the use of a survey. Due to the time horizon of this research a cross sectional survey was deemed most appropriate as it still allowed for an accurate description of a phenomenon or incidence (Saunders, Lewis & Thornhill, 2016, p. 200). The research compared the difference in perceived credibility of a celebrity and the
subsequent attitude toward that ad between the traditional media (i.e. magazine) and the online media (i.e. Instagram). This difference between media types was analyzed for two different celebrities. The deductive research was conducted by testing the 6 hypotheses through the use of statistical analyses for both celebrities. The statistical analyses were conducted based on data gathered through anonymous online questionnaires. This mono-method quantitative data collection method consists of an unorthodox type of subject design. In this between-subject design one group of participants were part of the control group (traditional media) for Kendall Jenner and the treatment group (Instagram) for the Lionel Messi, while the other group of participants were part of the control group for Lionel Messi and the treatment group for Kendall Jenner. This type of research design was chosen because it facilitated observation of the difference between the perceived credibility and subsequent attitude toward the ad of the
celebrities depending on the type of medium used, while also allowing for a comparison between the two celebrities’ credibility characteristics regardless of media type.
These two celebrities have been chosen, because they both advertised and endorsed a well-known global brand: Adidas (Nr. 55 of Best Global Brands; Interbrand, 2017) while both having completely different claims to fame. Kendall Jenner being a model and a reality-TV star
15
and Lionel Messi is regarded as being one of the best football players in the world. Having two completely different celebrities promoting the same brand to compare the effect of the online environment on their credibility allowed for another type of controlling condition for the effect of the type of medium in attempt to further increase the validity and generalizability of the research findings.
Participants
The participants were recruited based on non-probability sampling of convenience and voluntary self-selection to be able to achieve an adequate sample size. An anonymous link of the online questionnaire with a message inquiring voluntary cooperation was distributed on different social media and further shared in those platforms by acquaintances and relatives to a greater audience and mainly consisted of adults. A message with the link inquired for voluntary cooperation of relatives and acquaintances and therefore consisted of convenient voluntary self-selection (Saunders et al., 2016, p.303). This type of sampling technique can create homogenous samples and therefore be a threat to the validity of the research due to the biases that come with this homogeneity of voluntary self-selection. Participants are most likely to participate if they already have a certain disposition to the subject creating a bias in the sample (Saunders et al., 2016). Participants are also more likely to contact other potential participants who share similar characteristics with them, resulting in a homogenous sample (Saunders et al., 2016).
However, due to the time horizon and research objective this type of sampling was deemed most appropriate and resulted in a total of 146 participants of which 103 (70.55%) fully completed the survey.A total of 66 male (64.7%) and 37 female (35.3%) participants with an age
16
between 18 and 64 with an average age of 33 for males and 29 for females, contributed to the study. The minimum age included was 18 because the shown brand advertisements were targeted at an adult audience and it was estimated that the celebrities would be better known under adults and as such negate possibility for random guessing to enhance validity.
The participants’ pre-existing perceived image of celebrities based on their previous experiences and knowledge regarding the celebrity was crucial for evaluating the more thoughtful source credibility dimensions such as trustworthiness and expertise. Therefore it was necessary for participants to know the celebrities prior to the research to form this image. So to maintain validity of the questionnaire a manipulation check was conducted similar to Choi & Rifon (2012), where participants were asked to correctly identify the celebrity in a picture before continuing with the actual questionnaire. Of the 103 participants 35(33.98%) did not correctly identify Kendall Jenner and were therefore excluded from the analyzed data to preserve validity of the research. Only 2 participants misidentified Lionel Messi and were also excluded. So in total data of 68 participants was used for the analysis of Kendall Jenner’s advertisements and the data of 101 participants were used for the analysis on Lionel Messi’s advertisements.
To obtain a better general picture of the relevant characteristics of the participants in our sample a number of control questions were asked regarding social media. On average, 39.8% of the participants spend 1-2 hours a day on social media, 23.7% spend less than 1 hour and 24.7% spend 3-4 hours a day on social media. The majority of the participants (54.3%) considered themselves to be somewhat competent on social media. 71.3% of participants liked social media to a certain extent and 16% of the participants disliked social media. Of the 103 participants only 4.7% did not personally have a social media account.
