The Effectiveness of Attractive Endorsers in Fashion Advertising:
Assessing the roles of Viewer-Source Similarity and Self-Referencing
Sabrina A. Roy University of Amsterdam
Master Thesis
Author: Sabrina Alexandra Roy, MSc (10506640)
Institution: University of Amsterdam, Faculty of Economics and Business Programme: MSc Business Studies – Marketing Track
Supervisor: Dr. Frederik B.I. Situmeang 2nd assessor: Dr. Hsin-Hsuan Lee
Date: January 31, 2014 Draft: Final version
Acknowledgements
Foremost, I would like to express my appreciation to my supervisor, Frederik Situmeang, for having believed in this exciting topic since our first meeting. Although the fashion industry has not been studied much in business research, he acknowledged the fact that I was very passionate about this matter, and that I needed to research a topic that could potentially have practical implications.
My sincere appreciation goes to the friends who shared the survey and everyone who took the time to complete it. Without your participation, such a thesis and marketing research would not be possible.
I would like to extend my sincerest thanks to the team at BonLook and Alee for participating in the photoshoot, as well as Maxim for his patience and Ben Barry for the inspiration.
My gratitude goes to my mother and F. for their continuous support as well as to my father who helped me make this experience possible.
This thesis is dedicated to my grandparents for who education is the most valuable asset.
ABSTRACT
This paper examines consumer self-referencing as a mechanism for explaining the effects of the attractiveness of endorsers in fashion advertising. This study also assesses the role of viewer-source similarity on this relationship. Data was collected from women (N =299) who were randomly exposed to 1 of 2 advertising conditions (i.e. highly vs. realistically attractive models). Results show that (1) self-referencing mediates endorser attractiveness on model attitudes, (2) viewer-source similarity moderates the relationship between attractiveness and self-referencing at least for age, (3) higher attractiveness leads to higher purchase intentions, (4) ad and product attitudes mediate the model attitudes on purchase intentions, (5) realistically attractive models generate higher self-referencing. The results support the idea that brands should make efforts to stimulate the self-referencing of their target market when they are exposed to their advertisement campaigns. This study contributes to the research investigating potential use of alternative images in advertising that reflects more similarity between the viewer and source message.
Keywords: Attractiveness, Endorser, Advertising, Model, Self-Referencing, Women,
TABLE OF CONTENTS
ABSTRACT ... 3
INTRODUCTION ... 6
LITERATURE REVIEW ... 10
2.1 Highly attractive models vs Realistically attractive model... 10
2.2 Viewer-source similarity ... 13
2.3 Social comparison and self-referencing ... 16
2.3 Self-Referencing as a mediating variable ... 19
2.4 Self-Referencing and ethnicity ... 20
2.5 Self-referencing and viewer-source similarity ... 21
2.6 Theoretical framework & present study ... 23
RESEARCH METHOD... 24
3.1 Design ... 25
3.2 Ad models and ad stimulus ... 25
3.3 Data collection and procedure ... 28
3.4 Sample... 30
3.5 Measurement scales ... 30
3.5.1 Manipulation check ... 30
3.5.2 Independent variable: Attractiveness ... 32
3.5.3 Dependent variables ... 32
3.5.4 Mediator: Self-Referencing ... 33
3.5.5 Moderator variables ... 35
RESULTS ... 37
4.1 Overview ... 37
4.2 Correlation and significance ... 38
4.3 Mediator and moderator effects ... 41
DISCUSSION ... 45 5.1 Findings... 45 5.2 Additional Research ... 53 5.2.1 Internalizations of ideals... 53 5.2.2 Body Satisfaction ... 55 5.3 Practical implications ... 56 5.4 Limitations ... 58
REFERENCE LIST ... 62 APPENDIX I – Advertising Stimuli ... 66 APPENDIX 2 - Survey ... 67
‘The fashion industry is in the business of business; it will only be motivated to diversify if it’s a strategy that boots the bottom line.’ – Ben Barry
INTRODUCTION
Since the 19th century, advertising has been the main communication tool to promote consumer products, and women have been the prominent endorser of the message. Depicted as homemakers, superwomen, educators or sex objects, female models in ads have been selling anything targeting both men and women (Kurtz, 1997). Although major societal changes have boosted women’s leadership over the years, such as their entry in the workforce and the rise of their purchasing power, advertisements targeting them have barely changed (Taylor & Miyazaki 2013). Indeed, the women portrayed in ads have always embodied the ultimate ideal of beauty that is fostered by social standards. However, fueled by the advertisers on Madison Avenue, a cultural shift occurred in beauty standards in the 1940s. Since then, we have witnessed a body image revolution in the media that is extremely inconsistent with reality.
