Can Athlete Super Stars Just Do It?
The Influence Of Valence, Amplitude, And Relevance Of Celebrity Endorser Information On Consumer Brand Evaluation
MASTER THESIS
Student Wouter Jonkers Student number 10575960
University University of Amsterdam Faculty Economics and Business Study program Business Administration (MSc)
Specialization Marketing
Supervisor R. Pruppers Date of submission 29 – 6 - 2015
Version Final
All rights reserved. This master’s thesis is for personal, non-commercial use. Its content may not be sold, copied, stored, modified, retrieved, translated or published in any form or by any means, either electronically, mechanically, or any other way, without the prior written permission of the author.
Statement of Originality
This document is written by Student Wouter Paul Gerard Jonkers who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is 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.
Table of Content List of Tables List of Figures Abstract 1. Introduction………...1 1.1. Problem Definition………..3 1.1.2. Problem Statement………..3 1.1.3. Sub-Questions………..5 1.1.4. Delimitations………....5 1.2. Contributions………...6 1.2.1. Theoretical Contributions………...6 1.2.2. Managerial Contributions………..7 1.3. Outline………..8 2. Theoretical Framework………..9 2.1. Associative Networks………..9 2.1.1. Brand Knowledge………..10
2.2. Celebrity Brand Endorsers………..13
2.2.1. Classical Conditioning………..13
2.2.2. The Meaning Transfer Model………..14
2.2.3. The Match-up Hypothesis………....…15
2.2.4. The Elaboration Likelihood Model……….16
2.3. Processing of Information of Different Valence, Amplitude and Relevance………...17
2.3.1. Valence of Information……….18
2.3.2. Amplitude of Information………19
2.3.3. Relevance of Information……….20
4. Method………...……26 4.1. Stimuli Development……….26 4.2. Pre-Test………..28 4.3. Main Experiment………..32 4.3.1. Research Design………32 4.3.2. Procedure………...32 4.3.3. Measures………33 5. Results………37 5.1. Respondent Analysis……….37 5.2. Data Preparation………...38 5.3. Manipulation Checks………40 5.4. Control Variables………..46 5.5. Hypotheses Testing………...50 5.5.1. Hypothesis 1………...51 5.5.2. Hypothesis 2………...52 5.5.3. Hypothesis 3………...52 5.5.4. Hypothesis 4………...52 5.5.5. Additional Analyses………...53 6. Discussion………...62
6.1. Interpretation of the Results……….62
6.1.1. Methodological Explanation……….63 6.1.2. Conceptual Interpretation. ………..………64 6.2. Implications………...67 6.2.1. Theoretical Implications. ………..67 6.2.2 Managerial Implications.. ……….………71 7. Conclusion.. ………...75 7.1. Limitations……….………..………...76
7.2. Directions for Future Research………77
9. Appendices……….85
Appendix A Pre-test Scenarios………85
Appendix B Pre-test Questionnaire………90
Appendix C Pre-test Results per Scenario………..…92
Appendix D Main Experiment Scenarios………...……94
Appendix E Main Experiment Questionnaire………99
Appendix F Main Experiment Brand and Endorser Familiarity per Scenario……….…101
Appendix G Main Experiment Component Analysis………..……102
List of Tables
Table 1: Selected scenarios for main experiment, including mean scores and standard
deviations on Valence, Amplitude, and Relevance………..……30
Table 2: Familiarity and Attitude towards brand and brand endorser………..…...31
Table 3: Respondent nationality and education level………..…....37
Table 4: Reliability analysis for scales of Attitude towards the Endorser, Attitude towards the Brand, Valence, Amplitude, and Relevance………...38
Table 5: Eigen values for Extracted factors resulting from PCA………..…..39
Table 6: Principal Component Analysis for items for Valence, Amplitude, and Relevance……..…...40
Table 7: Scores for Valence, Amplitude and Relevance (M (SD)) for the eight experimental scenarios and control scenario, including means per manipulation level……....41
Table 8: Results of Univariate ANOVA for measured Amplitude………...42
Table 9: Results of Univariate ANOVA for measured Valence………...43
Table 10: Results of Univariate ANOVA for measured Relevance………....45
Table 11: Results of Univariate ANOVA for measured Perceived Realism………..….47
Table 12: Results of Univariate ANOVA for measured Involvement………....48
Table 13: Regression analysis results for AttEndorser………...51
Table 14: Regression Analysis results for AttBrand………...52
Table 15: Regression analysis results for AttEndorser using raw measured scores………54
Table 16: Regression Analysis results for AttBrand using raw measured scores………54
Table 17: Means and Standard definitions, M (SD), for Attitude towards the endorser and Attitude towards the brand for the eight experimental scenarios and the control scenario…………...56
Table 18: Results of Univariate ANOVA for AttEndorser………58
List of Figures
Figure 1: Sources of secondary brand knowledge identified by Keller (2003)……….12
Figure 2: Conceptual model displaying the relationships between valence, amplitude,
relevance, attitude towards the endorser, and attitude towards the brand………...22
Figure 3: Developed experimental scenarios……….………...…….27
Figure 4: Chosen scenarios for main experiment. The control scenario consisted of
Robin van Persie visiting Paris during his football free weekend with his wife………....33
Figure 5: Scree plot for Principal Component Analysis of items for items of
Valence, Amplitude, and Relevance scales……….………..….39
Figure 6: Interaction patterns for intended Valence, Amplitude, and Relevance
influencing measured Amplitude scores……….………...…….43 Figure 7: Interaction patterns for intended Valence, Amplitude, and Relevance
influencing measured Valence scores……….………...…….44
Figure 8: Interaction patterns between intended Valence and intended Amplitude
influencing measured Relevance scores……….………...………….45 Figure 9: Interaction patterns between intended Valence and intended Amplitude
influencing measured Perceived Relevance scores……….…..……….47 Figure 10: Conceptual model for testing hypotheses with PROCESS,
with test results per hypothesis………..…….50
Figure 11: Conceptual model for testing hypotheses with PROCESS,
with test results per hypothesis, using raw measured scores……….……….53 Figure 12: Results of Univariate ANOVA for AttEndorser
Figure 13: Results of Univariate ANOVA for AttBrand
with intended Valence, Amplitude, and Relevance as predictor variables………...………..59
Abstract
In today’s advertising industry, celebrity brand endorsers are used frequently as spokespersons for brands, in order to promote products or services and enhance consumer brand evaluation. Being public figures, celebrity endorsers are often the subjects of news items, rumors, and gossip: celebrity endorser information. Despite the vast amount of resources that are spent on celebrity endorsement deals, academic literature has yet to fully comprehend the impact that different kinds of celebrity endorser information has on the formation of consumers’ attitude towards a brand. The present study focuses on three characteristics of celebrity endorser information: Valence (positive versus negative information), Amplitude (the extent to which the information is positive or negative), and Relevance (the relevance of the information for the shared cognitive consumer category of both the brand and the celebrity endorser). A 2 (Valence: positive/negative) x 2 (Amplitude: low/high) x 2 (Relevance: low/high) between subjects experimental design was used to conduct the research. The experiment was conducted among 294 respondents by using an online questionnaire platform. The celebrity endorser information for the experimental conditions was selected during a pre-test. Results show that the positive effect of Valence on Attitude towards the brand, was fully mediated by Attitude towards the Endorser. Furthermore, Amplitude moderated the effect of Valence on Attitude towards the Endorser, high Amplitude enhancing the relationship. There were no significant effects of Relevance found. An explanation for the results could lie in the use of a real-life celebrity endorser and brand, the possibly imperfect conceptualization of Amplitude, and the probable misinterpretation of the shared cognitive consumer category for Relevance items. Further research could replicate the present study using hypothetical brands and celebrity brand endorsers, to prevent pre-existing associations influencing the results.
