Bad Girl Image versus Good Girl Image
How celebrity endorsers with a bad girl image affect brand recall,
brand attitude, and attitude towards the advertisement, when the
brand is congruent with their image
Lisanne van Beurden Master Thesis
Department of Communication Science University of Amsterdam
Bad Girl Image versus Good Girl Image:
How celebrity endorsers with a bad girl image affect brand recall, brand
attitude, and attitude towards the advertisement, when the brand is
congruent with their image.
Master Thesis
Lisanne van Beurden, 10726268
Thesis Supervisor: Dhr. Dr. Aart S. Velthuijsen University of Amsterdam
Abstract
In this study it is investigated to what extent celebrities with a bad girl image
positively influence attitude towards the brand, attitude towards the advertisement and brand recall, when they endorse a congruent brand, in comparison with celebrities with a good girl image that endorse an incongruent brand. The findings of this research may provide brands with criteria to select effective female celebrity
endorsers. Research about the match-up hypothesis was investigated, while research about the source credibility model was also taken into consideration. Four conditions were created, making this a 2 (celebrity image: bad versus good) x 2 (brand: fit vs. non-fit) between-subjects factorial design. In total 124 respondents participated in the online experiment. The results of the experiment show that a celebrity endorser with a bad girl attitude is not necessarily more beneficial for one’s brand than a celebrity with a good girl image. However, in the case of an advertisement that shows a brand that is congruent with the celebrity, a bad girl image is more beneficial than a good girl image celebrity that does not fit with that brand in terms of attitude towards the advertisement.
Content
1. Introduction……….….…...5
2. Theoretical framework………..……..8
2.1 Celebrity Endorsement……….…...8
2.2 Source Credibility Model……….…...8
2.3 Attitude Towards the Advertisement and Brand………...…….……11
2.4 Bad girl image vs. Good girl image………...…….…...12
2.5 The Product Match-Up Hypothesis: bad brand, bad celebrity?...14
2.6 Brand Recall and The Product Match-Up Hypothesis……….……..16
3. Method………...…..19
3.1 Type of method………...19
3.2 Participants and Sampling Method………19
3.3 Procedure………...19
3.4 Operationalizations dependent variables………...22
4. Results………...24
4.1 Sample Characteristics……….…..…24
4.2 Internal Consistency Dependent Measures………...….24
4.3 Results Main Analyses Experiment………..….24
5. Discussion………....30
5.1 Summary of Results………..….30
5.2 Limitations and recommendations……….…31
5.3 Conclusions and Practical Implications……….32
1. Introduction
“Well-behaved women rarely make history” - (Ulrich, 1976, p. 20).
The marketplace for various brands is cluttered, making it hard to gain attention from the right audience for one’s advertisement (Pieters, Warlop & Wedel, 2002).
Therefore, advertisers are always looking for ways to get attention from consumers. In order to reach this goal, organizations have to design strategies that create an advantage and make them stand out from other competitors (Erdogan, 1999). A well-known strategy is using celebrities in advertisements to endorse the products, which is considered a highly effective way to gain consumer interest in a cluttered marketplace (Hung, 2014). According to Erdogan (1999) the use of celebrities is not a new
occurrence; it has been a phenomenon since the late 19th century. An example from the usage in the early days is Queen Victoria in an advertisement with Cadbury’s Cocoa.
According to Edwards and La Ferle (2009) using celebrity endorsers in an advertisement is not without risk, given the extraordinary lifestyles they live. Besides bringing popularity to a brand, celebrity endorsers can also bring negativity to a brand. Moreover, with the fast-moving, paparazzi-infested media, it is easier to find negative information about a celebrity (Edwards & La Ferle, 2009). Previous research states that when a celebrity behaves negatively and controversial, this reflects badly on the brand that is endorsed (Amos, Holmes & Strutton, 2008; Edwards & La Ferle, 2009; Louie, Kulik & Jacobsen, 2001). Therefore, companies carefully select their endorsers to be certain that no bad publicity is reflected on their brand. However, sometimes brands do pick the “wrong” celebrity, as the endorser can get involved in undesirable events that might harm the brand. In this case, the company often stops the contract they have with the celebrity, to prevent negative associations of the publicity to transfer to the brand (Akturan, 2011).
Interestingly, there are celebrities who are behaving in a questionable way and take part in negative events, yet are still chosen as celebrity endorsers. One would expect that brands would not take any risk by closing a contract with celebrities who have a high risk of taking part in a scandal anytime soon. These so-called “bad” celebrity endorsers have a “rough” and “tough” reputation and are not hiding their bad
behaviors. Instead, it becomes a part of their image. It is probably the case that certain companies choose these celebrities because of this bad image.
Even though one can find a lot of research on the effects of celebrity endorsers who take part in a negative event or scandals (Akturan, 2011; Bartz, Mochanov & Stork, 2013; Dean, 2004; Fong & Wyer, 2012; Gupta, 2009; Knittel & Stango, 2013), research about celebrity endorsers who are chosen with a bad image is scarce.
Furthermore, according to Amos et al. (2008) there is no research on celebrities who portray this bad image and their effectiveness as celebrity endorsers, but they claim this research is needed. These sort “anti-celebrities” may transfer a certain image of danger, risk or toughness, which may scare away some consumers, and appeal to others (Amos et al., 2008). A reason why these celebrities might be effective is that these celebrities have a big following because of their controversy, and “bad publicity is better than no publicity at all” (Leeds, 2010, p.807).
The match-up hypothesis states that for effective advertising, the image of the celebrity should convey congruent messages as the brand/product (Edwards & La Ferle, 2009; Forkan, 1980; Kahle & Homer, 1985; Kamins, 1990). According to Misra and Beatty (1990) the perceived ‘fit’ between brand characteristics and
celebrity image (e.g. relevant characteristics as seen by the consumers) determines the match between celebrity and brand. Furthermore, the results by Misra and Beatty (1990) showed that indeed the advertisements with congruency between brand and celebrities were higher, specifically on the aspects of brand recall and affect. Certain brands may have a fit with celebrities who have a tough image, which might explain the fact that these celebrities are chosen to endorse them. Celebrities who have this bad, tough image may convey the messages that a certain brand wants to convey (e.g. being a rebel, disobedient or naughty, not following the rules). Brands such as Harley Davidson (slogan; American by Birth, Rebel by Choice) or Diesel (slogan; Be Stupid) would most likely choose a celebrity who has a tough, badass image as these
characteristics match with their brand.
More research on celebrity endorsers who portray a bad image is needed, as it has never been researched, while it is relevant to know if these celebrities are
effective endorsers. Therefore, this research will compare celebrity endorsers with a bad girl image to celebrity endorsers with a good girl image and look at their
effectiveness as endorser, while taking the match-up hypothesis and source credibility model in mind. Furthermore, to research the effectiveness of these celebrities, it will
be measured how they impact the consumers’ brand recall, attitude towards the brand and attitude towards the advertisement. To conceptualize credibility, research about the source credibility model is analyzed. Research based on this model contends that the effectiveness of a message depends on trustworthiness, expertise and
attractiveness of a celebrity endorser (Erdogan, 1999; Pornpitakpan, 2004).
