The effect of celebrity endorsers on consumers'willingness to pay

(1)

MSc. Business Studies – Entrepreneurship & Management in Creative Industries University of Amsterdam – Faculty of Economics & Business

Final version of MSc. Thesis

Sagitha Xenia Biekman – 10675043 June 30th 2014

First Supervisor B. Kuijken Second Supervisor N. Wijnberg

University of Amsterdam

The Effect of Celebrity

Endorsers on

Consumers’

(2)

Abstract

The leading question for this study is to find the effect of celebrity endorsers on consumers’ willingness to pay. Subsequently, the moderation effects of the image of the celebrity and the credibility of the celebrity on the effects of the celebrity endorser on

consumers’ willingness were investigated. The expert and non-expert celebrity endorsers and their different levels of involvement both had a significant effect on consumers’ willingness pay. The image of the expert celebrity had a very big impact on the effect of the celebrity endorser on consumers’ willingness to pay. However, this moderation effect was marginally significant. The moderation effect of neither the image of the non-expert celebrity, nor the credibility of both the expert and non-expert celebrity were significant.

(4)

1.0 Introduction

Many large brands use celebrities to increase appeal among potential consumers (Cronley et al., 1999; Kalra & Goodstein, 1998; Silvera & Austad, 2003; Agrawal & Kamakura, 1995; Kamins et al., 1989; Roy, 2006; Klaus & Bailey, 2008). To reach this goal, companies often go through great financial lengths closing celebrity endorsement deals, of which the economic benefits still have to be proven. Through closing deals with celebrities, celebrities are used as spokesperson, ambassador or as a mere feature in advertisements for the endorsed product or service. The brands that engage in endorsement deals are often established and international brands who want to broaden their customer base or new upcoming brands that still have to find out how to reach their target group, which can be the same group as the audience that the endorsing celebrity has as its fan base (Agrawal & Kamakura, 1995; Cronley et al, 1999; Doss, 2011; Erdogan, 1999; Roy, 2006). Previous research has shown that these celebrity endorsement deals can be very lucrative for the companies (Agrawal & Kamakura, 1995). In these studies the positive effects of the celebrity endorsements were supported by measures of increased revenues for the companies, increased performance of the stock markets and increased consumer attitude towards the brand (Agrawal & Kamakura, 1995; Cronley et al., 1999). However, in practice not all celebrity endorsements had these effects. Thus, not all celebrity endorsement deals guarantee a success for companies, as a consequence of a bad fit between the celebrity and the endorsed product an undesired effect can emerge from the endorsement (Klaus & Bailey, 2008). In their study, Klaus & Bailey (2008) show the undesired effects that could emerge when there is a bad fit between a celebrity and the endorsed product and the target audience. There has to be a contingency fit between the brand, the celebrity and the perception of the audience on the celebrity in order for the celebrity endorsement deal to be a success. One of the results is that

(5)

something as the gender of the celebrity endorser can have a desired or undesired impact based on the gender of the target audience (Klaus & Bailey, 2008).

As discussed before, this fit between celebrity and the endorsed brand is crucial for the success of an endorsement for the company, however the effect of the involvement of the celebrity has been neglected. This study attempts to explore the relation between the

celebrity/product fit and consumers’ willingness to pay. The fit between a celebrity and product will be examined through a celebrity’s expertise on the product field of the endorsed product. A celebrity with a perceived low expertise on the product field, i.e. there is a lack of overlay

between the celebrity’s current or previous profession and the product field, will be referred to as a non-expert. On the other hand, a celebrity with a perceived high expertise on the product field, which means that there is an overlap between the celebrity’s current or previous profession and the product field, will be referred to as an expert.

Celebrity endorsements deals can vary on the level of involvement of the celebrity endorser. As said before some celebrities function as a spokesperson for a charity and are actively involved with marketing the product, whereas others are merely appearing in the advertisement. Which raises the question whether just recommending a product would have a different effect on the consumers’ attitude, i.e. willingness to pay, compared the celebrity actively being part of the product or service.

This thesis attempts to bridge the concepts of celebrity endorsements and consumers’ willingness to pay. The theoretical contribution of this research is that it deepens the

understanding of the concept celebrity endorsements and most importantly will provide new insights on the effects of celebrity endorsement and involvement level in the endorsed products.

(6)

The practical contribution is that the research could provide evidence for possible positive effects of celebrity endorsements and their involvement with the products. If linking a celebrity with increased involvement to a product will increase consumers’ willingness to pay opposed to a celebrity with low involvement or no celebrity endorser at all, managers are able to have a better insight on the decision to whether or not enter an endorsement deal and whether or not it would be more favorable to have a celebrity who is highly involved in the product field. Thus, the research question that will be the guideline of the thesis is as follows:

What is the effect of celebrity endorsers on consumers’ willingness to pay?

The research proposal begins with a review of the current literature, where after the hypotheses and the theoretical framework are presented. Subsequently the methodology and the results of the experiment are explained. Finally, the discussion, conclusion, implications and

recommendations for future research are given.

(7)

2.0 Literature Review

2.1 Willingness to Pay

In economics willingness to pay has been a widely discussed topic. It describes the maximum amount of money someone is willing to pay, trade or sacrifice in order to obtain a product or avoid a displeasing product (Shogren, Shin, Hayes, & Kliebenstein, 1994).

Willingness to pay is often discussed in combination with the concept of willingness to accept, which refers to an amount someone is willing to gain in order to sell a product or experience something displeasing (Shogren, Shin, Hayes, & Kliebenstein, 1994). Willingness to pay

measuring methods can be divided in hypothetical and actual willingness to pay. In the latter the participants are often incentivized to actually purchase the product. Someone’s willingness to pay is dependent on one’s wealth and on the desire to own the product. In this thesis someone’s valuation of a product will be reflected through the amount of money somebody is willing to pay for the endorsed product. Hence, in this study the willingness to pay in the form of bid amounts at the auction, will give an insight on how the consumers valuate certain celebrity endorsers.