1 1. Introduction
In today’s advertising industry, entailing over three billion euros in the Netherlands alone (Nielsen, 2014), many companies are using celebrity endorsers as spokespersons for their brand in advertising efforts to elicit more positive brand evaluations with consumers (Kamins & Gupta, 1994). Enormous amounts of money are spent every year to ensure that brands are represented by the most appealing celebrities, as shown by the 50 million dollar endorsement deal that R&B singer Beyoncé Knowles signed with PepsiCo in 2012 (“Most outrageous celebrity endorsement’’, 2012).
Celebrity endorsers are frequently used in a wide range of advertisements for day to day products: from razor blades (e.g. Roger Federer endorsing Gillette) to video games (e.g. Lionel Messi appearing on the cover of Electronic Arts’ football simulator FIFA 15). Estimates are that, in the U.S., fifteen percent of all advertisements involve a celebrity endorser (Crutchfield, 2012)
One can imagine that companies would not spend these vast sums of money on an unproven concept. In the past decades, the use of celebrity brand endorsers has been studied from a wide range of scientific perspectives, one of the most elaborate being the view of associative learning.
In associative learning, studies show that presenting an affectively pleasant stimulus simultaneously with a brand can lead to more favorable brand attitudes by meaning transfer of associations (Priluck & Till, 2004; Till, 1998; McCracken, 1989, 1986). Showing higher likeability and physical attractiveness than their non-celebrity counterparts (Freiden, 1984), celebrity brand endorsers can in fact be seen as stimuli that generate positive affect with consumers. This makes them very useful as unconditioned stimuli in classical conditioning to transfer positive associations from themselves to a product or brand (Till, Stanley & Priluck,
2 2008). Considering a big share of the literature, this is exactly how celebrity brand endorsers are mostly considered: as people who transfer a positive affective meaning to the brand they endorse (Erdogan, 1999, Misra & Beatty, 1990).
However, brand managers must also consider the possibility that these favorable associations with the celebrity will change for the worse. Keller (2003) describes brand endorsers as a secondary source of brand knowledge, able to transfer information and meaning from itself to a brand, be it negative or positive. If a celebrity has become the center of negative attention by being involved in a scandal of some sort (e.g. the exposure of the recurrent unfaithfulness of Tiger Woods, or the unmasking of Lance Armstrong as a user of performance enhancing drugs), what are the consequences for the way that consumers evaluate a brand?
In the past decades, literature has described the general effects of negative celebrity endorser information on consumer brand evaluations. Till and Shimp (1998) and Amos, Holmes and Strutton (2008) have shown in their research that when the associative link between a brand endorser and a brand is strong, negative information about the celebrity brand endorser will decrease the consumer evaluations of the endorsed brand (e.g. Tiger Woods’ extramarital affairs would damage the consumers’ evaluation of the brand he endorses: Nike). Also, Money, Shimp and Sakano (2006) state that negative information about celebrities gets more attention and is more easy to recall for consumers than positive information.
The mentioned studies make it very clear that negative information about a celebrity brand endorser can severely damage consumer brand evaluations. What is less thoroughly examined however, is whether positive celebrity endorser information could have reverse effects on consumer brand evaluations (i.e. positive celebrity endorser information generating a more favorable consumer attitude towards a brand).
3 In the last decade or so, the focus of scholarly attention has mainly been the mitigating and enhancing factors that influence the effects of negative brand endorser information on brand evaluation. (Um, 2013; Fong & Wyer, 2012; Johnson, 2005; Ahluwahlia, Burnkrant & Unnava, 2000), leaving the possible effects of positive celebrity endorser information somewhat underexposed.
To fill this gap in knowledge, the present study aims to identify more, previously unexplored variables that influence the relationship between negative and positive brand endorser information and consumer brand evaluations.
1.1. Problem Definition
1.1.2. Problem Statement
Both the brand and the celebrity brand endorser exist in the minds of consumers as association sets. An association set ‘represents the preexisting associates or group of concepts that are related meaningfully to an object’ (Till & Shimp, 1998, p. 68). The associative nodes of the brand and the celebrity are connected by experiences of consumers with said brand or celebrity, and in this way form association sets (Till & Shimp, 1998). Spreading activation causes the link with the brand node to be accessed when a consumer thinks about the celebrity brand endorser.
There might be factors that can influence the strength of the associative link between brand and brand endorser information, thus influencing the ease of accessing brand and/or celebrity association sets (Amos et al., 2008; Till and Shimp, 1998) when consumers evaluate brand endorser information. Earlier research (Petty & Briñol, 2008; Petty & Cacioppo, 1979) has already shown that personal relevance of information increases the level of message processing for consumers. It is possible that this increase in information engagement also extends to different kinds of information relevance. In the present study, the effects of
4 information relevance (i.e. the relevance for the shared cognitive category that both the brand and the brand endorser belong to) on brand evaluation are examined.
For example: there could be a difference in the ease of accessing the association set for the sports brand Nike, if it comes to light that football player Cristiano Ronaldo has injured another player during a football match (information highly relevant to football, a core association for both the Nike brand as well as Cristiano Ronaldo), as opposed to information about Cristiano Ronaldo committing tax fraud (low relevance information).
Aside from the relevance of the information for the meaning of the brand, there are other possible factors that might influence the effect of brand endorser information on consumer brand evaluations. For instance, people process information with a varying degree of valence (i.e. slightly negative/positive versus very negative/positive) differently (Herr, Kardes & Kim, 1991). In the present study these different degrees of valence are dubbed the amplitude of the information. A real life scenario: would consumers evaluate information about famous athlete Oscar Pistorius shooting and killing his girlfriend (high amplitude negative information) differently than information regarding the use of marihuana (low amplitude negative information) by professional swimmer Michael Phelps?