It is true that an extensive array of literature has suggested that attractive models in advertising have positive effects on consumer attitudes (Kahle & Homer, 1985), product evaluations, willingness to purchase (Belch, Belch & Villareal, 1987) and actual purchase (Caballero & Solomon, 1984). In fact, for decades, attractive message sources have been appropriate to generate sales, especially in the fashion and cosmetics industries. However, it seems this strategy has been applied at its extreme. Indeed, Marketers targeting women choose to feature highly attractive/thin models that have often been completely transformed, retouched, falsified. They assume it is the most effective practice since idealized beauty has been the norm of the industry for a long time. According to them, skinny and pretty have selling power and keeps the demand high. They think it’s profitable to make women feel bad about them by selling
them a dream they will never achieve. However, few empirical studies support this opinion. In this context, professionals from the fashion and advertising industries act as cultural gatekeepers by defining the ideals of beauty, which might differ from the consumer’s standards (Salomon, Ashmore & Longo, 1992). Most of them also choose models that look very different from consumers. The fact that the average female model in the USA is a size zero, but that the average American woman is a size fourteen, illustrates well this dissimilarity (Wykes & Gunter, 2005). According to many scholars, negative affect can occur when women compare themselves to these attractive, yet very different models in advertising (e.g., Irving, 1990; Martin & Gentry 1997; Richins, 1991). In fact, this representation of highly attractive and predominantly young, thin and white models in advertising has a major impact on the body image and self-confidence of consumers (Halliwell & Dittmar, 2004). As a result, advertising has been criticized for decades now for exhibiting unrealistic and idealized images of beauty that consumers compare themselves with (Pollay, 1986). Nowadays, ‘before and after’ photos of celebrities, highlighting the extreme digital manipulation and touch-ups they have gone through leak everyday in social media and have become a sensation (e.g., Lena Dunham, 2014; Jennifer Lawrence, 2014). This quote from Susie Orbach (2009) has never been truer: ‘The body is turning from being the means of production to the production itself’.
The extreme digital manipulations of models in recent years have led female consumers to express their dissatisfaction with the physical stereotypes that are represented in the media. It seems they valorize diversity and aspire to advertising messages that encourage a positive body image. Governments have joined the debate by creating codes of conduct (e.g., The Québec Charter for a Healthy and Diverse Body Image) or policies against extremely thin models during Fashion Weeks (e.g., Montreal, Madrid, Milan). The Australian Government has even explicitly
advised the media that they should encourage body diversity in their channels and reduce – or notify – the use of airbrushing (Diedrichs & Lee, 2011). In Germany, an entire magazine took the decision – and the risk – to only use real women in their pages, and it became the most sold magazine of the country. Other magazines have published fashion editorials addressed to larger or older women (e.g., V Magazine) or have even featured plus-size models on their cover (e.g., Vogue, Elle). In November 2013, the director of the world's top modeling agency, IMG,
proclaimed: "We want to be an ageless, raceless, weightless agency". He recognized the fact that women actually want to see real women in ads. Around the same time, Gap launched a campaign with a Sikh Model that made headlines all over the world. These initiatives in fashion used to be limited to a few brands (e.g., Benetton), but it was more a strategy of brand differentiation or provocation that anything else. Some designers have also started to hire diverse models on the runways (e.g., Sunny Fong), others have worked with models from a minority like the disabled or albinos (e.g., Vivienne Westwood, Jean-Paul Gaultier) or even older models (Taylor & Miyazaki, 2013). Even though these initiatives are scarce and often limited to high-end brands, we have seen in the last years some examples of well-known brands working towards a better representation of women in advertising (e.g., Unilever for Dove, 2004-2014; Debenhams, 2013; Aerie, 2014). Nevertheless, the media landscape has barely changed, highly attractive and young women being represented most of times (Taylor & Miyazaki, 2013).