2.2 Celebrity endorsements

Celebrity endorsements have become an increased topic of discussion in the marketing and advertising field. A celebrity endorsement is a testimonial for a product, appearance at PR event, being a spokesperson or other form of engagement of a public figure in an advertising campaign in order to exploit their status in society to promote a product, service or charity goal (Erdogan, 1999). The use of celebrities in advertising campaigns has been very beneficial for organizations (Erdogan, 1999; Silvera & Austad, 2003; Agrawal & Kamakura, 1995; Kamins et al., 1989; Roy, 2006; Cronley et al., 1999; Doss, 2011; Keller, 2013). Studies have shown that the use of a celebrity endorser can increase the attitude of consumers towards the endorsed product, given that the consumer has a positive attitude towards the celebrity endorser (Silvera &

(8)

Austad, 2003; Erdogan, 1999; Keller 2013). In their study Silvera & Austad (2003) found that if there are increased assumptions that a celebrity actually likes the product, and is not only driven by for example monetary gains, consumers are more likely to adopt a favorable attitude towards the endorsed product. Based on the Silvera & Austad’s study (2003) the first hypothesis is:

H1: The use of a celebrity endorser has a positive effect on consumer’s willingness to pay

2.2.1 Level of involvement of Celebrity Endorsers

The relation between a celebrity’s expertise and the consumers’ attitude towards the endorsed relation discussed by Silvera & Austad (2003) can be deepened by introducing the involvement of the celebrity in the campaign besides its expertise. Focusing on the passive or active role of the celebrity endorser in the marketing campaign of the endorsed product can possibly have an impact on consumers’ attitude towards the product as well. This

correspondence inference is subject to the perspective of a celebrity’s expertise on the product field and their style preferences (Silvera & Austad, 2003). However, there was no distinction made between the different levels of involvement and expertise in the study of Silvera & Austad (2003). Their research has shown that people will have a more positive attitude towards the product when they perceive the celebrity as an expert on the product field, but did not examine the difference between an expert celebrity just giving a testimonial or actively participating in the production of the product or service. The same goes for a non-expert celebrity making a

commercial more appealing or functioning as a spokesperson. Where studies have shown that expertise can have a positive effect on consumers’ attitude, adding high involvement in the marketing campaign could possible increase this effect (Silvera & Austad, 2003). Translating someone’s attitude towards a product into someone’s willingness to pay for this product could mean in practice that someone would actually be willing to pay more for a product endorsed by a

(9)

celebrity who is actively involved in the promotion campaign or even in the production of the endorsed product or service.

In their study, Silvera & Austad (2001) argued that perceived expertise could have a positive effect on consumers’ attitude. Another result of their study was that if there is a

discrepancy between the celebrity endorser and the endorsed product, thus, the celebrity endorser was considered as a non-expert, the consumer attitude was not affected as much (Silvera & Austad, 2001). As we expect that high involvement of the celebrity can have an increasing effect on the already established relation between the celebrity expertise and the consumers’ attitude, it is expected that this will also be the same for non-experts celebrities. Thus, this would entail that by increasing their involvement in the promotion campaign or even the production process and subsequently sending the message that they’re not just there for the monetary gains, this would also have a positive effect on the consumers’ attitude even though the initial relationship would not be as strong. This would overcome the discrepancy found by Silvera & Austad (2001). Hence, the third hypothesis is:

H2: The level of involvement of a celebrity endorser has a positive effect on consumers’ willingness to pay.

2.2.2. Celebrity Endorser Image

Whereas the relation between the endorsed product and the celebrity endorser is key concerning the credibility of the endorser, this is not the case with the image of the celebrity endorser. The image of the celebrity endorser is constructed in the minds of individuals based on what they hear, see and read about the celebrity (Balmer, 2001). Just like a brand, celebrities uphold a

(10)

would like to be known for and with what they would like to be associated with in order to influence the perception of them by the external environment. However, as the message that the celebrity sends out to the external environment can be carefully monitored and managed, the perception of the external stakeholders is also influenced by a factor that is beyond the control of the celebrity. Just like any other brand, the image or reputation of the celebrity is affected by information about the celebrity that is being put out in the external environment due to the word of mouth (Balmer, 2001). This information is spread amongst the stakeholders and thus

represents an impact on the celebrity image that is beyond the control of the celebrity.

The image of the celebrity reflects the idea that the stakeholders have concerning the celebrity, it is expected that the image of the celebrity can positively affect the relation between the celebrity endorser and the endorsed product. As it is expected that a good image will increase the effect the celebrity endorser has on the consumers’ willingness to pay and that a bad image decreases the effect that the celebrity endorser has on consumers’ willingness to pay, the following hypothesis is formed:

H3: The image of a celebrity endorser has a positive effect on the impact of a celebrity endorser on consumers’ willingness to pay

2.2.3. Celebrity Endorser Credibility

As mentioned before, companies tap into the audience of celebrity endorsers and attempt to use their public appeal for the endorsed product. Studies have shown that celebrities who are highly credible are more effective in appealing to consumers than less credible celebrities (Ohanian, 1991; Atkin & Block, 1983). Throughout the literature three main dimensions that

(11)

(Ohanian, 1991). Thus, it is argued that highly credible endorsers either have all or a

combination of these dimensions. Physical attractiveness has increased in importance over the years in endorsement deals. Research has shown that people have the tendency to make positive assumptions and create positive stereotypes about physically appealing people compared to assumption associated with unattractive communicators (Ohanian, 1991). These inferences companies attempt to capture for their endorsed products by engaging in endorsement contracts. Expertise refers to the extent to which an endorser is perceived to make valid assertions based on traits they posses that support the claims he or she makes about the endorsed product or service (Ohanian, 1991). Finally, trustworthiness is based on the consumer’s confidence that the endorser recommending the product or service is doing that from an honest and objective perspective. The belief of the endorser doing it because he or she actually believes everybody should use the product or service if important in the trustworthiness (Ohanian, 1991). Thus, based on this study the expectation is that the credibility of an endorser could increase the effect that the celebrity endorser has on consumers’ willingness to pay. Hence, the final hypothesis is:

H4: Credibility of a celebrity endorser has a positive effect on the impact of a celebrity endorser on consumer’s willingness to pay.

2.3. Theoretical Framework

Based on the hypotheses discussed in the previous section the following theoretical framework is constructed. The main reason for endorsements is the expectation of a transfer of the associations with the celebrity endorser to the endorsed product or service (Agrawal & Kamakura, 1995; Erdogan, 1999; Doss, 2011; Keller, 2013). The model shows that it is expected

(12)

that there is a positive relation between the use of a celebrity endorser and the consumer’s willingness to pay, meaning that if a company uses a celebrity endorser, consumers’ willingness to pay for the endorsed product will increase. The same is expected for the level of involvement of the celebrity endorser in the marketing campaign. The anticipation is that the more actively a celebrity endorser is involved in the marketing campaign, the more consumer’s are willing to pay for the endorsed product.

As both variables are modeled as moderators of the relation between the use of a celebrity endorser and the consumers’ willingness to pay. As said before, the image of the celebrity is subject to change and external effects. However, it is expected that if the celebrity has a good image, this will strengthen the relation between the use of the celebrity endorser and consumers’ willingness to pay for the endorsed product. In the case of a bad image, it is

anticipated that this will decrease the effect of the celebrity endorser on consumers’ willingness to pay. Thus, by modeling celebrity image as a moderator, the effect of the image of the celebrity endorser on the endorsement can be analyzed. The same goes for celebrity credibility. By

positioning celebrity credibility as a moderator, the effect of credibility on the impact of the celebrity endorser on consumers’ willingness to pay is measured. Hence, it is expected that if the celebrity is perceived as a credible endorser for the product this will increase the effect of the celebrity endorser on consumers’ willingness to pay for the endorsed product. Vice versa, it is anticipated that if the celebrity is perceived as a non-credible endorser, this will decrease the effect of the celebrity endorser on consumers’ willingness to pay.