In the context of negative brand information, Einwiller, Fedorikhin, Johnson and Kamins (2006) found that a ‘severity threshold’ seems to exist, when the information about a company or brand is so negative, that even consumers who are committed to a certain brand cannot ignore the information anymore, resulting in decreased brand evaluation. The current study looks to expand on previous research and investigate whether such a threshold also exists when considering brand endorser information, thus looking into different amplitudes (i.e. degrees of severity) of information.
5 In short, the main goal of the present study is to answer the following question: How
does Valence, Amplitude, and Relevance of celebrity endorser information influence consumer brand evaluations?
1.1.3. Sub-Questions
Before the main research question can be addressed, a profound theoretical basis needs to be in place regarding the concepts that are treated in the present study. There has to be an understanding of how brands exist in the minds of consumers: how are these mental structures set up and how do they function when brand information is being processed by the consumer? Also, a thorough examination has to take place on what celebrity brand endorsement entails and what theoretical models there exist on celebrity brand endorsement. Finally, a broad insight needs to be gained on how positive and negative brand endorser information is perceived and processed by consumers in general and how this influences consumer brand evaluations.
1.1.4. Delimitations
First, the brands and celebrity brand endorsers that are incorporated in the present study are all sports-related: over the years, many events, both positive and negative, have occurred revolving around professional athletes, providing an abundance of scenarios to feat in an experimental setting.
Second, in contrary to some of the academic literature on the use of brand endorsers in advertising (Till & Busler, 2000, 1998; Kamins, 1990), the present study will not measure the degree of fit between brand and brand endorser. Instead, the current research will focus only on the processing of three different kinds of brand endorser information (i.e. varying in relevance, valence and amplitude, as clearly demarcated) by consumers, and the effects thereof on consumer brand evaluation.
6 Finally, to prevent a one dimensional measurement of consumer brand evaluation, the present research will not only examine consumer attitude towards the brand: purchase intentions will also be taken into account as a control variable.
1.2. Contributions
1.2.1. Theoretical Contributions
Existing literature has mainly focused on the general positive (Silvera & Austad, 2004; Kamins & Gupta, 1994) and negative effects of celebrity brand endorsers (Money, Shimp & Sadano, 2008; Till & Shimp, 1998) on consumer brand evaluations.
Recent studies have expanded the literature by looking into specific aspects of the celebrity brand endorser – brand - consumer relationship. These studies have, amongst other topics, the effects of negative brand endorser information on the endorsed product (White, Goddard & Wilbur, 2009) and the effects of negative information on brand endorser credibility and attractiveness (Thwaites, Lowe, Monkhouse & Barnes, 2012).
What has not been researched sufficiently however, are the possible effects of different sorts of information (i.e. varying in valence, relevance and amplitude) on consumer brand evaluations. By examining brand endorser information characteristics instead of merely focusing on particular brands or brand endorsers, the present research aims to add to the existing literature by providing an increased understanding of how consumers process different kinds of information in a branding context. In this way, consumers’ reactions to brand endorser information could be explained, not just for a specific brand of brand endorser, but in a wide variety of branding scenarios.
7 1.2.2. Managerial Contributions
Employing celebrity brand endorsers is a high risk (and high-cost) endeavor, as illustrated by the earlier mentioned celebrity endorser scandals. By knowing if and how variables such as relevance and amplitude of information influence consumer brand evaluations when confronted with celebrity endorser information, results from the present study could provide clear guidelines for managers on how to handle celebrity brand endorser issues.
In the case of negative information: when assessing what kind of celebrity scandal information leads to such a decrease in consumer brand evaluations that an early termination of the endorser’s contract is justified. On the other hand, the present study could also provide proof for keeping on a celebrity brand endorser, when it turns out that certain kinds of negative information do not impact consumer brand evaluations that severely.
When considering positive celebrity brand endorser information, the present study could provide insight into what kind of information leads to more of an increase in positive consumer brand evaluations. Another example: it could be that relevant brand endorser information (e.g. tennis player Novak Djokovic giving a ball boy his umbrella for shelter during a rainy match) increases consumer brand evaluations more than irrelevant information (e.g. Djokovic giving his umbrella to another person when it is raining while walking on the street). With this knowledge, managers could make better informed choices when deciding on leveraging celebrity brand endorser information in advertising, as they would be able to determine what kind of information would cause the highest increase in consumer brand evaluation.
8 1.3. Outline
In the introduction (Chapter 1), the relevance of the present study, as well as the most important research gap, research goal and research questions were stated. Subsequently, a literature review will follow, containing academic literature on associative networks, celebrity brand endorsers, and the evaluation of information with varying valence (Chapter 2). The hypotheses of the present study will be discussed in Chapter 3.
The stated research question will be addressed by conducting a quantitative experiment in the form of an online computer task. The way the experiment is set up and the procedure that is followed in order to carry out the experiment are described in the methods section (Chapter 4). Then, the results of the experiment are explained in Chapter 5. Moreover, in the discussion (Chapter 6), theoretical and managerial implications are discussed. Finally, the present study will be summarized, and the limitations of the present study and possible avenues for future research will be discussed in the conclusion (Chapter 7).
9 2. Theoretical Framework
The outline for the theoretical framework will be as followed: first, a definition is given for what a brand is. Subsequently, explaining how brands exist in the mind of consumers, associative networks and brand are described. After that, the mechanisms of celebrity brand endorsement are further examined, to find out which factors make the use of celebrity brand endorsers so desirable for branding and what are the important theoretical concepts the celebrity brand endorsement literature. Finally, the way that people process information of different valence, amplitude, and relevance is researched, specifically in the domain of consumer behavior.
In the theoretical framework, an overview will be given of relevant existing literature on the core topics of the present study. However, in order be able to discuss branding literature, a definition of what is a brand must be established first. A brand is ‘a name, term, sign, symbol, or design, or combination of them which is intended to identify the goods and services of one seller or group of sellers and to differentiate them from those of competitors’ (Kotler 1991; p. 442).
2.1. Associative Networks
In the last decades of the 20th century, cognitive psychologists have studied the workings of memory systems, and more specifically associative networks, intensively (Henderson, Iacobucci & Calder, 1998; Collins & Loftus, 1975; Anderson & Bower, 1972). In the literature, there is consensus that knowledge is represented in the human brain in the form of associative networks, built up from nodes (i.e. pieces of information like people, locations, or things), connected by associative links that vary in strength (Henderson et al., 1998; Keller, 1993). When considering the associative network context, brands and celebrity brand endorsers exist as nodes in the minds of the consumers (Till, 1998).