This research only focuses on female celebrity endorsers with a bad image, as it is expected that the results for male and female celebrity endorsers would have different outcomes. It is generally assumed by marketers that identification is a
significant component in the persuasion process, thus endorsers who are similar to the target group (e.g. same age, same gender) are more likely to influence the behavior of the audience (Fraser & Brown, 2002; Hsu &McDonald, 2002; Kelman, 1968; Wells, Burnett & Moriarty, 1989). Naturally, there are male celebrities with a “bad” image as well, however, they would not match with the same products as female celebrities (e.g. make-up, female clothes or lingerie). Furthermore, when only comparing female celebrities with a bad and good girl image, the research is demarcated. The findings of this research may provide brands with criteria to select effective female celebrity endorsers. This brings us to the following research question:
RQ: To what extent are celebrity endorsers who have a bad girl image positively
influencing attitude toward the brand and advertisement and recall of the brand, when they endorse a congruent brand, compared to celebrity endorsers who have a good girl image who endorse an incongruent brand?
2.0 Theoretical framework
2.1 Celebrity Endorsement
McCracken (1989, p. 310) defines a celebrity endorser 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.” Thus, the term celebrity does not only encompass movie stars, but also individuals from the world of sport, music and so forth. According to McCracken (1989) the effectiveness of a celebrity endorser depends on the meanings that the celebrity brings to the endorsement process. This theory is called the “meaning transfer” perspective, whereby the meanings that reside in the celebrity move from the celebrity to the product, and from the product to the consumer (Erdogan, 1999; McCracken, 1989). The meanings that celebrities contain are endless; there are distinctions of gender, age, status, and personality and lifestyle types (Erdogan, 1999).
There are many theories to when a celebrity endorser can be considered effective (Amos et al., 2008; Choi & Rifon, 2012; Edwards & La Ferle, 2009; Silvera & Austad, 2004). Many scholars have constructed various models that could be used for choosing an effective celebrity endorser. This paper focuses on two of those models, namely the source credibility model and the match-up hypothesis (Kamins, Kahle & Homer, 1985; Kamins, 1990), and uses these models to research the effectiveness of celebrity endorsers with a bad girl image, while comparing them to celebrities with a good girl image.
2.2 Source Credibility Model
According to McGuire (1978) the component source is comprised out of three main aspects: attractiveness, power, and credibility. Furthermore, the dimensions of the source credibility model commonly consist of expertise, attractiveness and
trustworthiness (Erdogan, 1999; Pornpitakpan, 2004).
Hovland and Weis (1951) were one of the first authors to study the credibility of the communicator and its influence on the effectiveness of the communication. One of the important findings of Hovland and Weiss (1951) was that high credibility sources produced more change than low credibility sources. Various other researchers reported the same finding that a highly credible source is more persuasive than a low credible source (Amos et al., 2008; Horai, Naccari & Fatoullah, 1974; Johnson &
Izzett, 1969; Johnson, Torvicia & Poprick, 1968; Kelman & Hovland, 1953; Lirtzman & Shuv-Ami, 1986; Watts & McGuire, 1964). However, it accounts for specific circumstances when a highly credible source is more persuasive than a low credible source. For instance, in commercial settings, sources that are more credible are more persuasive on brand and product attitudes than those that have less credible
dimensions (Fireworker & Friedman, 1977; Friedman & Friedman, 1979; Mowen & Brown, 1981).
Furthermore, it was also found that the credibility effect was reasonably strong when there was the lowest direct relevance and prior knowledge, but was weak and insignificant when the topic was very relevant and there was prior knowledge. According to Biswas, Biswas and Das (2006) consumer knowledge consists out of two major components: familiarity and expertise. Consumers with high product knowledge have a base knowledge with information about the product, which makes it easer for them to use this knowledge to make decisions (Biswas et al., 2006;
Cowley & Mitchell, 2003). Moreover, these consumers are less prone to be influenced by celebrity endorsers in their decision-making. However, when consumers have low product knowledge, they are more likely to use peripheral cues to analyze possible risks with a product purchase (Biswas et al., 2006; Rao & Monroe, 1988). Therefore, it is useful to consider the Elaboration Likelihood Model (ELM; Petty & Cacioppo, 1981; Petty & Cacioppo, 1986), which is a general theory of attitude change, and provides a framework for organizing, categorizing and understanding the basic processes that underlie the effectiveness of persuasion. Petty and Cacioppo (1986) came with the conclusion that there are two different routes to persuasion: central and peripheral. The central route to persuasion results from careful and thoughtful
consideration of the arguments and information that is presented and the peripheral route occurs when the receiver agrees with simple cues in the persuasion context that induced change, rather than agreeing with the message based on the strength of the arguments or ideas of the information (Petty, 1977; Petty and Cacioppo, 1986). From the ELM (Petty & Cacioppo, 1986) it can be argued that people use credible sources as peripheral cues when they have little product knowledge, when they are not very involved with the product, or when they don’t have time or attention to think.
There is also research that report different findings regarding the source credibility model and the effects of expertise, attractiveness and trustworthiness (Cantor, Alfonso & Zillman, 1976; Hovland & Mandell, 1952). However, according
to Pornpitakpan (2004), who has reviewed empirical evidence that has been published since 1950 from related disciplines regarding source credibility effects on persuasion, the main results show that higher source credibility persuades more in terms of attitude and behavioral measures than a source low in credibility. This is a very relevant finding, as this means that source attractiveness, expertise and
trustworthiness exercise a lot of influence on celebrity effectiveness in terms of influencing attitude.
The next few paragraphs define the components of the source credibility model, namely source attractiveness, source expertise and source trustworthiness and go more into depth in the research that has been conducted of these concepts.
2.2.1 Source Attractiveness
Most advertisements portray people who are generally perceived to be attractive (Erdogan, 1999). When referring to attractiveness, not only the physical attractiveness of the celebrity is meant, but also other characteristics such as intellectual skills, talents, personality and style (Amos et al., 2008; Erdogan, 1999). Physically attractive celebrities positively influence the products they endorse. Moreover, research shows that attractive celebrities have a more positive impact on the endorsed products than less attractive celebrities (Amos et al., 2008), and physical attractiveness of the source has a significant effect on opinion agreement and liking (Horai et al., 1974). Maddux and Rogers (1980) found that there was interaction between source attractiveness and source credibility, whereby the source that was an expert had more influence when he was unattractive than the non-expert that was attractive. Thus, physical attractiveness makes little difference when a source is highly expertise on the subject (Horai et al., 1974; Joseph, 1977). However, when the source is not an expert, consumers agree more with a highly attractive source than with medium or low-attractive source (Joseph, 1977). Furthermore, according to Yoon, Kim and Kim (1998) the perceived attractiveness of the endorser has more impact than expertise and trustworthiness on the attitude towards the advertisement.
2.2.2 Source Expertise
Source’ expertise is widely defined as the extent to which the source is perceived to make correct and valid assertions (Erdogan, 1999; Hovland, Janis & Kelley, 1953). However, this definition is still quite limited, as a person who makes few correct assertions about a certain field is not necessarily an expert. When a dentist knows a
person is having a heart attack and claims he needs surgery, even though this assertion can be true, one would not consider this dentist an expert in the field of cardiology. Therefore, adding to the above definition, it is also relevant to consider that expertise depends on factors such as training, experience and ability (Birnbaum & Stegner, 1979).