(13)

(14)

3.0 Methodology

Now that the literature is thoroughly discussed and the hypotheses are presented, the research design and the method used for the analytical strategy will be explained. First the research design will be discussed, followed by a brief discussion of the sample and finally the analyzing strategy.

3.1. Research Design

This study is a quantitative research that uses primary source data. The data will be collected through an experiment with a Vickrey auction. This experiment will be conducted with the online auction site Veylinx. In this sealed bid auction, participants will not know what others bid for the product, however, the participant with the highest bid, will only pay the second highest price. This method will help to examine consumers’ true willingness to pay for services and products by giving them an incentive to bid their true WTP, assuring that if they win, they will pay less than their maximum price. This experiment is preferred over a research survey because during a survey people often tend to answer what they think is most idealistic, i.e. how they expect to act in a certain situation. However, when they are actually in the situation in reality, some diverge from what they say they would do. Thus, with this experiment we are able to measure the consumers’ true willingness to pay by putting them in a real situation instead of their hypothetical willingness to pay.

Before mentioned hypotheses are modeled in a 1x3 matrix. The dimension that will be measured through the use of different treatments is the level of celebrity involvement during the auction. However in order to measure the different effects of non-expert and expert celebrities the involvement dimensions will be measured for non-expert and expert celebrity endorsers for products. By using an expert and non-expert setting, both the image and credibility can be tested thoroughly. The model will be as follows:

(15)

Figure 2 – Matrix of treatments used for online auction

3.1.1. Expert Endorser Auction

For the experiment two products will be auctioned through Veylinx in order to measure the effect of celebrity involvement in and expertise on the endorsed product on consumers’ willingness to play. The expertise level of the celebrity on the product is based on their career paths and public displayed personal interests. Besides making a distinction between the expert levels of the celebrities on the product field, their level of involvement will also be taken into account. Their level of involvement will be measured through their engagement in the

advertising campaign and the production of the product. Through this manipulation of the celebrity expertise and subsequently the level of involvement the true willingness to pay will be measured through the bids of the participants in the Vickrey auction.

For this study the empirical setting of the expert celebrity endorser is the live

(16)

event is a live performance of Lipizzaner horses that are trained in the highly traditional riding school (The Spanish Riding School, 2014). Besides being the center of classical dressage in the world, the school is a tourist attraction where they offer public performances as well. The riding school distinguishes itself due to its 400 years of experience and traditions in classical dressage and keeping them intact (The Spanish Riding School, 2014). The leading riders and their stallions often go out on tours around the world. The highest performing Dutch dressage equestrian of the Netherlands, Anky van Grunsven, will endorse this event. She won the most Olympic, Gold, World, National and European medals of the Netherlands in sportsmanship (Anky, 2014). Because of her track record in dressage and awards in the dressage, she will be categorized in the experiment as the celebrity endorser of the Spanish Riding School.

The treatments will be operationalized through one advertisement of the riding school without Anky which will function as the control variable, another advertisement with Anky, the low level of involvement will be operationalized through a positive recommendation on the advertisement from Anky van Grunsven about the show and the high level of involvement will be operationalized through an advertisement stating that the show will be presented by Anky van Grunsven and with the same positive recommendation and thus, integrating her in the production of the show. See the appendixes for the advertisements of the experiment. After the auction the participants had to answer to following questions in a random order:

Do you believe that Anky van Grunsven has a good or bad image?

Anky van Grunsven is a spokesperson for the Spanish Riding School; do you believe she is a credible ambassador?

(17)

3.1.2. Non-Expert Endorser Auction

For the non-expert level of celebrity endorsers the same treatments will be used for a different product and celebrity endorser. Recently Samsung has launched a new smart watch that could be used with the latest versions of their smart phones and tablets. The smart watch can be used to place calls without the need to take the phone out of your pocket, text, check email, has a camera and voice command. It is supposed to make it easier to do these things without needing to hold a smart phone or tablet. Due to a Bluetooth connection with a smart phone or tablet, the smart watch is able to diverge basic phone functions to a watch. In their advertisements they use Robin Thicke as a celebrity endorser (Daniel, 2013). Robin Thicke is an American-Canadian singer-songwriter. He had the biggest number 1 hit of 2013 called “Blurred Lines” which had features of Pharrell Williams and T.I (Interscope Records, 2003). In his career or education there is no track record of him being involved in the field of technological innovations or having a publicly displayed his interest for developing or creating new gadgets (Interscope Records, 2003).

For the first treatment, Robin Thicke will not be used in the advertisement for the smart watch; this advertisement will simply state the features and functions of the Samsung Galaxy Gear smart watch. In the treatment with low involvement Robin Thicke will be in the

advertisement but he will not visibly wear the smart watch or make any statements about the smart watch. In the final treatment with the high level of involvement Robin Thicke will be shown in the advertisement wearing the smart watch and an explicit comment stating that the smart watch is worn by Robin Thicke will on the advertisement. See the appendixes for the advertisements, which were used in the auction. After the auction the participants got the following questions in randomized order:

(18)

Do you believe that Robin Thicke has a good or bad image?

Robin Thicke is a spokesperson for the Samsung Galaxy Gear; do you believe he is a credible ambassador?

Have you ever considered buying the Samsung Galaxy Gear?

3.2. Sample Frame

For both experiments a sample group was randomly drawn from the Veylinx population, which consists of more than 3200 panel members. For the expert experiment a sample of 509 panel members responded with 53% male and 47% female participants. The completed

participation for the expert experiment was 100%, thus everyone who responded to the invite for the auction, made a bid. The panel for the non-expert experiment consisted of 406 members. 55% of the respondents were male and 44% was female. However, only 81% of the panel members, who received an invitation for the auction actually made a bid. Further details on the sample can be found in the table 1.