10 According to Ratcliff and McKoon (1988), a node is activated when an item is presented to the memory system that resembles the information stored in the node (e.g. if someone sees or thinks about the color red, they might think of a fire truck). Then, activation spreads through the associative network (Anderson, 1983), causing an increase in levels of activation for other nodes that correspond to the activated node (the case of seeing the color red that could go as followed: fire, fire truck, 911, police). The level of activation of a node is dependent of the proximity in the associative network: the closer a node lies to the node that was activated initially, the more it gets activated (e.g. continuing the earlier example: seeing the color red might activate ‘fire truck’ more so than it would ‘police car’, as a fire truck is red, and therefore closer to the initially activated node).
The strength of the link or association between nodes is determined by the number of experiences that one has had activating that link: the more experiences had, the stronger the link. (Henderson et al., 1998). Of course, in branding and advertising, these experiences are facilitated by marketing communications, through creating experiences where associations are constructed between the celebrity brand endorser and the endorsed brand. In this way, the two concepts become part of each other’s association sets: a collection of concepts which are related in a meaningful way to the targeted brand (Meyers-Levy, 1989). This mechanism for instance describes how people can see the celebrity brand endorser, and automatically think of the brand he/she endorses.
2.1.1 Brand Knowledge
Keller (1993), in his research on the concept of customer-based brand equity or CBBE (i.e. ‘the differential effect of brand knowledge on consumer response to the marketing of the brand’; p. 2), describes several factors that together form brand knowledge or ‘the cognitive representation of a brand’ (Keller, 2003; p. 596). The two main dimensions of brand knowledge are brand awareness and brand image. Brand awareness entails the strength of a brand node in the memory
11 of the consumer: the degree to which consumers can identify the brand in different circumstances. Two factors influence brand awareness: brand recall and brand recognition.
Brand recall is the ability of consumers to retrieve the brand name correctly from memory when given a cue about, amongst other types of cues, the product category, situations of product use, or the needs that are being fulfilled by the product. Brand recognition, on the other hand, involves the ability to confirm a previous exposure to the brand, when the brand is given as a cue. Using celebrity brand endorsers serves to increase the ability of consumers to recognize, as well as to recall, a brand (Kahle & Homer, 1985).
Brand image does not emphasize the strength of a brand node in consumer memory, it describes the content of the brand associations, which reflect the perceptions that a consumer has about a brand (Keller, 1993). There are many different types of brand associations. Attributes are the characterizing descriptive features of a product or service that make up the evaluation of the consumer of what the products or service, as well as what is concerned when acquiring them through purchase. Attributes can be product-related (i.e. the elements of a product that are crucial for its performance) or non-product-related (i.e. external product or service characteristics that are affiliated with purchase or consumption of the service or product).
Furthermore, benefits describe what consumers perceive that a product or service can do for them, and are divided into three types: functional (i.e. intrinsic benefits often corresponding with product-related benefits), experiential (i.e. what it feels like to use a product) and symbolic (i.e. extrinsic benefits mostly concerned with social approval and personal expression). However, the ability of a brand to differentiate itself does not solely depend on the type of brand association that consumers have. Most of all, brand associations in the mind of consumers should be favorable, strong, and unique (Keller, 1993).
12 According to Keller (2003) celebrity brand endorsers are able, as a secondary source of brand knowledge, to transfer the different types of associations from themselves to a brand (see Figure 1).
When connecting a source of secondary brand knowledge (e.g. the celebrity brand endorser) to a brand, Keller identifies three important factors that predict the degree to which the associations for the secondary source can be leveraged and linked to the brand. The first factor is the knowledge of the secondary source: what do people know about the secondary source of brand knowledge? The second factor Keller considers is the meaningfulness of the knowledge of the secondary source: to what extent is the knowledge of the secondary source relevant and meaningful for the brand? The third factor Keller describes is the transferability of the knowledge of the secondary source of brand knowledge: to what extent is it possible to link the knowledge of the secondary source to existing brand associations? When choosing a celebrity brand endorser, all three factors should be taken into account.
13 2.2. Celebrity Brand Endorsers
As described in the introduction of the present study, the use of celebrity brand endorsers is commonplace in today’s advertising industry. McCracken (1989) described celebrity brand endorsers as ‘any individual who enjoys public recognition and who uses this recognition on behalf of a consumer good by appearing with it in an advertisement’ (p. 310). In the current research, sports celebrity endorsers are examined in particular. Stafford, Spears and Hsu (2003) define a sports celebrity endorser as: ‘a famous athlete or coach who uses public recognition to recommend or co-present with a product in an ad’ (p. 13).
For example: repeatedly seeing Lionel Messi in Adidas advertisements, strengthens the link between the node containing the concept of Lionel Messi and the node containing the Adidas concept, so that when people are confronted with the one concept, the other will also be activated in their mind. This type of learning is called associative learning, or classical conditioning, and will now be further addressed.
2.2.1. Classical Conditioning
In his research, Till (1998) describes how classical conditioning can be adequately used to pair celebrity brand endorsers with brands, making them part of each other’s association sets. First introduced by Pavlov (1927), classical conditioning refers to a learning process where an unconditioned stimulus (US) that evokes a desired unconditioned response (UR), and a conditioned stimulus (CS) are paired repeatedly to elicit a desired conditioned response (CR). In advertising, brands and celebrity brand endorsers are paired together to transfer positive associations from one to the other (Till, Stanley & Priluck, 2008; Erdogan, 1999).
A sports-related example: imagine the famous Dutch sports commentator Jack van Gelder, who has been commenting on exhilarating sports events for years on end (i.e. making his voice a stimulus that evokes excitement), being used in an advertisement for a sports brand.
14 The exhilaration (UR) elicited by Van Gelder’s voice (US) might be associated with the advertised sports brand (CS), and therefore also elicited when people are presented the brand without van Gelder’s voice (CR). In order to make sure that the pairing of endorser and brand is successful, it is of great importance that the pairing is repeatedly shown to consumers (Till, 1998).
McSweeney and Bierley (1984) state that prolonged repetition of the US-CS pairing increases the likelihood that the occurrence of one stimulus forecasts the occurrence of the other stimulus in the brain, with the CS eliciting the same favorable response as the US. Also, it is easier to develop an associative link between a US (i.e. the celebrity endorser) and a CS (i.e. the endorsed brand) when the two have a good fit with each other. The significance of US/CS fit has also been addressed by literature on the meaning transfer-model (McCracken, 1986).