17
Furthermore, to control for a possible predisposition to like the ads or celebrities more because of personal interest in sports or fashion, control question were asked to evaluate the participant’s interest in these subjects. Of the participants 6.9% indicated to not have any interest
in fashion, whereas in regards to sports 2.9% indicated to have no interest at all. 31.4% of participants are very interested in fashion and for sports 49% considered themselves to be very interested.
Measures
To measure the effect of the online environment on the relationship between the source
credibility dimensions and the attitude toward the ad, five distinct variables are created. First of all, the three independent variables predicting the source credibility (Hovland et al., 1953) consist of: Attractiveness, Trustworthiness and Expertise. Second, the dependent variable Attitude toward the Ad was used as constructed by Sirgy (1985), to assess the actual consequences of the celebrity
credibility on the advertisement’s effectiveness. Finally, the moderating variable Type of Medium was constructed to assess the effect of the online environment on the relationship between de independent variables and the dependent variable, compared to the offline environment. Because in the survey the Type of Medium was realized as a condition rather than a typical scale, a dummy variable was created for each type of medium and their respective data with regard to the
perceived source credibility dimensions.
The variables of the source credibility dimensions were independently measured as Attractiveness, Trustworthiness and Expertise. For each characteristic 5 five descriptive items,
18
similarly to Goldsmith and colleagues (2000), a single variable measuring endorser credibility was also added by combining the three endorser credibility dimension subscales as constructed by Ohanian (1990) into a new variable: Credibility. This variable was created to be able to check for any significant relationship of the total credibility ignoring the individual dimensions.
The basic scales for attitude-toward-the-ad as used in traditional literature (Sirgy, 1985) consisted of 3 items measuring; Good, Favorable and Pleasant perceptions of advertisements were summed into a new variable consisting of the average total score named Attitude toward the ad. All constructed variables, besides the dummy variable Type of Medium, were assessed on a
7-point Likert scales and construed as: strongly agree = 1 … strongly disagree = 7, meaning that a high score (e.g. 7) indicates a worse perception of Attractiveness for example.
To be able to test for moderation of the relationship between the source credibility dimensions and the attitude toward the ad, interaction variables were also created for the interaction between the dummy variable of the type of media and the centered means of the different source credibility dimensions: Attractiveness*Type of Medium, Trustworthiness*Type of Medium and Expertise *Type of Medium.
As previously mentioned, control items were also created to check for any unexpected outstanding characteristics of the sample that might inadvertently influence the validity of the results. The control variables consisted of basic demographics: age and gender. With the regards to social media characteristics the following control variables were created: social media
platforms used, social media time spend, social media competence and social media opinion. And also to control for a pre-existing bias for the ad or the celebrity: interest in sport and fashion.
19
Conventionally, every scale was tested for reliability across both celebrities and respective medium type of advertisement. The subscales for each source credibility dimension and the basic scale for attitude-toward-the-ad were analyzed for reliability. As expected of previous research, the source credibility scales were all proven to be of good reliability for both celebrities and for each type of media. For Kendall Jenner across both types of media the constructs of
Attractiveness (α = .7661; α = .8432), Trustworthiness (α = .8171; α = .8602) and Expertise (α = .8221; α =
.8702) were measured for reliability. For Lionel Messi the same dimensions across both types of
media were also measured: Attractiveness (α = .8521; α = .8572), Trustworthiness (α = .7721; α = .8032)
and Expertise (α = .8511; α = .8332).
So in summary, every source credibility construct was concluded to be of good reliability except for Kendall Jenner Instagram Attractiveness (Cronbach’s α = .766) and Lionel Messi Instagram Trustworthiness (Cronbach’s α =.772), which were still reliable enough to use (Cronbach’s α > .700). The basic 3 item scale of Attitude toward the Ad had on average an even higher reliability (Cronbach’s α =.884) across both types of media and celebrities. For Kendall
Jenner the 3 items making up Attitude toward the Ad were measured for both media types and proven to be of good reliability ( α = .8571; α = .8772). Similarly Lionel Messi’s the Attitude toward
the Ad construct was tested and also proven to be of high (α = .9341; α = .8692). The new summed
variable Credibility was also tested for reliability and proven to be of good reliability across all conditions and groups (Cronbach’s α > .800) as seen in Tables 1-4 (p. 25-26).