N Minimum Maximum Mean SD

Expert Endorser Auction Gender 509 0 1 .47 .500 Age 509 1911 1995 1972.82 16.501 Education 412 0 5 3.69 1.247 WTP 509 0 12000 821.04 1464.27 Celebrity Endorser 509 .00 1.00 .6365 .48147 Level of Involvement 509 1 3 1.90 .788 Non-Expert Endorser Auction Gender 405 0 1 .44 .498 Age 405 0 1998 1969.87 99.437 Education 180 0 5 3.97 1.090 WTP 329 .00 23700 2398.17 3603.989 Celebrity Endorser 405 .00 1.00 .6765 .46837 Level of Involvement 405 1 3 .2.02 .819

(19)

3.3. Dependent Variable – Consumers’ Willingness to Pay

For this study the dependent variable is consumers’ willingness to pay. This variable is chosen because it reflects the extent to which consumers are willing to give up or sacrifice something in order to gain a pleasurable experience or desirable product. This also indicates how the cons In this study the consumer’s willingness to pay is measured through the amount they are willing to pay for the auctioned product or service, i.e. their bid amount. The bid amount of a Vickrey auction is chosen because it incentivizes participants to not bid more than they actually want to pay. A Vickrey auction is characterized by its rule that the highest bidder always pays the second highest bid. Thus, people are stimulated to bid the highest amount that they would pay for the product or service because if they win, they will always pay less then their top amount. On the other side, it also implicitly withholds people from bidding more than their top amount because if they win the auction, they run the risk of having to pay more than they actually value the product. Thus, this method of measuring consumer’s willingness to pay puts the participants in an actual situation where they are not able to give ambiguous answers. The Vickrey auction also decreases the risk of a survey where there is the possibility of people saying what they would do in a certain situation and how they would actually act in a certain situation.

3.4. Independent Variable – Level of Involvement of the Celebrity Endorser

Measuring the extent to which the level of involvement of the celebrity endorser has on the willingness to pay, companies will be able to manage and exploit their endorsement deals more effectively. The involvement level of the celebrity endorser is divided in three stages in this study. The first involvement level is the stage with no endorser at all, in the second stage the endorser will have a low involvement level which will be a recommendation about the product or a picture of the endorser in the advertisement and the highest level of involvement has the

(20)

different advertisements will be shown during three different auctions. See the appendices for the advertisements.

3.5. Moderators 3.5.1. Celebrity Image

The moderator used for this study is the image of the celebrity. The image of the celebrity can be perceived as the way the celebrity is seen by society (Balmer, 2001). This perception of society is based on what the celebrity communicates willingly and unwillingly to the environment. Thus, this is also based on how the celebrity carries itself in public and what is communicated through media. All this information builds the image that consumer’s mold in their minds of this

celebrity, which can be either positive or negative, either way the celebrity does not solely have an influence on this image. A parallel can be seen between the celebrity image and the corporate reputation where corporations also don’t have a sole influence on how people perceive them as they gain information from what the company says, does and what is said about the company (Balmer, 2001). The celebrity image is measured through the small survey given after the auction. The question to measure the celebrity image is “Do you believe that [insert name

celebrity endorser] has a good or bad image?” Other studies have shown that the celebrity image is one of the reasons companies engage in endorsement deals in the hopes that the positive associations of the celebrity will transfer to the endorsed producer (Erdogan, 1999; Silvera & Austad, 2003). Thus, by asking the aforementioned question after the auction the perceived celebrity image can be measured.

3.5.2. Celebrity Credibility

The second moderator used in this study is celebrity credibility. Credibility refers to the extent to which a source or message has enough grounds to be believed and dependable

(21)

category and expertise and the likelihood that the endorser would also like the product if he or she would not have an endorsement deal with the company (Ohanian, 1991). The study of Ohanian (1991) has shown that endorsements with an evident connection between the endorser and the endorsed product are better received compared to deals without this connection. Another influence of credibility is the amount of endorsement deals the celebrity has. A celebrity with a lot of endorsement deals runs the risk of diluting their own brand because it is involved in multiple deals which all send their own message to the environment (Ohanian, 1991). Thus, a celebrity with too many deals will not be a very credible endorser for each new deal. Hence, it is expected that credibility as a moderator can influence the effect of the celebrity endorser on the consumers. To measure the perceived the credibility of the celebrity endorser the following question was asked after the auction: “[insert name celebrity endorser] is a spokesperson for the [insert endorsed product]; do you believe he/she is a credible ambassador?”

(22)

4.0 Results

4.1. Expert Endorser Data Analysis

In the online auction experiment with the expert endorser invitations were sent out to Veylinx members, out of which 509 responded to the invitation. Out of the 509 panel members, every person participated in the auction, which resulted in 100% completed participation of the respondents. Thus, for this analysis there were no missing values in the dataset.

As reliability is focused on measuring the consistency of a variable by having multiple items that measuring a mutual variable, no such test was conducted in this study. The main reason being that the bid amount given by people in the experiment is the only item that measures the willingness to pay. Hence, the reliability statistic will not serve its purpose with only one item. Note that the bid amounts given in the remainder of this research is given in eurocents. Thus, a bid amount of 860.22 is the equivalent of an €8,60 bid. For the statistical analysis the variable bid amount is recoded into a new variable by taking the natural logarithm of the items. The Celebrity Endorser variable is constructed by recoding the treatments variables with a binary distribution. Treatment 1 where no celebrity endorser was used for the

advertisement was recoded with the value 0 and the treatments 2 and 3 both were recoded with value 1 as they both did have a celebrity endorser in the advertisement. Due to the large amount of 0-bids, for the regression analyses the top 50 percentile of the dataset is used.

(23)

The means of the bid amounts per treatment are given in the table below.

Treatment N M SD Median Minimum Maximum

No Celebrity Endorser 185 860.22 1333.41 100.00 0 7500 Celebrity Endorser 324 798.66 1525.55 84.50 0 12000

• Low Involvement 189 860.80 1577.37 100.00 0 10100 • High Involvement 135 711.67 1476.41 0.00 0 12000 Table 2 - Descriptive Statistics for consumers’ willingness to pay (N=509)

The first One-Way ANOVA test showed a p-value of .004 between groups, because the p-value is smaller than .05. It means that there is a significant difference between the mean bid amounts of the different treatments. The Post-Hoc LSD test is used to measure the significance of the mean difference between the groups. The test shows that even though there is no

significant difference between the means of the “no celebrity endorser” and a “low involvement” treatment with a p-value of .430, there is a significant difference between the “high involvement treatment” and the other two treatments. With a p-value of .001 there is a significant different between the means of the “no celebrity endorser” treatment and the “high involvement treatment”. The p-value of the mean difference between the “low involvement” and the “high involvement” treatment is with .012 also significant being smaller than the alpha of .05.

Sum of Squares df Mean Square F Sig.

Between Groups 19.323 2 9.662 5.616 .004

Within Groups 430.095 250 1.720

Total 449.418 252

(24)

Treatment (I) Treatment (J) Mean Difference Std. Error Sig.

No Celebrity Endorser Low Involvement .152 .192 .430

High Involvement .686 .211 .001

Low Involvement No Celebrity _Endorser -.152 .192 .430

High Involvement No Celebrity

Endorser -.686 .211 .001

Low Involvement -.533 .210 .012

Table 3B - LSD Post-HOC test for consumers’ willingness to pay in top 50%

The linear regression analysis of the effect of the use of a celebrity endorser of the expert ambassador on the bid amount has a B-value of -.374 and a p-value of .032. The B-value shows that there is a negative effect between the celebrity endorser and the consumers’ willingness to pay. The p-value is smaller than the alpha of .05, which indicates that there is significant evidence for the detected relation between the celebrity endorser and consumer’s willingness to pay. Hence, there is not enough evidence in support of H1, where a positive effect was expected.