2.2.2. The Meaning transfer model
The meaning transfer model by McCracken (1986) can be used to explain how meaning transfers from celebrity brand endorsers to brands. Meanings exists in the culturally constituted world as a result of dominant cultural categories and principles. Celebrity brand endorsers can harbor a wide variety of meanings (e.g. gender, profession, personality, life style, social class etc.) (McCracken, 1989). Furthermore, celebrities appear to convey their meanings more clearly and richly than anonymous actors or models in advertising, because they have developed and repeatedly shown the meanings they embody in public.
For example: Dutch football player Jaap Stam could transfer the meaning of masculinity more easily and vividly than an unknown model or actor, as Stam has shown this meaning in public during televised football matches, where he – amongst other displays of manliness – once got his eyebrow stitched without anesthetics after a collision with another player, continuing play afterwards.
15 Advertising can transfer meaning from the celebrity brand endorsers to a brand (McCracken, 1989). In order to do so, a celebrity brand endorser should be chosen that embodies the meanings that are supposed to be transferred. Moreover, the advertising campaign must imply that there are major similarities between the endorsed brand and the celebrity brand endorser, so that consumers recognize meanings from the culturally constituted world – represented by the celebrity endorser - in consumer goods. Finally, meaning can transfer from consumer goods to consumers when they come into prolonged contact with the consumer good, for instance by purchasing and using a product.
Similarity appears to be an important criteria for transferring meaning from a brand endorser to the endorsed brand of product. Providing more insight into the optimal degree of similarity between advertisement and brand endorsers is research on the match-up hypothesis (Kahle & Homer, 1985).
2.2.3. The Match-up Hypothesis
The match-up hypothesis (Kahle & Homer, 1985) describes, that for brand endorsers to be truly effective, in terms of increasing purchase intentions and consumer product or brand attitudes, there has to exist a certain ‘fit’ or ‘congruity’ between the brand endorser and the endorsed brand or product. This fit stems from the fact that both the endorsed product and the brand endorser are part of the same cognitive consumer category in the mind of consumers (Loken, Barsalou & Joiner, 2008; Loken & Ward, 1990).
A consumer category is defined as a set of marketing entities (e.g. products, brands, events, etc.) that appear, in the eyes of the consumer, to be related in some way. These cognitive categories are used by consumers to categorize new information, in order to determine in which category a new piece of consumer information fits best (e.g. a tennis player is a better
match-16 up for endorsing tennis shoes than a basketball player, as tennis shoes and tennis players are both part of the cognitive category ‘tennis’).
Earlier research on the match-up hypothesis mainly focused on brand endorser attractiveness (Kamins, 1990; Kahle & Homer, 1985). The results of these studies show that using attractive endorsers for products related to attractiveness (e.g. razor blades) creates an increased ‘fit’ between the endorser and the product, ultimately leading to higher consumer product evaluation. Other match-up criteria that have been examined and proven to increase the match-up between endorser and endorsed product, include endorser expertise (Till and Busler, 2000; 1998), muscularity (Lynch & Schuler, 1994) and ethnicity (Ryu, Park & Feick, 2006).
Despite the match-up hypothesis being well researched and backed by various scholars, there are other theories that provide a different, quite contradictory perspective on the importance of fit between brand and brand endorser.
2.2.4. The Elaboration Likelihood Model
Petty and Cacioppo (1984) introduced the elaboration likelihood model in order to explain how people are persuaded through different routes of message information processing. A central and a peripheral route of processing information were distinguished. When people have the motivation (e.g. because the information at hand is self-relevant) and ability to think (e.g. not being distracted) about presented information, involvement is high. In high involvement conditions, the message is processed via the central route, with high cognitive effort and a focus on argument strength. When either the motivation or the ability to think about the information is lacking, involvement is low, and information is processed through the peripheral route, receiving less exhaustive cognitive consideration. In the case of low involvement, heuristics are often used by consumers (e.g. the more expensive something is, the better the quality must be).
17 Following a central route, taking into account the content and strength of arguments, can lead to attitude change that remains stable over time and more or less resistant to counterarguments (Petty & Cacioppo, 1983). Attitudes formed by following the peripheral route of persuasion are less resistant to these factors, as they are based on cues of message strength (e.g. number of arguments or argument source expertise), rather than a detailed examination of the actual argument strength.
Petty, Cacioppo and Schumann (1983) show that celebrity endorsers serve as a peripheral cue: in low involvement conditions (i.e. when people are either unable or unmotivated to think about the presented information), a celebrity endorser stating an argument increases attitude towards the endorsed product, as opposed to when a non-celebrity endorser is stating the message.
2.3. Processing of Information of Different Valence, Amplitude, and Relevance
As identified by Keller (2003), brand endorsers serve as secondary sources of brand knowledge, making them useful for enhancing brand image and brand awareness, by transferring associations from their association set to that of the endorsed brand. One big problem is that secondary sources of brand knowledge do not exist solely for transferring desired associations to a brand. Having association sets of their own, the issue is that not every association within this set is desirable from the perspective of the endorsed brand. Of course, brands would prefer celebrity brand endorsers to only transfer positive associations from themselves to a brand (Erdogan, 1999, Misra & Beatty, 1990). The truth is however, that negative associations can also be transferred, possibly damaging brands. In the following section, earlier research on the effects of the Valence, Amplitude, and Relevance of celebrity brand endorser information on consumer brand evaluations are discussed.
18 2.3.1. Valence of Information
Before explaining the effects of negative and positive brand endorser information on consumers, first, there has to be an understanding of the impact of information of different valence (i.e. positive or negative) in general. Regarding negative information, the literature describes a ‘negativity effect’ or ‘negativity bias’, finding that people think of negative information as more important than positive information when evaluating a target (Klein, 1996; Skowronski & Carlston, 1989). This negativity effect could stem from the perception that negative information is more diagnostic, or contains more information, than positive information (Maheswaran and Meyers-Levy 1990; Skowronski and Carlston 1989). Other research seems to confirm this idea. Ito, Larsen, Smith, and Cacioppo (1998) show that negative information is evaluated more strongly than positive information of the same fortitude. Also, negative information is better remembered than neutral information in long-term memory (Kensinger & Corkin, 2003).
In the context of brands, Ahluwalia et al. (2000) examined the influence of negative publicity on consumer brand attitudes. A moderating effect of consumer-brand commitment was found, where low commitment consumers showed greater change in their attitude towards the brand as an effect of negative publicity, than high commitment consumers. This means that low commitment consumers are more susceptible to the effects of negative brand publicity than high commitment consumers. For positive brand publicity however, a reverse effect was found, where highly committed consumers exhibited a greater amount of attitude change. The results that Ahluwahlia et al. (2000) have presented show that brand-consumer commitment can play an important role in the way that consumers evaluate brand publicity of varying valence.