According to Braunsberger (1996) source expertise will lead to more positive attitudes to the endorser and the advertisement, when a source is high in expertise, in comparison to a source low in expertise. Celebrity expertise positively influences celebrity effectiveness (Amos et al., 2008). Thus, when a celebrity is considered an expert, this has a positive impact on the products the celebrity endorses. The
consumers’ action in response to the recommendations by the celebrity varies directly with the perceived level of the celebrity’s expertise and the agreement of the receiver with these recommendations (Amos et al., 2008). Moreover, when a celebrity is seen as an expert, consumers show more agreement with the recommendations by that celebrity than consumers exposed to a celebrity who was not considered as an expert.
2.2.3 Source Trustworthiness
Trustworthiness is defined as how much confidence is placed in the source’s intent to convey the assertions she/he considers most valid, and the degree to which an
audience perceives the assertions that are communicated to be the ones that the source considers to be valid (Amos et al., 2008; Erdogan, 1999; Hovland et al., 1953).
Furthermore, trustworthiness refers to the believability, integrity and honesty of the endorser. A highly trustworthy celebrity with a highly opinionated message can create an effective attitude change, and it is suggested that trustworthiness is an important predictor of celebrity endorser effectiveness (Amos et al., 2008; Erdogan, 1999).
According to Yoon et al. (1998) trustworthiness is the only component of the source credibility model that significantly impacts the attitude towards the brand and brand beliefs.
2.3 Attitude Towards the Advertisement and Brand
A person’s attitude is defined as his/her evaluation of people, objects, ideas or any entity in question, such as an advertisement, an endorser or brand (Eagly & Chaiken, 1998; Fishbein & Ajzen, 1977; Petty, Wegener & Fabrigar, 1997). Moreover, attitude consist of three components that together form this evaluation, namely; 1) an affective component, consisting of one’s emotional reactions toward the attitude object, 2) a
cognitive component, consisting of one’s thoughts and beliefs about the attitude object and 3) a behavioral component, consisting of one’s actions or observable behavior towards the attitude object (Crites, Fabrigar & Petty, 1994; McGuire, 1985; Zanna & Rempel, 1988).
When the goal of advertising is to change positive attitudes towards the ad and the brand, the best indicator of advertising effectiveness is creating a positive
emotional response to the ad (Aaker & Stayman, 1990; Brown & Stayman, 1992; Goldsmit & Lafferty, 2002). The positive emotional response of liking an
advertisement is positively related to brand attitudes and brand-related cognitions, brand attitudes and purchase intentions (Brown & Stayman, 1992; Stone, Besser & Lewis, 2000). Advertisers assume that an affective appeal of a certain stimulus/source (such as a celebrity endorser) will transfer to the attitude toward their brand (Kim, Lim & Bhargava, 1998). This affect transfer can be viewed as representing the peripheral route to persuasion in the ELM by Petty and Cacioppo (1981).
2.4 Bad girl image vs. Good girl image
As research about bad girl celebrities is scare, it is hard to define this concept of celebrities with a “bad girl image”. The only research that shed some light on a definition is Amos et al. (2008); they define celebrities that portray the bad girl image as anti-celebrities, and claim that these transfer a certain image of danger, risk or toughness to one’s brand or products. Therefore, in this research when there is referred to a female celebrity endorser with a bad girl image, it is meant that this celebrity has had certain “negative” experiences (e.g. fighting) or acts in a “bad” way, and is seen as a badass, tough and rough woman/girl. Thus, with a bad girl image it is not necessarily meant that this celebrity has a “bad image” in the sense that they are portrayed as bad people, but that this celebrity is seen as a badass, strong girl, someone who is not easily messed around with. Of course, some consumers might like these “bad image” celebrities, while others might perceive these celebrities as bad people, but this is merely an opinion.
When an undesirable event with a celebrity occurs, this might also transfer to the celebrity and thus can change the effectiveness of the celebrity as an endorser (Louie, Kulk & Jacobson, 2001). Furthermore, negative information about an
endorser can damage product evaluation through this meaning transfer between brand and celebrity (Till & Shimp, 1998). However, in the case of a female celebrity who is
chosen because of her bad girl image, this will probably work differently. The tough, cool image she has, and these meanings that reside in this image, will be transferred to the product/brand. A consumer who likes this image and wants to be just as tough and cool, would consider buying this product, causing these meanings to be transferred to the consumer.
Thus, it is a tricky thing; certain people might consider celebrities with a bad girl image as untrustworthy and/or dishonest. This is due to the fact that the, in some people’s opinion “unethical”, actions that bad girl celebrities might undertake can serve as a cue to the consumer, which can trigger a process of categorization. Hereby the celebrity will be assigned to a specific trait category, such as “bad”, untrustworthy and/or dishonest (Gupta, 2009). When considering this theory, it could be that
celebrities who have a bad girl image will be categorized as not being effective endorsers. However, this process of categorization can also go the other way around. Some consumers may like this bad, tough and daring behavior. When this is the case, the celebrity will be assigned to a specific trait category, such as “cool” and “badass”.. Thus, the reason for firms to choose these celebrities is because they want the
meanings that reside in these celebrities to transfer to their brands/products (McCracken, 1989). Consumers may want to be like these celebrities, and buy the brand in order to have the meanings that reside in the celebrity and products to transfer to them. Thus, they have a positive attitude towards the celebrity, and thus towards the brand and advertisement.
On the other side of the comparison are celebrities who have a good girl image. Celebrities with a good girl image are mostly seen as innocent, nonthreatening and sympathetic (Meyers, 2009). When consumers are assigning categories with celebrities with a good girl image, the specific trait category would probably be positive, such as “honest” or “sympathetic” (Gupta, 2009).
It is certainly not the case or expectation that celebrities with a good girl image are negatively impacting the consumers’ attitude toward the brand and advertisement. It can be expected that both a celebrity with a good girl image and a celebrity with a bad girl image positively impact the brand attitude, as they both have positive traits. However, celebrities with a bad girl image are expected to impact the attitude of the consumers even more positively than celebrities with a good image, as they are more noticeable. Because the celebrities with a bad girl image are generally more outspoken, and because of their “bad” behavior more striking, they are not easy
to ignore. Celebrity with a good girl image are expected to be perceived as less outspoken, maybe even boring by some, thus it is expected that the consumers’ attitude towards these celebrities is more neutral than towards celebrities with a bad image, leading to a less positive attitude towards the ad and brand. Therefore the following hypothesis is created, in the context that the consumer is female and between 18 and 22, and is interested in celebrities with a bad girl image.
H1: Celebrity endorsers who have a bad girl image will positively impact consumers’ attitude toward the brand and advertisement, in comparison with celebrities with a good girl image.
However, it may be important to consider whether the celebrity endorser and brand match together when deciding whether the celebrity endorser is effective (Cooper, 1984; Forkan, 1980; Kahle & Homer, 1985; Misra & Beatty, 1990). Therefore, in the next paragraph, research about the match-up hypothesis is discussed to further
hypothesize the effectiveness of a celebrity endorser with a bad girl image.