The linear regression test of the relation between the level of involvement of the celebrity endorser and the consumer’s willingness to pay resulted in a Beta of -.331 and a p-value of .002. The Beta reflects the negative relation between the involvement levels of the celebrity endorser and consumers’ willingness to pay. Even though the Beta shows that there is not a very strong effect, the p-value is below the alpha of .05, which indicates that the effect of the relation between the endorsement levels and consumers’ willingness to pay is significant. Therefore, there is no sufficient evidence to accept H2.

(25)

Model R Square Adjusted R Square Std. Error

Celebrity Endorser - WTP .018 .014 1.326

Level of Involvement - WTP .038 .034 .1.312

Table 4A - Model Summary for consumers’ willingness to pay

Variables N M SD B SE Sig.

Bid Amount in

top 50% (Constant) 253 6.85 1.335 7.085 .138 .000

Celebrity Endorser 253 .6364 .482 -.374 .173 .032 Level of Involvement 253 1.90 .788 -.331 .105 .002 Table 4B - Linear Regression Analysis for consumers’ willingness to pay

For the moderation analysis a moderator variable has to be constructed. This variable is constructed by computing the moderator variable as a product of items of the celebrity endorser variable and the items of the celebrity image variable. Subsequently a regression analysis is done with the celebrity endorser and newly formed moderator as independents and natural logarithm of consumers' willingness to pay as dependent. The p-value of this analysis is .015, which shows that the model is significant. Looking more deeply into the statistical results, it shows that the moderator has a Beta of 1.055 and a significance level of .125. These results show a strong, yet marginally significant effect of the image of the celebrity endorser on the effect of the impact of the celebrity endorser on the bid amount. Hence, there is partial evidence for H3 to be considered valid.

To test the moderation of the credibility of the celebrity endorser on the effect of the use of a celebrity endorser on consumers’ willingness to pay, the same analysis is done as described for the moderation effect of the image of a celebrity. First a new moderation variable was

(26)

constructed by multiplying the items of the celebrity endorser variable with the celebrity

credibility variable. After the new moderator was formed, a regression analysis was run with the variables celebrity endorser, celebrity credibility and the moderator. The model had a p-value of .193 with can be classified as a marginally significant effect. Looking more closely to the relations between the variables it is shown that the moderator variable has a Beta of .478 and a significance of .370. Thus, even though there is a small positive effect of the moderator on the impact of celebrity endorser on the bid amount, there is not enough evidence to argue that the effect is significant and H4 has to be rejected.

Celebrity Endorser - Image - WTP .205 .042 1.250 Celebrity Endorser - Credibility - WTP .016 .008 1.269

Variables N M SD B SE Sig. Bid Amount in top 50% (Constant) 240 6.90 1.280 6.829 .114 .000 Celebrity Endorser 240 .6333 .48290 -1.324 .665 .048 Celebrity Endorser Image 240 .93 .264 -.128 .582 .827 Moderator Image 240 .5792 .49472 1.093 .687 .113 Bid Amount in top 50% (Constant) 238 6.91 1.280 6.848 .114 .000 Celebrity Endorser 238 .6345 .48260 .126 .167 .000 Celebrity Endorser Credibility 238 .89 .318 -.208 .414 .616 Moderator Credibility 238 .5672 .49650 .492 .533 .357 Table 5B - Moderation Analysis for consumers’ willingness to pay

(27)

4.2. Non-Expert Endorser Data Analysis

The data that came out of the online experiment with the non-expert endorser required some filtering, because not all respondents actually finalized the whole experiment. Thus, by filtering out the cases that did not complete the experiment, i.e. made a bid, the total cases used for the data analysis are 329 out of the 406 participants. For the regression analysis the outliers were excluded from the analysis. First a frequencies analysis was constructed using the median, because the mean is too sensitive to outliers. This is done by first excluding all items below the median of the bid amount variable. Thus, only the top 50 percentile of the dataset is used. The natural logarithm of the variable bid amount is taken to standardize the variable for the analyses. Thus, the newly formed natural logarithm of consumers' willingness to pay variable is used as independent variable throughout the data analyses. The Celebrity Endorser variable is

constructed by recoding the treatments variables with a binary distribution. Treatment 1 where no celebrity endorser was used for the advertisement was recoded with the value 0 and the treatments 2 and 3 both were recoded with value 0 as they both did have a celebrity endorser in the advertisement.

For these types of experiments reliability is usually not tested because it is a one-time auction and thus, it is not possible to measure the consistency of different items referring to the one variable. The descriptive statistics of the real bid amounts per treatment are given in the table below.

(28)

Treatment N M St. D. Median Minimum Maximum No Celebrity Endorser 108 1979.83 3190.00 950 0 20000

Celebrity Endorser 221 2602.61 3779.93 1000 0 23700 • Low Involvement 110 2407.06 3254.03 1000 0 15000 • High Involvement 111 2796.40 4243.48 1000 0 23700 Table 6 - Descriptive Statistics for consumers’ willingness to pay (N=329)

The One-Way ANOVA test between the bid amounts and the treatments have shown no significant effects with a p-value of .117. But the Post-Hoc test done after the simple One-Way ANOVA test between the bid amounts and the different treatments has shown that there is a significant difference in the bid amounts between the treatment with “no celebrity endorser” and using a celebrity endorser with high involvement. The comparison of the means with the LSD Post Hoc analysis gave a significant p-value of .038 between the treatments. Comparison of the mean differences between the “no celebrity endorser” and the “low involvement” treatment and between the low involvement level and the high involvement level give results of respectively .284 and .296. Neither of the means of those treatments was significantly different.

Sum of Squares df Mean Square F Sig.

Between Groups 2.907 2 1.453 2.178 .117

Within Groups 110.790 166 .667

Total 113.697 168

(29)

Treatment (I) Treatment (J) Mean Difference Std. Error Sig.

No Celebrity Endorser Low Involvement -.165 .154 .284

High Involvement -.325 .156 .038

Low Involvement No Celebrity Endorser .165 .154 .284

High Involvement No Celebrity Endorser .325 .156 .038

Low Involvement .160 .152 .296

Table 7B - LSD Post-HOC test for consumers’ willingness to pay in top 50%

For the following analyses a linear regression test is conducted in order to model the relation between the different independent variables and the dependent variable of consumers’ willingness to pay.

The first regression analysis of the relation between the use of a celebrity endorsers and consumer’s willingness to pay had a B-value of .243 and a p-value of .073. This result shows that there is a positive relation between the use of the celebrity endorser and consumers’ willingness to pay. The p-value of the significance is slightly higher than the alpha of .05. However, because the difference is only .023, the effect is considered as significant. Hence, there is enough

evidence to accept H1 as valid.