Regarding celebrity brand endorsers, consumers pay more attention to negative information than positive information, encoding it better in memory and recalling the negative brand endorser information more easily than positive information (Ybarra and Stephan, 1996;
19 Folkes, 1988). Also, Till and Shimp (1998) and Amos et al. (2008) discovered that the strength of the associative link between the brand and brand endorser partly determines the impact of negative information on consumer brand evaluation: the stronger the associative link between the two entities, the more consumer brand evaluations decrease as a consequence of negative brand endorser information. Furthermore, Money, Shimp and Sakano (2006) report findings of decreased purchase intentions when negative brand endorser information is other-oriented as opposed to when brand endorser information is self-oriented.
2.3.2. Amplitude of Information
Despite the undeniable impact of negative celebrity brand endorser information, there are also factors that mitigate the effects of negative information on consumer evaluations regarding the brand. A study conducted by Einwiller et al. (2006) shows that the effects of negative information are moderated by consumer commitment to the company, making committed consumers less susceptible to change consumer company evaluation as a consequence of negative information. However, as mentioned in the introduction, a ‘severity threshold’ seems to be at play: when negative information is so severe, that even committed consumers cannot discard the information, resulting in decreased consumer company evaluations.
Extrapolating the results shown by Einwiller et al. (2006), consumers seem to react differently to brand information of varying amplitudes. Other studies appear to confirm this notion. Zhu and Chang (2013) for instance, have researched the effect of company founder image on corporate image. When a company founder, who can in some cases be considered as a company’s most important celebrity endorser (e.g. Richard Branson for Virgin or Steve Jobs for Apple) displayed unethical (i.e. negative) behavior, the perceived severity of the behavior determined the effect of the unethical behavior on company founder image (i.e. the higher the perceived severity of the unethical behavior, the less favorable the company founder image).
20 1946), which states that people want to remain consistent in their patterns of beliefs in liking and disliking other persons and inanimate objects, in order to achieve psychological balance. So, when the company founder acts unethical, and this is perceived as negative by consumers, people strive to be consistent and perceive the company itself as unethical – thus negative - as well.
2.3.3. Relevance of information
The effects of relevance of brand information have been researched in earlier academic literature. For instance, Broniarczyk and Alba (1994) show that brand preference ordering is partly determined by the relevance of the brand associations for the benefits sought in a particular product category. An example: when a consumer is looking to buy a tennis racket, he/she might want a racket that will last for the next three years. In this example, the consumer would prefer a brand for which he has durability associations (i.e. and is thus relevant for the benefit that he seeks in the product category) over a brand for which other, more irrelevant associations would exist.
In the brand crises context, Dawar and Lei (2009) have researched the influence of the relevance of a crisis on consumer brand evaluations. Crisis information was considered relevant when it appealed to key benefit associations for a brand (e.g. for the durable tennis racket brand in the previous example, a relevant brand crisis would be when a large quantity of rackets would break after only a few times of using them). In their research, Dawar and Lei (2009) found that relevance moderates the relationship between brand crises and consumer brand evaluation, where brand crises that are considered as relevant for a brand have a greater detrimental impact on consumer brand evaluations than brand crises that are considered to be irrelevant for the key benefit associations of a brand. When placing the earlier mentioned results into the context of celebrity brand endorsers, negative celebrity brand endorser information can be seen as a type of brand crisis, where similar effect might arise.
21 Following the line of reasoning of Dawar and Lei (2009) and taking into account the earlier mentioned research on brand-celebrity brand endorser fit (Loken et al., 2008; Loken & Ward, 1990), in the present study, information is considered as relevant when it fits into the shared consumer category that both the celebrity endorser and the brand are in, thus appealing to the key associations that exist for both the brand, and the celebrity brand endorser.
22 3. Hypotheses
The present study researches the mitigating and enhancing factors that influence the effect of celebrity brand endorser information on consumer brand evaluations. Specifically, the influence of different levels of valence, amplitude, and relevance was examined (see Figure 2 for all hypotheses).
Figure 2. Conceptual model displaying the relationships between valence, amplitude, relevance, attitude towards
the endorser, and attitude towards the brand.
As indicated in the theoretical framework, earlier research has proven that celebrity brand endorsers are capable of transferring associations from themselves to the brand they endorse (Keller, 2003; McCracken, 1989, 1986). Furthermore, it appears to be so, that the valence of celebrity brand endorser information, can affect the attitude consumers have towards the endorser, and via the endorser (i.e. through association transfer), the attitude towards the endorsed brand (Money et al., 2006; Till, 1998; Till & Shimp, 1998). However, endorser literature mainly focuses on the transfer of negative information, leaving the effects of positive information on brand attitude largely undiscovered.
23 To test whether information of both positive and negative valence can actually affect the attitude towards a brand through celebrity endorser attitude, Hypothesis 1 consists of two parts.
H1a. Attitude towards the endorser is more favorable when Valence is positive, as opposed to when Valence is negative
H1b. Mediating the relationship between Valence and Attitude towards the endorser, Attitude towards the brand is more favorable when Attitude towards the endorser is
favorable, as opposed to when the Attitude towards the endorser is not favorable.
Earlier research by Zhu and Chang (2013) and Einwiller et al. (2006) shows that amplitude is one of the possible mitigating or enhancing factors on the effect that valence has on attitude towards the endorsers. Perceived severity (of negative information) of brand endorser information can have a significant impact on the attitudes towards the brand endorser.
As proposed earlier, this can be explained by balance theory (Heider, 1946), which states that people strive to reach psychological balance through keeping consistent beliefs in liking or disliking people or objects (i.e. when people have negative beliefs about a brand endorser, they will also have negative beliefs about the brand). Following the reasoning of balance theory, the striving for consistent beliefs should also extend to positive beliefs, as balance in beliefs can be achieved both by liking or disliking someone or something. However, effects of positive information have not been properly examined in the celebrity brand endorser.
The effect of amplitude of positive information on attitude towards the endorser is to be researched in the present study, resulting in the formulation of the second hypothesis.
H2. The difference in Attitude towards the endorser between positive and negative Valence (H1a) is larger when Amplitude is high than when Amplitude is low
24 Previous research on the fit between brand and brand endorser (Till & Shimp, 1998; Misra & Beatty, 1990; Kamins, 1990; Kahle & Homer, 1985) has found that positive and negative associations can shift from brand endorsers to the endorsed brand. In other words, consumer attitudes towards a brand endorser influence the attitude consumers hold towards the endorsed brand. This is partly caused by the brand and brand endorser being part of the same cognitive consumer category (Loken et al., 2008; Loken & Ward, 1990).