A good argument to convince brands and advertisers to embrace more diversity in fashion advertising is by showing them that using similar models in advertising would not only be more socially responsible, but also more lucrative. Indeed, as research suggested, advertisers miss an opportunity ‘by failing to make more strategic casting decisions’ that would respect their target market (Barry, 2007). To promote diversity and models that fully represent the population
could be an effective public health intervention. However, this study goes further by tackling the interests of the cultural gatekeepers who have the influence to change media imagery. Following the latest research on viewer-source similarity, we will try to demonstrate that using models in fashion advertising that are similar to the targeted consumers is a good thing for brands.
Apart from a few authors, most of academic research looked at the relationship between models in advertising and self-esteem or body image without studying the impact of purchase intentions and consumer variables. Studies in this field have also mainly focused on other product categories. This study will evaluate the effectiveness of viewer-source similarity in fashion advertising. More particularly,
o How do different types of models in fashion advertising influence consumer behavior?
o What type of model will increase the purchase intentions? And how will the consumer attitudes affect the intentions to buy?
o What is the process that women engage in when evaluating realistically attractive models vs. very attractive models in advertisements promoting fashion?
o How do the age, race and body shape of female consumers influence advertising effectiveness?
The goal of this thesis is to determine if matching models attributes to the ones of the consumers is more effective than using the traditional ‘recipe’ of highly attractive, Caucasian, young and thin models in fashion advertising. The study will be intended for a specific product category, fashion accessories. Also, the main past research on viewer-source similarity (Barry, 2007) was conducted with a qualitative approach; yet, a quantitative research with a random sample of women was required to generalize their results. For this quantitative study, we
analyzed the data that was collected from 299 women who randomly saw one of two stimuli conditions and answered a survey of 20 questions. This paper will start with the literature review, followed by the research method explaining in detail the ad stimuli and procedures of data
collection. We will then summarize the results and analyze them fully in the discussion section, where will also be covered additional research and implications.
LITERATURE REVIEW
2.1 Highly attractive models vs Realistically attractive model
Marketing literature on the effectiveness of attractiveness is extensive. In fact, abundant are the studies that have exposed the positive effects of attractive message sources on brand and product perceptions and evaluations (e.g., Joseph, 1982; Kahle & Homer, 1985; Caballero & Solomon, 1984, Barry, 2007). Attractive people are seen as more sociable, self-confident, mentally healthy, sexually warm (Feingold, 1992) and above all, more familiar, regardless of prior exposure (Monin, 2003). Considering that familiarity leads to higher liking (Zajonc, 2001), female consumers like attractive communicators more; thus, they are in a more favorable
position to analyze the message of the ad/source in a more positive way. This positive evaluation of the model and message is transferred to the product. The similarity-attraction theory
elaborated by Byrne (1971) also suggests that people find other people attractive when they share similar attributes.
By definition, highly attractive models have beautiful faces and bodies characterized by thinness (Richins, 1991). Women idealize these models and compare the way they look to their beauty because physical attractiveness is very valued in the society. However, it turns out that
using highly attractive models might not be as effective as assumed (e.g. Bower & Landreth 2001; Caballero & Salomon, 1984). In fact, when a comparison between a consumer and a beautiful model in an ad creates sufficient negative affect, the source derogation is prone to lead to a negative view of the spokesperson, the advertised product and the brand, as well as to a decrease in intentions to purchase (Bower, 2001). Thus, it seems that if brands would represent their products differently in their ads, consumers would respond more positively to the message. Although studies have raised these issues for a while now, brands have not really responded since they have been meeting their target sales anyway. This finding is very interesting because it proposes that the evaluations of spokesperson and product argument can have significant effect on the product evaluations and purchase intentions.