20 Procedure
The data on the participants in the survey study was gathered through the use of an anonymous online distributed questionnaire (see Appendix). The questionnaires for each of the two randomly selected groups consisted of 2 parts each. The first part consists of items measuring the source credibility of Kendall Jenner and the participant’s attitude toward that ad. The second part
consists of items measuring the source credibility of Lionel Messi and the subsequent attitude toward the ad. Group 1 first answered questions regarding de perceived credibility of Kendall Jenner while being exposed to her Adidas Instagram Endorsement Ad, while Group 2 answered the same source credibility questions while being exposed to an Adidas Magazine Advertisement Kendall Jenner. In the second part of the survey regarding Lionel Messi, Group 1 answered questions regarding an Adidas magazine advertisement of Lionel Messi, while Group 2 answered questions on the Adidas Instagram endorsement by Lionel Messi.
Prior to answering the source credibility questions regarding the advertisements involving Kendall Jenner both groups had to answer a screening question similar to the study of Choi and Rifon (2012), where the participants were asked to correctly identify the celebrity in the
advertisement. The source credibility questions were based around measuring the three core dimensions of the source credibility model: Attractiveness, Trustworthiness and Expertise. Each of these three dimensions were measured based on five descriptive questions that evaluated comparable characteristics of Kendall Jenner regarding the same dimension as constructed by Ohanian (1990). The participants were asked to what extent they agree with the statement
regarding the characteristic of Kendall Jenner. So for example, participants had to answer to what extent they agree, based on a 7-point Likert scale (strongly agree = 1 .., strongly disagree = 7), with the statement that “Kendall Jenner is attractive”. Five items stating that Kendall Jenner is;
21
“attractive”, “classy”, “beautiful”, “elegant”, and “sexy” measured attractiveness; five statements consisting of “sincere” “dependable” “honest” “reliable” and “trustworthy” assessed
trustworthiness; and five items based on “expert” “experienced” “knowledgeable” “qualified” and “skilled” were used to measure expertise. So the extent to which the participant agreed with
the five statement together evaluated the perceived Attractiveness, Trustworthiness and Expertise. The second part of the questions revolved around the Attitude toward the Ad of the different media types of Adidas advertisements with Kendall Jenner. The participants were asked to rate their Attitude toward the Ad based on the three statements where again the extent to which the participant agrees with the statement is measured. The three items based on the basic Attitude-toward-the-Ad scale by Sirgy (1985) stated that the advertisement was “good”, “pleasant” and “favorable”. Together agreeableness of these items measured the Attitude toward the Ad of the
participant.
As previously mentioned, the second part of the questions evaluated the source credibility and subsequent attitude toward the ad of Adidas advertisements involving Lionel Messi for both types of media. Group 1, which first answered questions about the Instagram endorsement of Kendall Jenner, now answered questions regarding an Adidas magazine advertisement of Lionel Messi. Group 2, which first answered questions about the magazine advertisement involving Kendall Jenner, now answered questions on the Adidas Instagram endorsement by Lionel Messi. The same items as stated for Kendall Jenner’s advertisements were used to measure the
Attractiveness, Trustworthiness and Expertise, for Lionel Messi for both types of media.
Subsequently, both groups were asked about their attitude toward either of the shown advertisements of Lionel Messi.
22
Lastly, after the main questions about the credibility of the celebrities and the attitude toward the ad, the participants were asked to answer the control questions regarding: age, gender, which social media the participant uses, social media time spend, social media competency, how much they like social media and their level of interest in sports and fashion.
Data Analysis
Before analyzing the relationships between the independent and dependent variables data initial descriptive analysis was done to check for reliability of scales and for possible correlations as displayed in Tables 1-4. After the descriptive analysis, to be able to test: H1a: Celebrity Attractiveness has a generally positive effect on Attitude toward the Ad; H1b: Celebrity
Trustworthiness has a generally positive effect on Attitude toward the Ad; H1c: Celebrity
Expertise has a generally positive effect on Attitude toward the Ad; a simple linear regression
analysis was conducted for each source credibility dimension separately regarding their effect on the attitude toward the ad.