In the present study, expanding the scope of earlier literature brand endorser information is taken into consideration. Brand endorser information is perceived as relevant when it fits within the shared cognitive consumer category of the brand and brand endorser (e.g. a football player endorsing for Nike football shoes would have a high fit, as they both fit in the category ‘Football’). The degree of relevance could determine the thoroughness of processing the information (Petty & Briñol, 2008; Petty & Cacioppo, 1979), thus influencing the degree to which the information influences brand and brand endorser attitudes. Examining the influence of informational relevance on the effect that attitude towards the endorser has on attitude towards a brand, lead to the formulation of the following hypothesis.
H3. The positive effect of Attitude towards the endorser on Attitude towards the brand (H1b) is stronger when Relevance is high than when Relevance is low
Finally, as discussed earlier, Einwiller et al. (2006) appeared to have found what appears to be a ‘severity threshold’: when information about a company is so negative, that even the most company-committed consumers cannot ignore the information (as they would do with moderately negative company information), resulting in decreased consumer company evaluation.
For instance: company-committed consumers for Nike could dismiss information that that is moderately negative as being necessary for the company (e.g. hundreds of layoffs
25 while a parting manager gets a million dollar bonus), not affecting the positive beliefs they have about the company Nike. However, when information gets too negative (e.g. Nike uses child labor and maintains horrible working conditions, causing children to die of exhaustion), even company-committed consumers can not dismiss this information anymore, affecting the positive beliefs they have about a company in a negative way.
In the present study, it is expected that high relevance of information enhances the effect of attitude towards the endorser on attitude towards the brand (H3). When translated to the present study, the severity threshold can be described as the amplitude of information being so high, that regardless of the relevance of the information, it will influence the effect of attitude towards the endorser on attitude towards the brand nonetheless, as was exemplified earlier by mentioning Nike athlete Oscar Pistorius shooting and killing his girlfriend. This information is not relevant for the shared category of Pistorius and Nike (i.e. Athletics), but is so severe that consumers’ beliefs about Pistorius, and therefore Nike, could be negatively affected. This line of reasoning leads to the fourth and final hypothesis.
H4. The interaction between Attitude towards the endorser and Relevance (H3) is weaker when amplitude is high than when amplitude is low.
26 4. Method
In order to test the hypotheses that were posed in the previous chapter, an empirical experiment has been conducted. A 2(Valence: positive/negative) x2(Amplitude: high/low) x2(Relevance: high/low) between subjects design was used. In this experiment, respondents were asked to read news headlines with information about celebrity brand endorsers (i.e. news scenarios), after which they had to fill out a questionnaire. However, before the main experiment could be carried out, a wide variety of proposed news scenarios were developed and pre-tested.
4.1. Stimuli development
In order to create stimuli that could adequately portray the various manipulation goals of the experimental conditions, a focus group (n = 5) was conducted. In this focus group, combinations of celebrity endorsers and brands were created, as well as information about the celebrity endorser (scenarios), displaying on of the experimental conditions. During the focus group, a set of criteria were determined that would have to be satisfied in order to optimize the stimuli for experimental use.
For instance, the brand, the celebrity brand endorser, and the endorsement deal between the two should be well known by the general public, in order to control for indifference about the celebrity brand endorser information (i.e. if one does not know an endorser/brand, one might dismiss the information about the endorser/brand as uninteresting, as they do not have strong existing beliefs about the endorser/brand). Also, neither the brand nor the endorser should elicit an extremely positive or negative attitude, as this would influence the way that respondents perceive information about the endorser/brand (i.e. if someone is overly fond of a certain brand, they might dismiss negative information as unrealistic or mere slander). Furthermore, the information in the developed scenarios should be hypothetical (preventing respondents from drawing from their earlier cognitions about the sketched scenario), as well as realistic
27 (preventing people from not paying attention to the endorser information as a result of thinking that this scenario would never occur in real life).
Finally, it was decided that during the pre-test, at least one set of scenarios should contain a male celebrity endorser and one should contain a female, to control for gender effects. In Figure 3, all scenarios that were developed are displayed. As can be seen, there were 2 sets of scenarios considering football, one set about tennis, and one about pop stars.
Figure 3. Developed experimental scenarios
From the developed scenarios, the scenarios regarding Robin van Persie endorsing Adidas and Serena Williams endorsing Nike were selected. Members of the focus group agreed that both van Persie, as well as Serena Williams were likeable, but at the same time also had a darker, more angry side. This made van Persie and Williams suitable for playing the lead role in positive as well as negative information. The pop music scenarios were not selected for a variety of reasons. First, it came to light that 40 % of the focus group (i.e. two out of five respondents) did not know who Jay-Z was. Second, 60% of the focus group (three out of five respondents) was not aware that Jay-Z had an endorsement deal with Samsung. Third, the initial
28 focus of the present study was on sports, making van Persie and Williams more appropriate candidates for being examined in the pre-test.
4.2. Pre-test
The scenarios that were shown in the main experiment were first examined in a pre-test, to see which scenarios would fit best with the different experimental conditions. Eighty-nine people participated in the pre-test. The pre-test was conducted both on paper and online using survey platform Qualtrics. Each participant assessed four randomly presented scenarios out of a total of twenty-four scenarios (see Figure 3 for synopsis, and Appendix A for full text). The scenarios consisted of information about either football player Robin van Persie, endorser for sports brand Adidas, or professional tennis player Serena Williams, endorsing the Nike sports brand. Two celebrity endorsers and two brands were tested to control for brand and/or endorser specific effects.
Before evaluating the various scenarios, participants were first asked to describe the endorsers and the brands in a maximum of five words, as a measure of endorser and brand recognition. Also, participants’ attitude towards the celebrity endorser, as well as the brand, were queried on a 7-point Likert scale (ranging from 1 ‘very negative’ to 7 ‘very positive’), to control for participants that feel extremely positive/negative about either the endorser or the sports brand. After reading each scenario, participants were first asked to evaluate the perceived realism of the information in the scenarios, as the degree to which participants deem the information in the scenarios realistic might influence their responses on the items regarding relevance, amplitude, and valence. Subsequently, respondents were asked to fill out a questionnaire that consisted of eleven self-developed items (see Appendix C). All items were answered on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree). The questionnaire consisted of the following items.
29 The independent variables relevance, amplitude and valence were measured using self-developed scales, as there were no pre-existing scales that covered the content of the independent variables adequately. Relevance was measured with four items (‘I think this information is related/connected to/relevant for/has something to do with the brand Nike’, α = 0.88). The items were developed to reflect the semantic meaning of relevance. Amplitude was also measured using four items (‘I think this information has a big impact/big consequences/ is important’ and ‘I am impressed by this information’, α = 0.83) developed to give respondents the possibility to mention the levels of magnitude of the information, and to what extent they are impressed by the information (i.e. is the information a ‘big thing’ or not). Valence was measured using two items (‘I think this information is favorable/positive’, α = 0.93).