2.5 The Product Match-Up Hypothesis: bad brand, bad celebrity?
According to various practitioners of advertising a match-up between the
characteristics of the spokesperson and the characteristics of the brand/product is of importance to improve advertising effectiveness (Cooper, 1984; Forkan, 1980; Kahle & Homer, 1985; Misra & Beatty, 1990). This idea of endorser/product fit has been labeled the up hypothesis (Kahle & Homer, 1985; Kamins, 1990). The match-up hypothesis suggests that the messages conveyed by the image of the celebrity and the image of the brand/product should be congruent.
Research has shown that when there is a high congruency between brand and celebrity endorser, it leads to a greater believability compared to an advertisement with a less congruent match (Erdogan, 1999; Silvera & Austad, 2004). Moreover, the celebrity comes across as more believable and credible, which is beneficial for the celebrity, but also for the brand (Keel & Nataraajan, 2012). When there is no
congruency, consumers might think that the celebrity is only endorsing the brand for the money and her own benefits (Erdogan, 1999; Keel & Nataraajan, 2012).
Kahle and Homer (1985) claim that when a celebrity’s physical attractiveness matches the degree to which the product advertised enhanced attractiveness, the
match-up hypothesis would predict a positive impact upon the product and ad
evaluations. Furthermore, incongruence between celebrity and product/brand will lead to declined evaluations. However, with their research Kahle and Homer (1985)
merely focused on attractiveness related products.
Kamins’ (1990) study tested the hypothesis with an attractiveness-related product (luxury car) and an attractiveness-unrelated product (home computer). The celebrity endorsers who were used in the experiment were either perceived to be highly physically attractive (Tom Selleck) or perceived low in physical attractiveness (Telly Savalas). It was found that for the attractiveness-related products, Tom Selleck significantly enhanced the measures of advertisement-attitude, in comparison with Telly Savalas.
Most empirical work on the match-up hypothesis has focused on the physical appearances of the celebrity endorser (Till & Busler, 1998), with the argument that attractive endorsers are more effective in endorsing products that enhance
attractiveness. However, there are other important characteristics in celebrities that can be matched with brands/product as well. The research by Till & Busler (1998) even showed that expertise is more important than physical attractiveness when matching a brand with an appropriate endorser.
For celebrities with a bad girl image some brands may be more suitable than others. When a brand wants to send out a tough/raw image, they would be more eager to choose a female celebrity with a bad girl image than a celebrity that is not
perceived to be tough at all. The more similar the brand and the celebrity are, the more likely they become integrated within an associative network (Till & Busler, 2000). In this research not only the physical attractiveness of the celebrity is relevant when applying the match-up hypothesis, but characteristics such as behavior (e.g. being rebellious or innocent). It is expected that when the brand fits better with the celebrity, a celebrity endorser will be more effective. A celebrity who has a bad girl image who endorsers a brand with similar characteristics would be more effective than a celebrity with a good girl image who does not have similar characteristics with the brand she endorsers, according to the match-up hypothesis. Therefore, the last hypothesis is formulated:
H2: When there is a match between a celebrity endorser with a bad girl image and the brand, the attitude towards the ad and brand is more positively influenced in
comparison to a celebrity endorser with a good girl image who doesn’t match with the brand.
2.6 Brand Recall and The Product Match-Up Hypothesis
According to Percy & Rossiter (1992) brand awareness is often overlooked in
advertising strategies. Brand awareness is the consumers’ ability to identify a brand in detail within a category in order to make a purchase. Brand recognition and brand recall are two different types of brand awareness, whereby the difference depends on the first communication effect that occurs in the consumers’ mind. In this research, the focus is on brand recall. According to Percy and Rossiter (1992) brand recall occurs when a consumer has a certain category need and relies on his/her memory to generate possible solutions. Thus, in order to make a purchase, a consumer recalls a brand, or various brands, from memory.
Research has also shown that generally congruent information is better recalled than information incongruent with existing schemas (Cantor & Mischel, 1979; Cohen, 1981; Taylor & Crocker, 1981). Furthermore, some research has already shown that the perceived fit between the attributes of the spokesperson and the brand attributes results in better recall of brand information. For instance, Misra and Beatty (1990) found that recall of brand information is significantly higher when the celebrity fits with the brand. This holds true for recall measured right after exposure of the advertisements, and measured after a week delay.
The filtering model (Misra & Beatty, 1990; Taylor & Crocker, 1981) proposes that an existing schema helps to structure the new information in a meaningful way. Therefore, when an individual receives new information about a brand, information that is not congruent with or irrelevant to the characteristics of the existing schema may be filtered out. Congruent information is more easily encoded, thus leading to recall superiority for information that fits (e.g. with brand characteristics).
According to Misra and Beatty (1990) individuals create schemas for
celebrities as well. When an advertisement features a celebrity endorser, the schema related to this celebrity is likely to be evoked. According to the filtering model, when the consumer is exposed to the brand information, any information incongruent with, or irrelevant to, the celebrity schema would not be encoded as well as congruent information. Therefore, when information matches with the celebrity schema, the recall is higher than when the information does not match with the celebrity schema.
It is expected that a celebrity endorser with a bad girl image who endorses a brand that has congruent characteristics as her (e.g. tough, badass characteristics) will lead to significantly higher recall than a celebrity endorser with a good girl image who does not have congruent characteristics.
H3: If there is a match between a celebrity endorser with a bad girl image and a brand, the brand recall is higher compared to a celebrity endorser with a good girl image who does not have a match with a brand.
2.7 Experimental Stimuli and Target group
In this research two experimental stimuli are used: celebrity endorsers and brands. First of all, two celebrities are chosen, based on their image (either good or bad). Second, brands will be chosen as experimental stimuli, based on their congruency with the celebrities. In order to be sure that the target groups’ prior knowledge is not extremely high, brands that are very popular, such as Apple or Samsung, are not chosen. This is done because consumers often have a set opinion about these kinds of brands, which means that a short individual exposure to an advertisement would not have a high chance of changing that opinion. Moreover, according to Biswas et al. (2006) knowledgeable consumers are very unlikely to be influenced by a celebrity endorser in their decision-making.
Various scholars claim that identification increases with similarity and identification is considered to be a significant component in the persuasion process. Individuals are more likely to identify with celebrities that are for instance similar in gender (Fraser & Brown, 2002; Hsu &McDonald, 2002; Kelman, 1968; Wells et al., 1989). Therefore, only females were asked to participate, as it is expected that females would be more persuaded by female celebrity endorsers than males.
Moreover, the brands that were chosen are expected to be more interesting for women than for men.
This research tried to target a group that would be more interested in celebrities with a bad girl image, than celebrities with a good girl image. The
participants who were selected were approximately between the ages 18 and 22. Thus, this means that the respondents can either be in the last year of high school or already in their first years of college. The reason for choosing females of this age is that the brands that are chosen are more interesting for girls between 18 and 22; especially the
brand for the bad girl image. It was expected that females of different ages would react differently to bad girl image celebrities, and that in particular young females would be fonder of such celebrities. For instance, according to Choi and Rifon (2007) for older people it might be more important that the celebrity is sincere and genuine, while for younger consumers it might be important that celebrities have a hip, outgoing image.