For the second hypotheses the effect of the level of involvement of the celebrity endorser on consumers’ willingness to pay was theorized. The overall model showed a significance level of .038. The regression test to analyze the effect of the level of involvement on the consumers’ willingness to pay shows, that there is a significant effect between the two. The regression analysis provided a Beta of .163 and a p-value of .038. Even though the positively relation

(30)

between the level of involvement and consumers’ willingness to pay is not a very strong effect, it is still very significant. Therefore, there is H2 is proven to be true.

Celebrity Endorser - WTP .019 .013 .81719

Level of Involvement - WTP .026 .020 .8145

Variables N M SD B SE Sig.

Bid Amount in

top 50% (Constant) 169 8.097 .8226 7.932 .111 .000 Celebrity Endorser 169 .6805 .4677 .243 .135 .073 Level of Involvement 169 2.01 .809 .163 .078 .038 Table 8B - Linear Regression Analysis for consumers’ willingness to pay

For the moderation analysis a new moderator variable has to be created. The first

moderator that will be formed is to see whether the variable celebrity image has an effect on the relation between the celebrity endorser and consumers’ willingness to pay. This variable is formed by multiplying the items of the variable of celebrity endorser with the items of celebrity image. For the moderator analysis a linear regression is done with consumers’ willingness to pay as dependent variable and celebrity endorser, and the moderator as independents. The complete model had a significance of .137, which shows that there is no evidence found for the model to be considered as significant. The regression resulted in a B-value of .065 and a p-value of .717 for the moderator. Thus, this shows that the image of the celebrity slightly increases the effect of the use of the celebrity endorser on consumers’ willingness to pay. However, due to the p-value

(31)

being larger than the alpha of .05, there is no significant evidence for this effect. Hence, H3 has to be rejected.

For the moderator analysis of the celebrity credibility on the effect of the celebrity endorser on consumers’ willingness to pay, another moderator has to be created. This moderator variable is constructed by multiplying the items of the celebrity endorser with the items of the celebrity credibility. Subsequently, the celebrity endorser, celebrity credibility and the moderator are used as independent variable in the regression analysis with consumers’ willingness to pay as dependent variable. The regression analysis resulted a p-value of .068 for the complete model. Which shows that there is significant evidence that the suggested model is significant. With a B-value of -.183 the effect of the moderator on the relation between celebrity endorsers and consumers’ willingness to pay is quite small. The p-value of .290 is larger than the alpha of .05; therefore there is no evidence that the effect of credibility as a moderator is significant. Thus, H4 has to be rejected as well.

Celebrity Endorser - Image - WTP .027 .014 .81857 Celebrity Endorser - Credibility - WTP .036 .023 .80886

(32)

Variables N M SD B SE Sig. Bid Amount in

top 50% (Constant) 144 8.124 .8222 8.281 .080 .000

Celebrity Endorser 144 .6597 .4755 .232 .184 .211 Celebrity Endorser Image 144 .67 .471 -.332 .239 .167 Moderator Image 144 .4444 .4986 .065 .179 .717 Bid Amount in top 50% (Constant) 145 8.134 .8137 8.294 .078 .000 Celebrity Endorser 145 .6552 .4770 .232 .184 .211 Celebrity Endorser Credibility 145 .59 .493 -.411 .218 .062 Moderator Credibility 145 .4138 .4942 .065 .179 .717 Table 9B – Moderation Analysis for consumers’ willingness to pay

(33)

5.0 Discussion

For the discussion the results of the expert and non-expert auction are first discussed separately and finally the similarities and differences between the two auctions are laid out.

5.1. Expert endorser auction

The results of the auction with the expert celebrity endorser were least to say counter-intuitive. Even though previous research had shown that the use of a celebrity endorser could positively increase the attitude of consumers towards the endorsed product (Erdogan, 1999; Ohanian, 1991; Silvera & Austad, 2003), the results of the experiment with the expert celebrity endorser showed something else. Where a positive effect of the use of a celebrity endorser was hypothesized on consumers’ willingness to pay, the data showed that the use of this particular expert endorser had a significant negative effect on consumers’ willingness to pay. Hence, this indicates that when the expert celebrity was endorsing the auctioned product, people were inclined to pay less for the auctioned product compared to the people who did not see the celebrity in the advertisements endorsing the auctioned product. This shows that the presence of the celebrity endorser did not had the desired expected effect of increasing the willingness to pay. A reason for this unexpected turn of events can be the debate about the training methods of Anky van Grunsven, the expert endorser of the auctioned product. The current negative publicity about the training subject could be a reason for the result. Whereas in the literature, the argument is made that there is merely a positive relation between the endorser and the consumers attitude towards the endorsed product (Agrawal & Kamakura, 1995; Erdogan, 1999; Ohanian, 1991; Silvera & Austad, 2003), the data shows that the use of a celebrity endorser could lead to the undesirable event of consumers’ devaluating the product based on the affiliation with the celebrity endorser. Another reason for the surprising outcome of this experiment could be the categorization. Whereas, Anky van Grunsven is known for being an exquisite equestrian, she is

(34)

not being perceived as a host. Thus, people categorizing Anky van Grunsven as an equestrian and not as a show presenter, could be an explanation of the lower willingness to pay. However, further research on this categorization issue is needed.

Looking more closely to the different levels of involvement of the expert endorser another significant negative effect on consumers’ willingness to pay has manifested itself. In other words, the more the celebrity endorser was involved with the endorsed product, the less people were willing to pay for the product. This negative effect could be a consequence of the current controversy concerning the celebrity endorser Anky van Grunsven, who is being

questioned about possibly harmful equestrian training methods as mentioned before. This shows that also on different levels of the endorsements, the celebrity endorser in question increasingly had a negative effect on consumers’ valuation of the endorsed product. A remarkable finding here was that the difference between the advertisements with no celebrity endorser and low involvement showed not a big fluctuation in the bid amounts, because the low involvement only showed a quote from the endorser recommending the show. The high involvement advertisement where it was clearly stated that the celebrity endorser was participating in the show showed a real decrease in the bidding amounts.

A really strong effect is found concerning the impact of the influence of the celebrity endorser on the previously established relation between the celebrity endorser and consumers’ willingness to pay. The experiment showed in this case that the image is a definite moderator in this relation. Having a very strong effect on the impact of the celebrity endorser on consumers’ willingness to pay, the image of the celebrity is able to neutralize the negative effect of the celebrity endorser on consumers’ willingness to pay. Meaning that the negative effect that this

(35)

the celebrity’s image. Thus, this auction showed that when there is a negative impact of the celebrity endorser on consumers’ willingness to pay, a good image of the celebrity endorser could neutralize this negative effect. On the other side, a bad image of the celebrity endorser will increase the negative effect of the celebrity endorser on consumers’ willingness to pay even more. Either way, this strong positive, yet marginally significant effect shows support for the theory that already argued that the image of the celebrity is important to appeal to the consumers (Erdogan, 1999; Silvera & Austad, 2003). However, where this previous research indicates that the image of the celebrity functions as a precedent of the decision to use the celebrity endorser, this study has shown that the image of the celebrity endorser also has a moderation effect on the use of the celebrity endorser and consumers’ willingness to pay for the endorsed product.