Furthermore, the difference in means of the celebrity credibility dimensions between the two types of media was tested. This analysis was conducted to check whether people perceive the credibility of the celebrity differently based on the type of medium. An independent samples t-test was used, because the independent sample assumptions hold true for each separate celebrity. The observations are independent because every case for each separate celebrity represents a different person. No participant evaluated the source credibility dimensions of the same celebrity twice. The participant only evaluated each celebrity once, meaning that per celebrity each
23
observation was independent. Normality of the dependent variable and homogeneity of the standard deviation were also appropriately assumed.
Finally, a multiple regression analysis was conducted with the use of the interaction variables to check for moderation to test: H2a: Type of medium moderates the effect of Celebrity Attractiveness on Attitude toward the Aa; H2b: Type of medium moderates the effect of Celebrity
Trustworthiness on Attitude toward the Ad; H2c: Type of medium moderates the effect of
Celebrity Expertise on Attitude toward the Ad. To compare for moderation the linear regression
analysis was conducted on 2 different models. The first model consisted of the separate source credibility dimensions, together with the Type of Medium as independent variables and tested for the effect on the dependent variable of Attitude toward the Ad. In the second model the
interaction variables of Attractiveness, Trustworthiness and Expertise with Type of Medium were added to test for a moderation effect. These 2 models were compared to check for possible moderation.
24 RESULTS
Firstly after collecting the data, a preliminary correlation analysis was conducted to check for possible already-existing correlations. The correlations were tested between every subscale and for every different type of medium and celebrity. The analysis revealed that there are distinct differences in correlations between celebrities and types of media. The correlations between the Attitude toward the ad and the different source credibility dimensions were significant for some
and insignificant for others indicating an important difference between each celebrity in combination with the type of media.
On Instagram only Kendall Jenner’s Attractiveness was significantly correlated with Attitude toward the Ad (r = .502, p < 0.01), while in the magazine ad Trustworthiness and total Credibility were significantly correlated with Attitude toward the Ad (r =.435, p = 0.01; r =.399, p
< 0.05). Furthermore, on Instagram all of Lionel Messi’s source credibility dimensions:
Attractiveness, Trustworthiness and Expertise and total Credibility were significantly correlated
with Attitude toward the Ad (r =.435; r =.452; r =.381; r =.660, p < 0.01)3, while in the magazine
ad nothing had a significant correlation. These distinct differences in significant correlations, as seen in Tables 1-4, gave even more reason to look into the moderating effects of media type and celebrity type on attitude toward the ad.
25
Table 1. Descriptive statistics & correlations: Kendall Jenner Instagram Reponses N Mean SD Aad Cronbach’s alpha
Gendera 30 1.467 .507 Age 30 26.555 11.043 Attractivenessb Trustworthinessb Expertiseb Credibility 30 30 30 30 2.633 3.913 3.853 3.466 .697 .803 .871 .608 .502** .021 .131 .264 (.766) (.817) (.822) (.857) Attitude towardb the ad (Aad) 30 3.422 1.128 1 (.857)
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
a 1 = male, 2 = female, 3 = neutral, b = Likert scale (strongly agree = 1 … strongly disagree = 7)
Table 2. Descriptive statistics & correlations: Kendall Jenner Magazine Reponses N Mean SD Aad Cronbach’s alpha
Gendera 37 1.487 .507 Age 37 26.750 10.960 Attractivenessb Trustworthinessb Expertiseb Credibility 37 37 37 37 2.928 4.001 3.903 3.616 .949 1.043 1.116 .824 .237 .435** .269 .399* (.843) (.860) (.870) (.886) Attitude towardb the ad (Aad) 37 3.694 1.290 1 (.877)
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
26
Table 3. Descriptive statistics & correlations: Lionel Messi Instagram Reponses N Mean SD Aad Cronbach’s alpha
Gendera 53 1.377 .489 Age 53 32.625 14.784 Attractivenessb Trustworthinessb Expertiseb Credibility 53 53 53 53 4.208 3.098 2.117 3.141 1.144 .843 .792 .646 .522** .452** .381** .660** (.852) (.772) (.851) (.818) Attitude towardb the ad (Aad) 53 3.289 1.485 1 (.934)
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
a 1 = male, 2 = female, 3 = neutral, b = Likert scale (strongly agree = 1 … strongly disagree = 7)
Table 4. Descriptive statistics & correlations: Lionel Messi Magazine Reponses N Mean SD Aad Cronbach’s alpha
Gendera 47 1.298 .462 Age 47 29.710 13.462 Attractivenessb Trustworthinessb Expertiseb Credibility 47 47 47 47 3.996 3.034 2.000 3.012 1.173 .866 .717 .636 .291 .131 .226 .282 (.857) (.803) (.833) (.813) Attitude towardb the ad (Aad) 47 2.773 .993 1 (.869)
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
27
The effect of Source Credibility dimensions on Attitude toward the ad
Firstly, a simple linear regression analysis was conducted to test the following hypotheses:
H1a: Celebrity Attractiveness has a generally positive effect on Attitude toward the Ad.