3. Method
3.1 Type of method
The purpose of this research is to study the effects of celebrity endorsers with a bad girl image on brand attitude, advertisement attitude and brand recall compared to celebrities with a good girl image, while taking source credibility model and match-up between celebrity and product into consideration. A 2 (Celebrity image: bad versus good) x 2 (product: fit versus non-fit) between-subjects factorial design experiment was conducted.
3.2 Participants and Sampling Method
Female participants between 18 and 22 were asked to participate. For each condition 30 participants were required, thus 120 participants in total. The participants were recruited from either the high schools Zadkine College and Schoonhovens College (students that were in their graduation year), or from Erasmus University (Bachelor students).
In order to collect the data, e-mail addresses were collected via the school board from the high schools. Furthermore, the link of the online experiment was posted online on Facebook, requesting female students between 18 and 22 to
participate, and send privately via Facebook chat to the university students. Thus, the sampling method for this research is a non-probability sampling technique, namely convenience sampling. This entails that the participants were mostly based on accessibility, which made it effortless and less costly for the research (Marshall, 1996).
3.3 Procedure 3.3.1 Pre-test
In order to choose the stimulus material, two online pre-tests were conducted with 12 participants. The goal of these pre-tests was to choose the experimental stimuli: two celebrities and the brands that are shown in the main experiment. This is done to be sure that the celebrities in both the good and the bad girl condition were considered to have this (bad or good) image. The brands were pre-tested to be certain that these are considered to be a good match with either the bad girl celebrity or the good girl celebrity. Moreover, the product that fits with the bad girl celebrity also has to have a
bad fit with the good girl celebrity and vice versa. This way the fit and non-fit conditions were created.
Twenty celebrities selected from multiple popular Dutch online magazines (e.g. Cosmopolitan.nl, Vogue.nl) from current articles, to be certain that there is a high chance that the participants are familiar with the celebrities. Together with their pictures and name they were shown to the participants.
During the first pre-test the bad girl image and good girl image celebrities were chosen, which were selected based on previous information that the
experimenter had about them (e.g. they act rebellious or were mostly portrayed as innocent, sweet girls). Two definitions were shown to the respondents, the first one of a bad girl image, which was inspired by the research by Amos et al. (2008); “Female celebrities with a bad girl image are anti-celebrities, who have had certain “negative” experiences (e.g. fighting) or act in a “bad” way, and is seen as a badass, tough and rough woman/girl.” The second definition was of a good girl image, namely that ”celebrities with a good girl image are mostly seen as innocent, nonthreatening and sympathetic” (Meyers, 2009). After this the pictures of 20 celebrities were shown, whereby the participants were asked the question, “Do you consider this celebrity to have a bad girl image or good girl image?” Then the respondents could choose from a seven-point bipolar scale that ranged from a very bad girl image to a very good girl image.
After the first pre-test was conducted and analyzed, Doutzen Kroes (M=6.10, SD=.568) was selected as a celebrity with a good girl image and Amber Rose
(M=1.70, SD=.675) was selected as a celebrity with a bad girl image. They were selected based on the level of the mean, as the celebrity with the highest mean is most considered to have a very good girl image by the respondents, while the lowest mean is most considered to have a very bad girl image. Jennifer Aniston was actually the celebrity that had the highest mean (M=6.40, SD=.699), but after some consideration, it was decided that it was not a good idea to choose her as the celebrity endorser for the celebrity with a good girl image, and she shouldn’t have been in the pre-test in the first place. She is relatively older than all the other celebrities, and also older than Amber Rose, which might have influenced the outcome of the research.
The second pre-test was conducted to select the two brands. To get a match-up that is useful for this paper, ten brands were selected as possible experimental stimuli. The brands that were selected were believed to fit with the celebrity endorser with a
bad girl image or with a good girl image. The brands that were chosen and believed to fit with the bad girl image are:
1. Make up brand MAC Cosmetics 2. Fashion brand Nasty Gall 3. Fashion brand River Island 4. Fashion brand Cheap Monday 5. American Apparel
The products that were chosen that were believed to fit with the good girl image are: 1. Cosmetics brand Maybelline
2. Cosmetics brand L’Oreal 3. Fashion brand Mango 4. Fashion brand H&M 5. Fashion brand DKNY
The participants were exposed to Doutzen Kroes and Amber Rose with each brand. After each brand they were asked to choose from a 7-point bipolar scale ranging from a very bad fit to a very good fit. Eventually L’Oreal was chosen as having the best fit with Doutzen Kroes (M=6.36, SD= .505), while Amber Rose (M=1.55, SD= .820) had the worst fit with L’Oreal. Doutzen Kroes and Cheap Monday had the worst fit (M=2.09, SD=1.044), while Amber Rose and Cheap Monday had a good fit (M=5.18, SD=1.537).
3.3.2 Main experiment
After the pre-test had been conducted, L’Oreal and Cheap Monday were chosen for the fit and non-fit conditions, and Amber Rose and Doutzen Kroes were chosen for the bad and good image condition.
The participants for the main experiment were first asked via e-mail or
Facebook whether they agreed to participate in this study. Respondents who agreed to participate could immediately click on the link that was included in the e-mail or on Facebook. It was only revealed to the participants that the research is about celebrity endorsement, as this method was covert, the real reason of the research was not shared with the participants to ensure that they would not act in favor of the outcome of the experiment.
Four groups were created: (1) a group of participants who were exposed to an advertisement with a good girl image celebrity and a brand that is congruent with this image, (2) a group participants who were exposed to an advertisement with a bad girl
image celebrity and a brand that is non-congruent with this image, (3) a group participants who were exposed to a bad girl image celebrity and a brand that is
incongruent with this image and lastly, (4) a group participants who were exposed to a good girl image celebrity and a brand that is incongruent with this image.
Furthermore, it is important to know that the advertisement were especially made for this experiment, thus they are fictional. The advertisements can be found in image 1.
For the main experiment, the respondents were randomly assigned to one of the four conditions via the Qualtrics questionnaire. First, the participants had to give their informed consent, followed with questions for demographic data. After that they had to fill in several multiple choice questions that measure brand recall, credibility, attitude towards the brand and advertisement and the fit of the celebrity with the brand.
3.4 Operationalizations dependent variables 3.4.1 Brand Recall
In order to measure brand recall, the participants were asked which brand was advertised; where they could either fill in the correct brand name or I do not remember/I do not know. The results show that 25.8% of the respondents did not recall the brand, whilst 74.2% correctly recalled the right brand.
3.4.2 Source Credibility Model
The questions for the source credibility model was inspired by the research of Edwards and La Ferle (2009), measuring the attractiveness, trustworthiness, and expertise of the celebrity, each on a seven-point bipolar scale. Attractiveness was measured from a scale ranging from highly un-attractive to highly attractive, from very bad looking to very good looking, from very ugly to very pretty, and from very unsexy to very sexy. Expertise was measured from a scale ranging from highly inexpert about product X to highly expert about product X, highly unknowledgeable about product X to highly knowledgeable about product X, highly un-experienced with product X to highly experienced with product X and knows very little about product X to knows very much about product X. Trustworthiness was measured with the scales ranging from highly untrustworthy to highly trustworthy, highly dishonest in her recommendation of product X to highly honest in her recommendation of product X, highly undependable to highly dependable and highly insincere to highly sincere.