Previous studies have argued that the credibility of a celebrity endorser also has an effect on the consumers’ attitude towards the endorsed product (Ohanian, 1991). The experiment showed that there is indeed a small, yet positive effect of the credibility of the celebrity endorser as a moderator on the relation between the celebrity endorser and consumers’ willingness to pay. Resulting in the fact that if the use of a celebrity endorser has a negative effect on consumers’ willingness to pay; this effect is decreased if the celebrity is deemed as a trustworthy source of information. This is in line with the study of Ohanian (1991) arguing that credible endorsers contribute to shaping the perception of the endorsed product by the consumers. However, where Ohanian (1991) argues that credibility is a precedent of the use of a celebrity endorser, this study enlightens a different role of the credibility as a moderator. Even though a positive effect was found by analyzing the data of the experiment, the effect is not significant.

(36)

5.2. Non-expert auction results

As said before the previous studies have shown that there is a positive effect between the use of a celebrity endorser and consumers’ attitude towards the endorsed product (Erdogan, 1999; Ohanian, 1991; Silvera & Austad, 2003). This auction showed support for this theory as well. As expected a small, yet positive effect of the use of a celebrity endorser on consumers’ willingness to pay was discovered during the data analysis. Even though it was small, the positive effect was significant. Which means that in this case the use of the celebrity endorser resulted in an increase in the amounts auction participants were willing to pay for the endorsed product. The increase in the bid amounts when the celebrity endorser was involved is suspected to be a consequence of the current popularity of the celebrity endorser, Robin Thicke. The high rotation on radio stations and other media of his music could be the reason for this impact. When looking more deeply into the use of the celebrity endorser by different levels of involvement, the same effect was found. The level of involvement of the celebrity endorser showed to have a positive effect on consumers’ willingness to pay. Meaning that the more the celebrity was

involved in the marketing campaign, the higher the bidding amounts were. This effect was also a small positive, yet significant effect as was hypothesized.

For this auction the moderation effect of the celebrity image and the credibility of the celebrity endorser on the effect of the use of the celebrity endorser on consumers’ willingness to pay is tested as well. The literature has argued that the image of the celebrity endorser functions merely as a precedent of the use of the celebrity endorser (Erdogan, 1999; Silvera & Austad, 2003). Meaning that if a celebrity has a good image, it should be more appealing to the

consumers compared to a celebrity with a bad image. However, for this study the image of the celebrity as a moderator on the relation between the use of the celebrity endorser and consumers’

(37)

willingness to pay explores a different role of the celebrity image. The experiment has shown that in this case the image of the celebrity endorser also has a positive effect on the relation between the celebrity endorser and consumers’ willingness to pay. Thus, this means that as expected if the celebrity is perceived as having a good image, this will increase the positive effect of the celebrity endorser on consumers’ willingness to pay for the endorsed product. On the contrary, if the celebrity would be perceived as having a bad image, the positive effect of the celebrity on consumers’ willingness to pay would in this case be decreased. Unfortunately, there was no indication of a significant effect.

As aforementioned the existing literature argues that the credibility of the celebrity endorser is one of the pillars on whether or not a celebrity would be suitable for the endorsed product or not (Ohanian, 1991). Thus, in this theory the credibility of the celebrity is situated as a precedent in the model, whereas in this study the position of credibility as a moderator is

explored. For the credibility a negative moderator effect was confirmed by the data analysis. A small, yet negative effect of credibility was found on the effect of the use of celebrity endorser on consumers’ willingness to pay. Hence, when the use of a celebrity endorser has a positive effect on consumers’ willingness to pay, being perceived as a credible endorser could decrease this effect. In the case of being not perceived as a credible endorser, this could increase the positive effect of the use of the celebrity endorser on consumers’ willingness to pay. However, just like the moderation effect of the celebrity image, no significance was found for celebrity credibility as moderator that indicates that we can only establish for now that in this case there is negative moderation effect. An explanation for this effect could be the categorization of the celebrity endorser. Even though Robin Thicke is no expert in gadget technology, his music in which he excels, was used for the marketing campaign. Thus, this could be the effect of

(38)

consumers perceiving him as an artist and him taking on the role of artist in an endorsement deal for a product which he is not an expert in. however, this categorization issue needs further investigation.

5.3. Expert Celebrity Endorser results vs. Non-Expert Celebrity Endorser results Even though both experiments were done exactly the same, two different products and two different endorsers were used. For the first experiment tickets for a live performance of the Spanish Riding School were auctioned with Anky van Grunsven as the expert endorser. In the second experiment a Samsung Galaxy Gear smart watch was auctioned off with the singer-songwriter Robin Thicke as non-expert endorser. The expert classification that was made between the endorsers was constructed on the grounds of their knowledge about and affiliation with on the product category based on their careers (Ohanian, 1991).

Looking at the differences in the effect of the overall use of the celebrity endorser on consumers’ willingness to pay, it is striking to see that in the case of the expert celebrity there is a negative effect on consumers’ willingness to pay, while in the case of the non-expert the celebrity has a positive effect on consumers’ willingness to pay. However, even though the effects are going in different directions, both are significant. The data has also shown that the negative effects of the expert endorser are stronger than the positive effects of the non-expert endorser. Whereas the theory argued that the celebrity endorser would have a positive effect on the consumers (Erdogan, 1999; Silvera & Austad, 2003), the data shows otherwise. Reason for this could be the current controversy in the equestrian industry concerning the expert endorser, Anky van Grunsven for her training methods, whereas Robin Thicke is currently very popular and relevant because of his hit songs. Thus, an explanation for the differences in the direction of the relationship could be due to the public attention and media coverage on the celebrities.

(39)

However, both auctions showed that the model illustrating that the use of a celebrity endorser has an effect on consumers’ willingness to pay was valid. The same results occurred for auction concerning the relation between the level of involvement of the celebrity endorser and consumers’ willingness to pay are compared. Whereas the auction with the expert endorser shows a negative relation between the level of involvement and consumers’ willingness to pay, there is a positive relation between the level of involvement and consumers’ willingness to pay for the non-expert endorser. Thus, this shows as well that the endorser can affect the direction of the relation between the level of involvement and consumers’ willingness to pay. Since both relations are significant, there is enough evidence to argue that consumers’ willingness to pay for the endorsed product is also affected by the level of involvement of the celebrity endorser.