H1b: Celebrity Trustworthiness has a generally positive effect on Attitude toward the Ad.
H1c: Celebrity Expertise has a generally positive effect on Attitude toward the Ad.
The separate source credibility dimensions combined, as proposed in the hypotheses, explained for 22.2% (adjust. R2 = .222) of variability of Attitude toward the Ad in the proposed model (Table 5). The regression analysis concluded that Attractiveness (β = .325, p < 0.001) and Expertise (β = .346, p < 0.001) had a significant positive effect on Attitude toward the Ad. The effect of
Trustworthiness on Attitude toward the Ad (β = .142, p = 0.123) was shown to be insignificant regardless
of media type or celebrity as shown in Table 5.
Table 5. Regression statistics of Source Credibility Dimensions on Attitude toward the Ad
B SE B β t p (Constant) Attractiveness Trustworthiness Expertise .396 .341 .186 .359 .428 .079 .120 .101 .325 .142 .346 .925 4.320 1.528 3.566 .356 .000 .123 .000
Dependent variable: Attitude toward the Ad Adjusted R2 = .222
28
The Source Credibility dimensions differences between Type of Medium
Prior to the moderating regression analyses an independent samples t-test was conducted. An independent samples t-test was conducted because for each separate celebrity each observation is made by a different participant and therefore independent of each other, as has previously been further explained above. This test was conducted to check whether the source credibility dimensions of the celebrity were perceived significantly differently between the two types of media. Whether there is a significant difference, in perceived source credibility dimensions of a celebrity between the types of media, is important because that finding could also provide an explanation for the possible difference in perceived Attitude toward the Ad between the two types of media. So this test can likely help prevent false positives (type 1 error) for the moderation analysis, because it removes another possibility that could have also caused the difference in Attitude toward the Ad.
The t-test was done for each source credibility dimensions independently and also for the 3 variables together under one combined variable: Credibility. Separate tests were conducted for each celebrity comparing their mean Attractiveness, Trustworthiness and Expertise between the two types of media. Results from the independent samples t-test indicated that for both celebrities the means of the source credibility dimensions did not significantly differ from each other
between the two types of media (t(67) < 1.420, p>0.05, two-tailed). Moreover, for Lionel Messi the combined variable Credibility did not have a significant difference in means between the two types of media (t(100)=0.827, p= 0.412). Neither for Kendall Jenner was there a significant difference in means of total Credibility between Instagram and the magazine (t(67)=0.854, p= 0.397). This means that there are no significant differences in participants’ perceptions of the credibility of the celebrities between the two types of media. So the type of medium has no
29
significant effect on the evaluation of the source credibility dimensions alone for the different celebrities. However, this is without taking Attitude toward the Ad and its relationship with the source credibility dimensions into account.
The moderating effect of Type of Medium
Finally, the following moderation hypotheses will be tested:
H2a: Type of medium moderates the effect of Celebrity Attractiveness on Attitude toward the Ad.
H2b: Type of medium moderates the effect of Celebrity Trustworthiness on Attitude toward the Ad.
H2c: Type of medium moderates the effect of Celebrity Expertise on Attitude toward the Ad.