Image 1. The four advertisements as experimental stimuli
3.4.3 Attitude towards the brand and advertisement
Attitude towards the advertisement was measured on seven-point bipolar scales, ranging from highly unappealing to highly appealing, highly undesirable to highly desirable, very boring to very interesting and highly unexciting to highly exciting. Attitude towards the brand was measured on seven-point bipolar scales, ranging from highly unreliable to highly reliable, very low quality to very highly quality, highly unfavorable to highly favorable and highly undesirable to highly desirable, highly ineffective to highly effective, very unknown to very well-known, very common to very distinctive.
3.4.4 Product fit/celebrity fit
Product fit/celebrity fit was measured on a seven-point bipolar scale, ranging from highly inappropriate to highly appropriate, highly ineffective to highly effective, very poor endorser to very strong endorser and very bad fit to very good fit.
4. Results
4.1 Sample Characteristics
In total 173 participants started the experiment, but after cleaning out the data it was found that 48 participants did not complete it. Furthermore, it was found that one person was 60 years old, which may have been a typo, however, to be certain that this person is not a large outlier, this participant was deleted out of the dataset. Thus, in total there were 124 respondents study.
The average age of the respondents is 21 years (N=124, M=21, SD=1.68). Of the respondents 38.7% has a bachelor’s degree, 33.9% has a high school diploma, 9.7% has a HBO diploma, 9.7% has a master’s degree, 5,6% has a MBO diploma, and finally, 2.4% of the respondents checked “other”. When examining the nationality of the respondents, it was clear that 88,7% are Dutch, 2.4% are Hungarian, 4% are German, 0.8% are Dutch-Polish, 0.8% are American, 0.8% are Dutch-Pakistani and 0.8% are Dutch-German.
4.2 Internal Consistency Dependent Measures
Before computing the various measures together, the internal consistency was measured using the Cronbach’s Alpha. When looking at the measures of the source credibility model, the internal consistency of attractiveness can be considered excellent (α=.91), expertise can be considered good (α=.85) and trustworthiness can be considered acceptable (α=.77). Furthermore, the internal consistency of the measures of attitude towards the advertisement can be considered good (α=.89) and the measures of the attitude towards the brand can be considered excellent (α=.92). Lastly, the measure of brand fit was good (α=.87). Thus, these variables were computed into individual variables.
4.3 Results Main Analyses Experiment 4.3.1 Source Credibility Model Results
First of all, the three components of the source credibility model are analyzed, which consists out of attractiveness, expertise and trustworthiness. There are no hypotheses formulated based on the theory of the source credibility model, this theory is merely incorporated to take into account that it is a possibility that one of the two endorsers is more effective than the other, as their credibility is significantly higher. In figure 1 the mean scores show that the celebrity with a good girl image (Doutzen Kroes) was
considered significantly more attractive and trustworthy than the celebrity with the bad image (Amber Rose).
The participants who were exposed to the celebrity with a bad girl image, considered the celebrity endorser significantly less attractive (M=3.76, SD=.99), than the participants who were exposed to the good girl image celebrity (M=5.35, SD=.94). Furthermore, the participants who were exposed to the celebrity with a bad girl image (M=3.91, SD=1.12) considered the celebrity endorser significantly less trustworthy than the participants who were exposed to the celebrity with the good girl image (M=4.55, SD=1.11). Although the celebrity with the bad girl image (M=4.30,
SD=1.29) was considered slightly less an expert than the celebrity with the good girl image (M=4.42, SD=1.23), there was no significant difference. There was no
expectation for the source credibility model and the celebrities; however, it needs to be taken into consideration that Doutzen Kroes is considered significantly more attractive and trustworthy than Amber Rose.
A one-way multivariate analysis of variances (MANOVA) was conducted to examine whether the credibility between the two celebrities was significantly
different. The analysis shows that there was a statistically significant difference in credibility in the bad girl celebrity image and good girl celebrity image
F(3,120)=36.94, p=.000; Wilk’s Λ=.520, partial η2=.48. Furthermore, from the tests of between-subjects effects one can see that celebrity image has a statistically significant effect on attractiveness (F(1,122)=84.90; p=.000, partial η2=.41) and trustworthiness (F(1,122)=10.31; p=.002, partial η2=.08). There was no significant difference in the two celebrities for expertise (F(1,122)=.29; p=.590, partial η2=.002).
3,76 4,3 3,91 5,35 4,42 4,55 0 1 2 3 4 5 6
Attractiveness Expertise Trustworthiness
Bad Image Good Image
4.3.2 Hypothesis testing
According to hypothesis one it was expected that celebrity endorsers who have a bad girl image would positively impact consumers’ attitude towards the brand and advertisement, in comparison with celebrity endorsers who have a good girl image. Figure 2 show the mean scores, where one can see that exposure to the celebrity with a good girl image led to a significantly more positive attitude towards the brand.
Contrary to expectations, it was found that the participants who were exposed to the celebrity with a good girl image (M=4.75, SD=1.26) had a significantly more positive attitude towards the brand than the participants exposed to a celebrity with a bad girl image (M=4.19, SD=1.27). The participants who were exposed to the
celebrity with a bad girl image (M=4.13, SD=1.50) had a slightly more positive attitude towards the advertisement than the participants exposed to the advertisement with the good girl celebrity (M=4.07, SD=1.40).
A one-way multivariate analysis of variances (MANOVA) was conducted, with attitude towards the advertisement and attitude towards the brand as dependent variables, and celebrity image as independent variable. The analysis shows that there was a statistically significant multivariate effect F(2,121)= 9.78, p=.000; Wilk’s Λ=.861, partial η2
=.14. To determine how the dependent variables differ for the independent variable, the tests of between-subjects effects table is analyzed. From this table one can see that for the celebrity image there was a significant main effect for attitude towards the brand (F(1, 122)= 6.062, p=.015; partial η2=0.05). There is no significant main effect for attitude towards the advertisement (F(1, 122)=.066, p= .798; partial η2=0.00). As the attitude towards the brand was significantly better for an advertisement with a good girl celebrity, and the attitude towards the advertisement was only slightly better for the advertisement with the bad girl celebrity, hypothesis one has to be rejected, and the null hypothesis accepted.
Figure 2. The mean scores for the attitude towards the brand and advertisement
Hypothesis two states that when there is a celebrity endorser with a bad girl image that endorses a matching brand, the attitude is more positively influenced in comparison with a celebrity endorser with a good girl image that endorses a brand that doesn’t match. In figure 3 the means can be found for the attitude towards the brand and attitude towards the advertisement.