Most women do not share the same corporal attributes of these exceptional women and, consequently, they feel like they don’t measure up. Because of this, they experience negative feelings, such as frustration, anxiety, envy and jealousy (Richins, 1991; Salovey & Rodin, 1984). It’s been shown that many women loose the motivation to shop after viewing these ads because they ‘they do not want to focus on their bodies or they feel excluded from the advertised
products’ (Barry, 2013). They have ‘advertising experience’ and know that only extreme – or even impossible – techniques, such as dangerous diets and airbrushing, can make them reach the idealized beauty in the media. Their skepticism is increased, which leads to negative brand outcomes. Therefore, the goal of an advertiser should be to make trustworthy advertising content to increase positive outcomes and ad effectiveness (Barry, 2007). According to many studies, women trust ads exposing endorsers that seem to be credible, honest and sincere. More specifically, they want to perceive these positive qualities in nonverbal cues such as physical attributes, i.e. not solely in the ad copy. They want to see endorsers who display healthy body
shapes, normal aging signs as well as diverse racial traits (Wood, Boles Johnston, & Bellenger, 2008). Bower and Landreth (2001) studied the negative consequences of highly attractive models on ad processing and persuasiveness by examining how consumers see themselves compared to these models. They found that normally attractive models are not perceived to be trustworthier than highly attractive models in low involvement products but that they do are in high
involvement products. Also, spokesperson trustworthiness does not mediate the effects of model beauty on ad effectiveness. According to this research, it seems that the ‘degree’ of the
attractiveness of models has an impact on the effectiveness of an advertisement depending on the type of product, i.e., high or low involvement / enhancing or problem-solving products (Bower & Landreth, 2001). Many scholars have also studied the negative effects of thin ideals and the representation of skinny models in advertising on women’s self-esteem, self-perceptions, body satisfaction, mood and even well-being (Stice & Shaw, 1994). With 59% of Americans and 40% of Europeans being overweigh or obese, it is not surprising that consumers don’t relate to these models (Smeesters, Mussweiler & Mandel, 2010). It has been shown a several times, over many years, that women who view neutral images instead of these thin models express more self-esteem and body satisfaction.
Moreover, findings showed that when featuring an exceptionally attractive model in an ad, American women experience less self-esteem because they engage in self-evaluation. Similarly to what Bower (2001) found for highly attractive models, this process harmfully impacts the ad and product evaluations (Micu & Coulter, 2012). Moreover, these authors have found that when an average-looking model is featured in advertising, the self-perceptions of women do not differ among comparison motive conditions (Micu & Coulter, 2012). Even though women’s perceptions were different in Asia where this study was also conducted, the authors
showed that, in general, encouraging women to compare themselves with beautiful models for the purpose of self-evaluation affects advertising attitudes in a negative way (Micu & Coulter 2012). In sum, according to many researchers, highly attractive models in advertising can have a negative effect on consumers and realistically attractive models should lead to higher positive attitudes. Finally, studies on the attitude-behavior relationship have shown that purchase
intentions are linked to product attitudes and that product attitudes are often correlated to model and brand attitudes (Ryan & Bonfield, 1972). These findings lead to a first hypothesis:
H1: Realistically attractive models lead to higher positive attitudes (towards the model, ad,
product and brand) and purchase intentions than highly attractive models for female consumers in fashion advertising
2.2 Viewer-source similarity
The inflated attractiveness of the models is not the only physical attribute that makes women perplexed, unsatisfied or irritated in fashion advertising. Diedrichs, Lee and Kelly (2011) revealed that in many cases, consumers were not satisfied with the uniformity of body sizes in media imagery. They showed that consumers want to see more diversity in terms of body weight and that they have more positive reactions to average-size models in the media. Additionally to this, it has even been shown that advertisements presenting thin models or average-size models are equally effective (e.g., Diedrichs & Lee, 2011; Dittmar & Howard, 2004). Thus, advertisers can consider using larger models in their ads and deliver equally successful - or perhaps more - campaigns. By doing so, they might also contribute to reducing body-focused anxiety among women (e.g, Halliwell & Dittmar, 2004). However, Gustafson, Hanley and Popovich (2008) also
discovered that women could have mixed feelings about advertisements portraying only plus-sized models. Indeed, these authors studied the Dove campaigns and found that women value a healthy body image that does not correspond with overweight persons. Research by Smeesters and Mandel (2006) is another example of work with similar findings; they discovered that females have higher self-esteem after being exposed to moderately thin models than moderately heavy models because of the similarity focus effect, i.e. they felt more similar to the moderately thin models. According to Smeesters, Mussweiler and Mandel (2010), females with a higher BMI adopt a similarity focus when being exposed to moderately or extremely heavy models and a dissimilarity focus when exposed to moderately or extremely thin models but that in both cases, women experience lower self-esteem and negative effect. Also, they found that women with a low BMI always express higher self-esteem after exposure to advertising models of any size whereas women with high BMI feel the opposite. Thus, it seems that women, regardless of how they perceive themselves, want to see models in advertising that reflect a healthy body image.