For these moderations to be able to exist there should be an initial direct effect of the source credibility dimensions on Attitude toward the Ad and there should be no direct effect of Type of Medium on Attitude toward the Ad. Consequently, preceding the actual testing for moderation of
the relationship between the source credibility dimensions and Attitude toward the Ad, the direct effects of the independent source credibility dimensions on Attitude toward the Ad and the direct effect of Type of Medium on Attitude toward the Ad were analyzed. As previously mentioned and also illustrated in Table 5, there is a direct effect of Attractiveness and Expertise on Attitude toward the Ad. Furthermore, the direct effect of Type of Medium on Attitude toward the Ad is insignificant (β = 0.062, p = 0.428). So after the initial requirements for moderation were
30
dimension in combination with Type of Medium. To account for possible moderation the linear regression analysis was based on 2 different models as shown in Table 6. The first model consisted of the separate source credibility dimensions together with the Type of Medium as independent variables and tested against the dependent variable of Attitude toward the Ad. The first model explained 22.2% (adjust. R2 = .222) of variability of Attitude toward the Ad. In the model only Attractiveness and Expertise had a significant effect on the dependent variable (β = .322; β = .346, ps < 0.001). In the second model the interaction variables of Attractiveness,
Trustworthiness and Expertise with Type of Medium were added to test for a moderation effect.
The addition of these interaction variables significantly increased the strength of the model with a 4.9% increase in variability explanation (Change in R2 =.049, p < 0.05). In total the second model accounted for 25.8% (adjust. R2 = .258) of the variability in Attitude toward the Ad. In the second model, out of the separate source credibility dimensions only Expertise still has a significant effect (β = .322, p < 0.05) on Attitude toward the Ad albeit less significant compared to model 1 (Table 6). Of the added interaction variables, the interaction between Attractiveness and Type of Medium had a significant effect on Attitude toward the Ad (β = .344, p < 0.01). This significance
of the interaction variable indicates a moderation effect of Type of Medium on the relationship between Attractiveness and Attitude toward the Ad as plotted in Figure 2. No other significant moderating effect of Type of Medium on the source credibility dimensions has been proven with the regression analyses even when controlling for the type of celebrity (p>0.05, two-tailed).This finding indicates that, of all three source credibility dimensions, only the effect of the
attractiveness of a celebrity on the attitude toward the advertisement is moderated by the type of medium where the endorsement advertisement is used. In summary, in the online environment higher celebrity attractiveness improves the attitude toward that ad more so then in the traditional media. Celebrity expertise is still a significant predictor of attitude toward the ad but is not
31
moderated by the type of medium as opposed to celebrity attractiveness. Celebrity
trustworthiness has an insignificant effect on attitude toward the ad regardless of media type.
Table 6. Regression statistics of Source Credibility Dimensions on Attitude toward the Ad
Model 1 Model 2 B SE B β B SE B β (Constant) Attractivenessa Trustworthinessa Expertisea 3.175*** .338*** .189 .359*** .124 .079 .120 .101 .322*** .145 .346*** 3.161*** .081 .238 .299* .121 .111 .157 .130 .077 .142 .346* Type of Medium .167 .176 .065 .170 .172 .066 ATT*MEDIA .501** .155 .344** TRST*MEDIA -.048 .237 -.024 XPT*MEDIA .131 .199 .084 R2 .240 .2901 Adjusted R2 .222 .258
Dependent variable: Attitude toward the Ad
*** Regression is significant at the 0.001 level (2-tailed), ** Regression is significant at the 0.01 level (2-tailed), * Regression is significant at the 0.05 level (2-tailed).
1R2 change (.049) is significant at the 0.05 level (2-tailed).
a Independent variables Attractiveness, Trustworthiness & Expertise are mean centered.
32 DISCUSSION
H1a, b & c proposed that the Attractiveness, Trustworthiness and Expertise of the celebrity all
have a significant positive effect on the Attitude toward the Ad. The results from this research suggest that out of three source credibility dimensions, only Attractiveness and Expertise have a significant positive effect on Attitude toward the Ad. Therefore, H1a and H1c are supported. The finding that the perceived attractiveness and expertise of the celebrity used in endorsement advertising positively contributes to the consumer’s attitude toward the ad, is commonly accepted
and in line with traditional celebrity endorsement literature (Choi & Rifon, 2012; Goldsmith et. al., 1999). In this previous literature the general consensus is that all three of the source
credibility dimensions (Hovland et al., 1953) have a positive effect on the consumer’s attitude toward the ad, including the celebrity’s perceived trustworthiness.