As expected, the participants who saw the advertisement displaying the celebrity with a bad girl image and a congruent brand (M=4.33, SD=.25) had a significantly better attitude towards the advertisement than the participants who saw the advertisement displaying the celebrity with a good girl image and a incongruent brand (M=3.27, SD=.28). Furthermore, the participants who were exposed to the celebrity with a bad girl image and a congruent brand (M=4.20, SD=.22) had a slightly better attitude towards the brand than the participants who were exposed to the celebrity endorser with a good girl image and a incongruent brand (M=4.10, SD=.25).
Another one-way multivariate analysis of variances (MANOVA) was conducted. The analysis shows that there is a statistically significant difference in attitude towards the brand and advertisement based on the celebrity image that was shown F(2,119)=9.10, p=.000; Wilk's Λ =.867, partial η2 =.133. Furthermore, there is also a statistically significant difference in attitude towards the brand and
advertisement based on the brand (congruent or incongruent) that was shown F(2, 119)=5.56, p=.005; Wilk's Λ =.915, partial η2 =.09. Furthermore, the interaction effect was non-significant, F(2, 119)= 2.52, p=.085; Wilk's Λ= .959; partial η2 =.04.
4,19 4,13 4,75 4,07 3,6 3,8 4 4,2 4,4 4,6 4,8 5
Attitude towards the
brand Attitude towards the ad
Bad image Good image
To determine how the dependent variables differ for the independent variables, the tests of between-subjects effects table is analyzed. Here, one can see that there is a significant result for attitude towards the advertisement and product fit (F(1,120) =11.09; p=.001; partial η2 =.09), and a significant result for attitude towards the brand and product fit (F(1,120)=5.70; p=.019; partial η2 =.05). There were no significant results for attitude towards the advertisement and celebrity image (F(1,120)=.76; p=.387; partial η2 =.01), nor a significant result for attitude towards the brand and celebrity image (F(1,120)=3.73; p=.056; partial η2 = .03). Thus, hypothesis two can be partly accepted.
Figure 3. Mean scores for attitude towards the brand and advertisement with the match-up hypothesis
According to the third hypothesis, when there is a fit between the brand and the celebrity with a bad girl image, the brand is better recalled in comparison to a celebrity with a good girl image without a fitting brand. In figure 4, the mean scores show that the participants that were exposed to the advertisement with a bad girl image and a matching brand, had a better brand recall than the participants that were exposed to the advertisement with a good girl image and an incongruent brand. Thus, as expected, the participants who were exposed to the advertisement with the celebrity with a bad girl image and a congruent brand (M=1.65, SD=.49) better recalled the brand than the participants who were exposed to the celebrity with a good image and an incongruent brand (M=1.58, SD=.50). 4,2 4,1 4,33 3,27 0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5
Attitude towards the
brand Attitude towards thead
Bad image and congruent brand Good image and incongruent brand
A two-way analysis of variance (ANOVA) was conducted to examine whether the effect of the match-up hypothesis and image of the celebrity endorser (bad vs. good) was significant. There was a statistically significant interaction between the effects of celebrity image and brand type on brand recall, F(1,120)= 7.62, p=.007. However, the analysis shows that there is no significant effect for brand recall for product F(1,120)= 1.70, p=.195, nor for celebrity image F(1,120)=.261, p=.611. Thus, as there was no statistically difference between the groups, hypothesis three has to be rejected and the null hypothesis has to be accepted.
Figure 4. The mean scores for brand recall 1,65 1,58 1,54 1,56 1,58 1,6 1,62 1,64 1,66 brand Recall
Bad image and congruent brand Good image and incongruent brand
5. Discussion
5.1 Summary of Results
First of all, the results show that Doutzen Kroes (good girl image) is considered significantly more attractive and trustworthy than Amber Rose (bad girl image). There was no significant result for expertise, although Amber Rose was considered slightly less as an expert than Doutzen Kroes.
Second of all, participants who were exposed to the advertisement featuring Doutzen Kroes had a significantly more positive attitude towards the brand than when the advertisement included Amber Rose. There was no significant result for attitude towards the advertisement, although the participants who had viewed the
advertisement with Amber Rose had a slightly more positive attitude towards the advertisement than the participants who were exposed to the advertisement with Doutzen Kroes. All in all, this means that hypothesis one, that states that celebrity endorsers who have a bad girl image will positively impact consumer’s attitude towards the brand and advertisements in comparison with celebrities with a good girl image, has to be rejected. However, when one looks at matching brands, one can see that the participants who saw the advertisement with Amber Rose who has congruent characteristics with the brand she endorses (Cheap Monday) had a significantly better attitude towards the advertisement than the participants who saw the advertisement with Doutzen Kroes who endorses a non-congruent brand (Cheap Monday).
Furthermore, the respondents who were exposed to the advertisement with Amber Rose and the congruent brand had a slightly better attitude towards the brand than the participants that were exposed to the advertisement with Doutzen Kroes and the non-congruent brand. However, this result is not significant. Thus, the second hypothesis can only be partly accepted.
Lastly, there was no significant difference between the participants in all four conditions for brand recall. However, the participants that viewed the advertisement that included Amber Rose and Cheap Monday (congruent) slightly recalled the brand better than participants that had viewed the advertisement with Doutzen Kroes and Cheap Monday (incongruent). There are no significant results, thus the last hypothesis has to be rejected.
5.2 Limitations and Recommendations
This research contains several limitations that are important to mention. First of all, the most important limitation is that the good girl image and bad girl image differ significantly with each other on attractiveness and trustworthiness. Doutzen Kroes is considered significantly more attractive and trustworthy, and also slightly more expert than Amber Rose. For an experiment it is very important that the experimental stimuli are very similar in most areas, which is not the case with this research. Therefore, not much can be said about the good girl image and bad girl image, as attractiveness and trustworthiness play a confounding role. Doutzen Kroes is considered significantly more attractive and trustworthy than Amber Rose and this may have influenced the outcome of the research. For future research, it is therefore advisable to not only measure the image of the celebrity in a pre-test, but also use the source credibility model.
Second, it is also advisable to measure the popularity of the celebrity
beforehand, which can be done by asking the respondents how familiar they are with the celebrity or how popular they consider them to be. This is relevant to do, as another limitation is that Doutzen Kroes is most likely more popular than Amber Rose. Although a pre-test was conducted in order to choose the celebrity endorsers, it was not thought of that Doutzen Kroes is most likely more famous and/or popular. Furthermore, in the theoretical framework it was hypothesized that the attitude towards the celebrities with a good girl image is more neutral than the attitude towards the bad girl image, as a celebrity with a good girl image might be perceived as boring. Even though this might be the case for some celebrities, this is probably not the case for Doutzen Kroes; of course, this is merely an opinion. Doutzen Kroes has a large following on many social media platforms, which might indicate her popularity. Doutzen Kroes was chosen in the pre-test as having a very good girl image, yet this does not mean respondents find her boring. Therefore, in future research it is advisable to use celebrities who are equally popular in order to not influence the outcome of the research.