For the moderator effect of the image of the celebrity on the impact of the celebrity endorser on consumers’ willingness to pay comparing the results leads to some surprising

findings. The strong positive effect of the image of the expert celebrity endorser on the influence of the celebrity endorser on consumers’ willingness to pay completely cancels out the initial negative effect that the celebrity endorser had on consumers’ willingness to pay. Even though there is marginal significance, it is evident that the image can have a big impact on the on the effect of the celebrity endorser on consumers’ willingness to pay, as discussed before this can be due to the current controversy concerning the expert celebrity endorser. The non-expert celebrity endorser image also showed to have a positive effect on the influence of the celebrity endorser on consumers’ willingness to pay. Since there already was a positive relation detected between these two variables, the image of the non-expert celebrity increases the positive effect as a moderator, whereas the image of the expert celebrity is particularly balancing out the negative effect as a moderator.

(40)

The moderation effect of the credibility of the celebrity endorser on the relation between the celebrity endorser and consumers’ willingness to pay showed no similarities on both the direction of the relation and even though both effect were insignificant. Where the expert celebrity auction showed a positive effect of credibility as moderator, the non-expert celebrity auction showed a negative effect of the credibility as moderator on the impact of the celebrity endorser on consumers’ willingness to pay. In the case of the expert celebrity this positive

moderation effect is able to neutralize the negative effect of the celebrity endorser on consumers’ willingness to pay. For the expert celebrity the negative moderation effect decreases the

established positive relation between the celebrity endorser and consumers’ willingness to pay. However, neither of the data sets has shown any significance for these effects. Where previous studies argue that credible sources can eventually lead to more positive attitudes towards the endorsed products from the consumers (Ohanian, 1991), no evidence has been found. Thus, this could mean that even though consumers could have a more positive attitude towards the

endorsed product, this does not mean that they are actually willing to also pay more for the product if the endorser is considered as a credible source of information. As discussed before, a categorization issue could be the cause of this problem whereas Anky is diverging from her role as an equestrian, Robin Thick is converging with his role as an artist. However, future research is needed for the effects of categorization by the consumers on these relations.

Both data sets have shown that there is indeed a significant relation between the use of celebrity endorsers and consumers’ willingness to pay. This study has revealed that the use of a celebrity endorser undeniable affects the amount consumers are willing to pay for the endorsed product. However, the direction of the relation remains unanswered. The experiment with the expert celebrity endorser showed a negative relation between the celebrity endorser and

(41)

consumers’ willingness to pay, whereas the non-expert endorser showed a positive relation between the two variables. Since both experiments were done with different products and different celebrities, there is more research needed to conclude whether the directions of the relation are also depending on the expertise of the celebrity endorsers.

(42)

6.0 Conclusion

A celebrity endorsement is a testimonial for a product, appearance at PR event, being a

spokesperson or other form of engagement of a public figure in an advertising campaign in order to exploit their status in society to promote a product, service or charity goal (Erdogan, 1999). Through an experiment the aim of this study is to find out whether celebrity endorsers really have a positive effect on consumers’ attitude in that sense whether consumers are willing to pay more for a product endorsed by a celebrity than a product without an endorsement. The effect of the image of the celebrity and the credibility of the celebrity on the relation between the celebrity endorser and consumers’ willingness to pay is investigated as well.

The set up of the experiment is a Vickrey auction. This is a second-bid auction where the winner of the auction, i.e. the highest bidder, has to pay the second highest price. For the

experiment two auctions were done; one auction with an expert celebrity endorser and tickets for a live-performance and on auction with a non-expert celebrity endorser and a smart watch.

For each auction three different advertisements were made, one without the endorser, one with low involvement of the endorser and on with high involvement of the endorser. For the expert celebrity the results showed that the celebrity endorser and the different involvement levels both had a significant negative effect on consumers’ willingness to pay. The image of the celebrity had a strong positive moderation effect and the credibility of the celebrity showed a positive moderation effect as well. However, whereas the image of the celebrity reached

marginal significance, there was no significant evidence for the moderation effect of credibility. The non-expert auction showed that the celebrity endorser and the different levels of

(43)

negative moderation effects on the relation between the celebrity endorser and consumers’ willingness to pay. However, neither of the moderation effects was significant.

This study has partially answered the research question, arguing that the use of celebrity endorsers indeed have an effect on the way consumers valuate endorsed products. Both auctions have shown that the use of a celebrity endorser and even different involvement levels of

endorsing does affect consumers’ willingness to pay for the endorsed product. However, the direction of the relation is yet not answered due to the differences in the types of relation

between the expert and non-expert endorser auction. The different directions could be due to the expertise of the endorsers or the relevancy and public attention of the endorser at the time. Thus, further research is required to answer this question. This question could be answered by doing the same experiment with one celebrity endorser and one product with different advertisements based on expertise.

In this study the direct effect of celebrity endorsers on the valuation by the consumers of the endorsed products was studied. The study has contributed in to the theory by adding a new dimension to the model by focusing on the consumer behavior and the act of buying an endorser product. Hence, where previous studies has stuck to the attitude of the consumer towards the endorsed product and intention of buying, this study has gone one step further by focusing on the direct connection between the celebrity endorser and people valuating the endorsed product by putting them in a real life situation to purchase the product. For management that is involved with expensive endorsement deals, it shows that it is possible to measure how the consumers of the endorsed product will accept the celebrity endorser. When it may seem to be a good match on paper, the study has shown in the case of the expert celebrity endorser that sometimes an

(44)

6.1. Implications, Limitations & Recommendations for future research

The experimental auctions that have been undertaken for this research consisted of different items and endorsers for the experiment. For the expert celebrity endorser auction a tickets for a live performance were sold and for the non-expert celebrity endorser a smart watch was auctioned. Due to the different products and different endorsers, it was not possible to do a data analysis to see whether the expertise of the celebrity also would have an effect on

consumers’ willingness to pay. Thus, expertise was one of the elements that determined the outlook of the auction. Yet it was not a variable that could be tested as well. For future research the same experiment could be repeated with the one product and one endorser and expertise could be added as a variable to the model. If expertise is added a as moderator variable such as celebrity image and celebrity credibility, more insights could be gained on whether the expertise of the celebrity does actually matter for the consumers’ willingness to pay for the endorsed product.

Another implication of the study was that the endorsers were considered an expert or non-expert based on their career paths, awards and public display of interest in the product category. Thus, it is unclear whether the respondents considered the celebrity endorsers as experts or non-experts and whether that mattered to them. Once tested this could create a new perspective for the research.

As mentioned in the discussion section, a possible categorization issue was raised for the expert celebrity experiment. Because the expert celebrity fulfilled a different role which made her more involved in the marketing and production of the endorsed product, but diverged from the role consumers were used to seeing here in. Thus, categorization could be an important factor for the use of celebrity endorsers as well. For example, this could be tested with the same