However, the results from the analysis in this research did not report a significant positive effect of Trustworthiness on Attitude toward the Ad and as such H1b is rejected. In this study the trustworthiness of the celebrities did not significantly contribute to the overall attitude toward the ad of the participants. There are several possible explanation for this finding. When considering the Elaboration Likelihood Model, celebrity trustworthiness might not have a significant
influence on the attitude toward the ad, because in the model celebrity endorsement is considered to take a peripheral route to persuasion. In this peripheral route superficial cues influence the ultimate attitude toward the ad without necessitating scrutiny of the trustworthiness of the advertisement (Petty & Cacioppo, 1986, p.125). Thus it can be argued that the perceived
trustworthiness of the celebrity in the endorsement advertisement is not necessary for persuasion and significant attitude change, which can consequently explain the insignificant effect of Trustworthiness on Attitude toward the Ad.
33
Furthermore, trustworthiness might not have a significant influence on the attitude towards the ad even though most literature suggests it, because it can be argued that the purchases associated with the Adidas brand are generally of lower involvement and therefore do not necessarily require a strong trusting relationship between the brand and the consumer. Furthermore, when people purchase sportswear they mainly care about the look and the functionality of the product (Lim, Kim, Cheong, 2016). Through the proven importance of the alignment of the celebrity and the advertised product, the Celebrity/Product Congruence added another dimension to the celebrity credibility model to further explain the subsequent attitude toward the endorsement advertisement (Choi & Rifon, 2012). Further drawing on this consolidation between the celebrity and the product, it can be argued that, in terms of product characteristics, look and functionality of the product coincide with the celebrity’s Attractiveness and Expertise and are therefore more important in celebrity endorsement sportswear advertisements compared to Trustworthiness. An explanation for this discrepancy between the findings of this research and the traditional celebrity endorsement literature is that previous studies on the effect of the source credibility dimensions on the attitude toward the ad mainly used electronics brand advertisements in their analyses, which are typically regarded as higher involvement purchases and as such make trustworthiness more important (Choi & Rifon, 2012; Goldsmith et. al., 1999).
H2a, b & c proposed that the type of medium used for the endorsement advertisement
moderated the effect of the perceived celebrity Attractiveness, Trustworthiness and Expertise on Attitude toward the Ad. The results from the regression analyses indicate that out of the three
source credibility dimensions, only the effect of celebrity Attractiveness on Attitude toward the Ad is moderated by the type of medium used and therefore H2a is accepted. This means that,
34
positive effect on the endorsement advertisement. Specifically, as seen in Figure 2, higher celebrity endorser attractiveness leads to a greater increase in positive attitude towards the
Instagram ad compared to the printed magazine advertisement. However, this difference between the two types of media regarding the effect of celebrity endorser characteristics on the attitude toward the advertisement is not evident for all source credibility dimensions. The results do not provide sufficient support for the moderating effect of type of medium on celebrity
trustworthiness and expertise in determining the attitude toward the ad and therefore H2b & c are rejected. Although celebrity expertise is not moderated by the type of medium it still has a
significant predicting effect on attitude as opposed to celebrity trustworthiness.
A possible explanation for these findings lies in the nature of the online environment and in this research particularly of social media. The participants spend on average 2 to 3 hours a day on social media, 70% of them said they liked social media and the majority considered
themselves to be competent on the different social media platforms. This high social media compatibility of the participants provides an explanation, because avid social media users adopt a certain behavior as a result of the inherent need to keep up with the information overload on today’s social media. When people use social media such as Instagram, they typically tend to quickly scroll through each post and make a fast decision whether they “like” or “dislike” it on
the spot, without taking the time to really consider the more intricate perceptions of the image due to time constraints and the overwhelming amount of information. Consequently, the more superficial characteristics such as attractiveness are ultimately more decisive in the forming of an attitude toward that post (Gomez-Rodriguez, Gummadi & Schoelkopf, 2014). This also coincides with the idea that in the online environment a more peripheral route to persuasion is present, where superficial cues such as the attractiveness influence the ultimate attitude toward the ad
35
without necessitating scrutiny of the trustworthiness of the information presented (Petty & Cacioppo, 1986, p.125).