Third, even though a pre-test was conducted to select the brands that are used in this experiment, the selection did not go faultlessly. It was important that the participants did not have high prior knowledge of the brands, however, this was most likely the case for one of the brands that was shown. The brand L’Oreal is quite known, which means that the participants can already have salient beliefs about this
brand. This makes it harder to measure if their attitude towards the brand and advertisement changed based on the advertisement in this research. Therefore, it is recommended for future research to use brands that are less known. This way, the knowledge the participants already have will not influence their attitude. Furthermore, the brand Cheap Monday is most likely less known than the brand L’Oreal. It may be that this has influenced the outcome of brand recall and the participant’s attitude towards the brand and advertisement.
The fourth limitation is that the quality of the stimulus material, the
advertisements, could have been better. The experimenter herself created these fictive advertisements at an amateur level, with an online program that delivers lower quality than most Photoshop programs. Thus, this could also have influenced the
respondents’ attitude towards the advertisements. Therefore, for future research it is recommended to use advertisements that are made by someone who is skilled with Photoshop or use real advertisements.
Lastly there are two more final recommendations for future research. First of all, another method can be used to investigate the phenomenon of bad girl celebrities; for instance using qualitative research instead of quantitative. It might be interesting to conduct in-depth interviews about celebrity endorsers with a bad girl image, as one can get more of an insight into why they would be effective endorsers for certain target groups and brands. Now there was a comparison with celebrities who have a good girl image to measure their effectiveness, while it might be better to investigate in which ways they are effective and what their qualities actually are, without
comparing it to another kind of celebrity image. Second, future research can make use of male celebrity endorsers, as this research only investigated female celebrities. It is interesting for future research to investigate whether male celebrity endorsers with a bad image are more effective than male celebrity endorsers with a good image or even compare them with female celebrity endorsers.
5.3 Conclusions and Practical implications
This research investigated the phenomena of a celebrity with a “bad girl image”; as it seems a vast array of brands make use of such rebellious women in their advertising campaigns. The results indicate that a celebrity endorser with a bad girl image positively impacts attitude towards the brand, attitude towards the advertisement and brand recall, but not necessarily more than a celebrity endorser with a good girl
image. When an advertiser is considering choosing a celebrity with a bad girl image, it is highly recommended to merely make this choice when the celebrity is congruent with the brand or product she will endorse. This research showed that only when there was a match between the characteristics of the celebrity with a bad girl image and the brand, the attitude towards the advertisement is significantly more positively
influenced than when the consumers are exposed to an advertisement with a good girl image and a non-matching brand. Furthermore, for attitude towards the brand, the participants had a slightly more positive attitude when they were exposed to the bad girl image and congruent brand than the participants that were exposed to the good girl image and non-congruent brand.
This research shows that a celebrity with a bad girl image is not always the best choice. The respondents that were exposed to a celebrity with a good girl image had a significantly higher attitude towards the brand than the respondents that were exposed to a celebrity with a bad girl image. However, this research also measured the credibility of the celebrities by using the source credibility model. This shows that the participants that were exposed to the celebrity with the good girl image considered the endorser to be significantly more credible, than the participants that saw the celebrity with a bad girl image. This might have influenced the outcome of the
research. Previous research has shown that higher source credibility is very persuasive in attitude measures (Pornpitakpan, 2004), meaning that this could have influenced the attitude of the respondents towards the brand and advertisement. Furthermore, the brand L’Oreal is a brand with high prior knowledge, which could also influence the outcome.
All in all, no certain conclusions can be made about the influence of a celebrity with a bad girl image and a good girl image on the attitude towards the advertisement and attitude towards the ad. However, previous literature and the outcome of the experiment does indicate that the matching the endorser with the brand does positively influence brand attitude and brand attitude, and works better than using a endorser who does not match with one’s brand.
Previous research has shown that congruent information is generally better recalled than information that is incongruent (Cantor & Mischel, 1979; Cohen, 1981; Tayloor & Crocker, 1981), and that recall of brand information is significantly higher when the celebrity fits with the brand, also right after exposure of the advertisement (Misra & Beatty, 1990). This experiment does not show the same significant results.
There was a better brand recall for Cheap Monday when the celebrity was congruent, than when the celebrity was incongruent; but this was only a slight difference.
However, as mentioned before, it might be that the usage of L’Oreal has influenced the research, as it’s a brand with high prior knowledge. The respondents who had seen the L’Oreal advertisement probably already had a base knowledge with information about the brand (Biswas et al., 2006; Cowley & Mitchell, 2003). L’Oreal is not significantly better recalled than Cheap Monday, or vice versa. However, it could have been the case that the congruent brand for the bad girl image was significantly better recalled in comparison to the incongruent brand for the good girl celebrity, if a less known brand for the good girl image was used. Thus, again, as with the other two hypotheses, no certain conclusions can be drawn from this research for the hypothesis of brand recall.
Thus, what is the answer to the main research question of this paper; To what extent are celebrity endorsers who have a bad girl image positively influencing attitude toward the brand and advertisement and recall of the brand, when they endorse a congruent brand, compared to celebrity endorsers who have a good girl image who endorse an incongruent brand? This research has shown that celebrity endorsers with a bad girl image who endorse a congruent brand positively influence the attitude towards the ad significantly more than celebrity endorsers with a good girl image that endorse a incongruent brand. Further conclusions cannot be drawn about brand recall and attitude towards the brand, and also the significant outcome of attitude towards the advertisement is not very trustworthy, due to the unreliable experimental stimuli.
On a final note, we end where we began. According to Laurel Thatcher Ulrich (1976, p.20), “well-behaved women seldom make history” where she compares Antinomians and witches with devout matrons, claiming that the later have had little chance at all. Although from the results of this research, no certain conclusions can be drawn, one thing is certain: Doutzen Kroes, mother of two and a devoted wife,
definitely stands a chance against Amber Rose in making history with her “well-behaved” image.
Literature
Aaker, D.A., & Stayman, D.M. (1990). Measuring audience perceptions of
commercials and relating them to ad impact. Journal of Advertising Research, 30(4), 7-17.
Amos, C., Holmes, G., & Strutton, D. (2008). Exploring the relationship between celebrity endorser effects and advertising effectiveness: A quantitative synthesis of effect size. International Journal of Advertising, 27, 209-234. doi: 10.1080/02650487.2008.11073052
Akturan, U. (2011). Celebrity advertising in the case of negative associations:
Discourse analysis of weblogs. Management Research Review, 34, 1280-1295. doi: 10.1108/01409171111186405
Bartz, S., Molchanov, A., & Stork, P.A. (2013). When a celebrity endorser is disgraced: A twenty-five-year event study. Marketing Letters, 24, 131-141. doi: 10.1007/s11002-013-9229-2
Birnbaum, M. H., & Stegner, S.E. (1979). Source credibility in social judgment: Bias, expertise, and the judge’s point of view. Journal of Personality and Social Psychology, 37(1), 48-74. doi: http://dx.doi.org/10.1037/0022-3514.37.1.48
Biswas, D., Biswas, A., & Das, N. (2006). The differential effects of celebrity and expert endorsements on consumer risk perceptions. The role of consumer knowledge, perceived congruency, and product technology orientation. Journal of Advertising, 35(2), 17-31. doi: 10.1080/00913367.2006.10639231
Braunsberger, K. (1996). The effects of source and product characteristics on persuasion. Dissertation Abstracts International: Humanities and Social Sciences, 57(6), 2575.