When selecting the best fitting scenarios for the different experimental conditions, the scales for Valence, Amplitude, and Relevance were leading, as the scores on these scales were most representative for the structuring of the experimental conditions. To control if the various scenarios are seen as intended, negative and/or low amplitude/relevance scenarios should preferably score significantly below the midpoint of the 7-point Likert scale (i.e. below 4). Vice versa, positive and/or high amplitude/relevance scenarios should score significantly above the midpoint of the 7-point Likert scale (i.e. above 4).
In Table 1, the eight scenarios are displayed, which showed the most promise for fulfilling the various experimental conditions (for the results of all twenty-four scenarios, see Appendix B).
30 Table 1
Selected scenarios for main experiment, including mean scores and standard deviations on Valence, Amplitude, and Relevance.
First of all, results show that, for six of the eight experimental conditions, scenarios containing Robin van Persie and Adidas fitted better with the manipulation goals for the main experiment, as opposed to scenarios containing Serena Williams and Nike. Furthermore, both combinations of celebrity endorser and brand met the criteria set beforehand (familiarity above 90% and mean attitude towards endorser and/or brand neither very positive (M > 6,0) nor very negative (M < 3,0)). This makes the two combinations of endorser and brand interchangeable in that regard (see Table 2). With these observations in mind, the decision was made to only use scenarios with Robin van Persie for the main experiment.
Relevance
Amplitude Low High
Negative Low Scenario 1 (n = 10) Scenario 2 (n = 11) Valence 1.7 (0.75)* Valence 2.04 (1.08)* Amplitude 2.35 (1.40)* Amplitude 2.29 (1.05)* Relevance 2.09 (1.07)* Relevance 1.61 (0.78)* High Scenario 3 (n = 13) Scenario 20 (n = 9) Valence 1.34 (0.63)* Valence 1.44 (0.73)* Amplitude 5.02 (1.38)* Amplitude 4.63 (1.58)
Valence Relevance 3.4 (1.58) Relevance 3.51 (1.56)
Positive Low Scenario 13 (n = 12) Scenario 14 (n = 12) Valence 5.38 (1.07)* Valence 5.2 (1.68)* Amplitude 3.45 (1.67) Amplitude 3.17 (1.55) Relevance 2.44 (1.64)* Relevance 3.23 (1.99) High Scenario 23 (n = 10) Scenario 8 (n = 7) Valence 5.7 (0.82)* Valence 5.29 (0.81)* Amplitude 4.35 (1.12) Amplitude 4.64 (1.43) Relevance 3.53 (1.18) Relevance 4.35 (1.72)
31 Table 2
Familiarity and Attitude towards brand and brand endorser
Familiarity % Attitude M (SD)
Robin van Persie (n=64) 93.8 4.30 (1.10)
Serena Williams (n=25) 96.7 5.02 (1.00)
Adidas (n=64) 98.4 5.02 (1.00)
Nike (n=25) 100 5.47 (0.90)
While further analyzing the results of the pre-test, it became apparent that scores on the relevance scale never rose significantly higher than the midway point of the 7-point Likert scale. Thus, participants did not perceive one of the twenty-four scenarios as highly relevant. These results most likely stem from the fact that the questions all referred to being relevant for the Adidas/Nike brand, instead of being relevant for the shared category that Robin van Persie/Serena Williams and Adidas/Nike have: Football/Tennis. This means that the questions were not valid for examining Relevance as it was defined in the present study.
This insight led to a re-evaluation of the items making up the relevance scale for the main experiment, of which the exact formulation will be further discussed in the method section. In short, the items within the relevance scale were reformulated in such a way that they referred to the relevance of the scenario for the product category that both the brand Adidas, and celebrity Robin van Persie Williams are a part of: football. This revision of the relevance scale also had consequences for which scenarios would be best usable for the main experiment. For instance, scenario 2 was selected over scenario 18 (which initially had better fitting scores on the valence, amplitude, and relevance scales) as the researchers argued that scenario 2 was more relevant for football, and therefore more likely to score high on relevance in the main experiment with the re-imagined relevance scale.
32 4.3. Main experiment
4.3.1. Research Design
For the main experiment, a 2 (Valence: positive/negative) x2 (Amplitude: high/low) x2 (Relevance: high/low) between subjects design has been used. In the eyes of the researchers, a between subjects design (i.e. where respondents rated one scenario only) was the best fitting for the present study, as it prevents respondents from comparing various scenarios with each other while filling in the questionnaire, which could affect the way people look at the various levels of the independent variables (e.g. a positive scenario might be rated differently if it is judged after a negative scenario, than if it would be judged on its own).
4.3.2. Procedure
The main experiment was programmed using the online survey platform Qualtrics. The experiment was fit to be conducted on personal computer, as well as smartphones and tablets.
First, participants would receive a welcome message that made clear that they were about to read news headlines about a famous athlete. After this welcome message, one of the nine scenarios (i.e. eight experimental conditions and one control condition, see Figure 4 for a synopsis, and Appendix D for full text) would be randomly selected to appear on the screen. After reading the news headlines, a statement followed that asked participants to fill in the questionnaire based on the information that they had just received in the form of the news headlines. Subsequently, a questionnaire was filled in (see Appendix E). All questions were answered by using a 7-point Likert scale. After answering all questions, respondents were shown a last message thanking them for their participation. Also, a disclaimer was displayed, stating that all scenarios were fictional and created for research purposes only.
33 Figure 4. Chosen scenarios for main experiment. The control scenario consisted of Robin van Persie visiting
Paris during his football free weekend with his wife. 4.3.3. Measures
Dependent variables
Endorser/Brand Familiarity. To measure endorser familiarity, a single item was used
(i.e. ‘I am familiar with Robin van Persie’) with a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree). A similar item was used for measuring brand familiarity (i.e. ‘I am familiar with the Adidas brand’).
Endorser/Brand Attitude. Endorser attitude was measured using a modified version of
the three item scale also used in earlier celebrity brand endorser research by Till & Busler (1998). The three items (α = 0.94): ‘My attitude towards Robin van Persie is positive/favorable’ and ‘I like Robin van Persie’ were scored on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree). The same 3 item scale (α = 0.95) was used for measuring attitude towards the Adidas brand (‘My attitude towards the brand Adidas is positive/favorable’ and ‘I like the brand Adidas’).
Relevance Amplitude
Low High
Low Walking straight into night club without waiting Verbally abused the referee Negative
High Sex party with minority girls
Verbally and physically abused referee, breaking jaw with
punch Valence
Low Buying pizza for homeless man
Kicked out ball for opposing team's injury treatment Positive
High Donating money and building houses for homeless people
Scoring winning goal Champions League final w/