As such it can be theorized that celebrity trustworthiness is less important on social media and only celebrity attractiveness is significantly moderated by the type of medium. On the
contrary, when reading magazines, people generally spend more time thoroughly reading the articles and developing more intricate perceptions of the images which might explain why celebrity attractiveness, trustworthiness and expertise were indeed all significantly correlated with the attitude towards the magazine advertisement. This can provide a possible explanation for the findings with regard to the second set of hypotheses.
Especially this second set of hypotheses and the acceptation of H2a add to the current celebrity endorsement literature, because the effect of the online environment on the celebrity endorsement effectiveness, more specifically on celebrity credibility, has not notably been researched as of yet. The findings of this study contribute to current theorization of the source credibility dimensions in the online environment that, higher celebrity endorser attractiveness leads to a greater increase in positive attitude towards the online ad compared to the offline advertisement. Furthermore, the significant effect of celebrity expertise on the attitude toward the ad is not substantially different between the two types of media and thus not moderated, as opposed to celebrity attractiveness. Celebrity trustworthiness has an insignificant effect on attitude toward the ad regardless of media type. These unexpected results regarding the source credibility model also further justify the skepticism of Karson and Korgaonkar (2001) regarding the unquestioned application of the traditional celebrity endorsement models in the new online
36
With regard to the original research question; What is the effect of the online environment on celebrity endorsement effectiveness?, the current research provides an answer by determining that
the effect of the online environment on celebrity endorsement effectiveness is that, higher
celebrity endorser attractiveness leads to a greater increase in positive attitude towards the online ad compared to the offline ad. Within this answer an important implication is made, namely that the attitude toward the endorsement advertisement determines its ultimate effectiveness. This implication is made based on the fundamental theories of celebrity endorsement literature (Goldsmith et. al., 2000).
A limitation to the generalizability of the findings of this study is the difference between each individual celebrity. Out of the whole variety of celebrities, only two have been evaluated in this study. Choi & Rifon (2012) also recognized this limitation in the field of celebrity
endorsement studies and its negative consequence for the generalizability of the results, since every celebrity is indeed unique. However, they suggested that diverse celebrities from different fields with a distinctly different claims to fame should be used to minimize this negative effect on the generalizability. Therefore in this study a female reality-TV star and model (Kendall Jenner) and a professional football player (Lionel Messi) were used. Although a thing to note is that, Lionel Messi, being a professional sports player, is inherently more connected to the sports fashion brand of Adidas compared to Kendall Jenner. This inherent connection could have influenced the results of the study as the perceived Celebrity/Product Congruence (Choi & Rifon, 2012) subsequently influences the attitude toward the ad explaining the on average better attitude toward both ads involving Lionel Messi. The generalizability of the findings is also limited by the usage of just one brand (Adidas) and its industry (sports fashion). Moreover, the generalizability
37
of the findings for the population is limited by the usage of a voluntary convenience sample from a single geographical region.
A limitation of the validity of this study is that the forming of the attitude toward the ad cannot be fully contributed to the celebrity. The difference in attitudes between the ads might not be fully contributable to the celebrities and their (credibility) characteristics, but also to the advertisement message surrounding the celebrity. An ad might be a more universally pleasing picture or better fulfilling its message or advertisement purpose and thus be generally more effective regardless of the celebrity endorser used in the advertisement. Another limitation to the validity might be that the magazine advertisement measuring offline media was shown in an online survey instead of a printed version, a struggle also noted in previous endorsement research (Goldsmith, 2000). The study as a whole including the Instagram ad would benefit from a more naturalistic setting to observe the real time perceived reactions to the different celebrities and endorsement ads instead of a questionnaire.
A final limitation could be the usage of the attitude toward the ad scale for an Instagram endorsement. The scale (Sirgy, 1985) was originally designed to measure the attitude toward traditional advertisements. This provides a possible explanation for the, on average, better
attitude measured for the magazine advertisements compared to the Instagram ads. The measured attitude toward the ad could be generally better for the magazine, because the scale is designed for evaluating actual advertisements and Instagram endorsement might be less obvious
advertisements and therefore the 3 item scales used might not accurately reflect the attitude toward those ads.