This leads to another question: What about other physical criteria such as age and cultural background? Much research has confirmed the underrepresentation of older women in
advertising (e.g., Baumann & de Laat, 2012), but do women actually want to see older women? Few studies on this topic have investigated on the types of models, appearance-wise, consumers actually want to see in advertising. In fact, Caballero and Solomon (1984) are among the few who contributed to the matter with their research on the impact of the use of physically attractive models in advertising on purchase behaviors. What they discovered was groundbreaking:
Customers may be more responsive to models perceived as similar to themselves. However, because their experiment focused on point-of-purchase advertisements, more research on the
topic was necessary. Some researchers recently advanced the ‘Beauty Match-Up Hypothesis’, which stipulates that women distinguish several types of good looks, and that in an advertising context some beauty ideals are more suitable for specific products than others (Solomon, Ashmore & Longo 1992). According to this hypothesis, if the model in an ad matches the product, the message is more consistent and the advertisement more accepted.
The author that been contributing a lot to research on viewer-source similarity is Ben Barry (2007) who conducted a very interesting research on the impact of the relative
effectiveness of traditionally attractive models versus realistically attractive models in advertising. He wanted to determine if women’s attitudes towards a brand and purchase intentions are the result of their evaluations of the model in an ad. He concluded that women loved seeing beautiful models but that they were more likely to buy the products being advertised if those models resembled them in age, body size and racial background. In other words, realistically attractive models and the brands with which they were paired, received more positive evaluations than traditionally attractive models (Barry, 2007). For his doctoral paper, Barry extended this research by adding India, China and Brazil, and by opting for a more quantitative approach. He confirmed his previous research and proved that at least in Western countries, women responded more positively to models that are similar to them. Barry performed his experiment using advertisements representing full body-length models and specific product categories, i.e. cosmetics (2007) and clothing (2012). Following the same idea, in 2004, the findings of a survey of more than 3000 women from ten countries were striking: 75% of the respondents said they wanted to see models of all ages and sizes in advertising (D’Agostino et al. 2004).
2.3 Social comparison and self-referencing
These studies have rarely taken into consideration the self on the comparison dimension, while simultaneity manipulating the comparison standard’s position. Smeesters, Mussweiler and Mandel (2010) have innovated in their research by taking into account the body mass index (BMI) of the subjects in their studies and, by doing so, demonstrated that ‘different comparison processes and outcomes occur for individuals differing in their relative positions on the
comparison dimension’ (Smeesters, Mussweiler & Mandel, 2010, p. 931). They showed that it is not the absolute size but instead the relative difference between the size of the consumer and the size of the model that is problematic and have negative consequences on the consumer’s self-esteem. They explored two different processes that define these social comparison outcomes: the similarity focus and the accessibility of self-knowledge (Mussweiler, 2003).
Since Festinger (1954) first introduced the social comparison theory, many researchers have explored this concept under various angles. According to him, individuals have a need to self-evaluate themselves by comparing their features with others and they do so with upward or downward comparisons and similar comparisons. When individuals compare themselves to others that are ‘superior’ on the judgment dimension (upward comparison), they disturb the way they feel about themselves, which results in negative emotions (Epstude & Mussweiler, 2009). These automatic social comparisons also occur in advertising settings, when female consumers are exposed to models (Bower & Landreth, 2001; Richins, 1001; Smeesters & Mandel, 2006). In general, because advertising represents ideals of beauty, social comparisons are usually upward (Martin & Kennendy, 1994). Thus, when female consumers compare themselves with models in the ads that are socially seen as very attractive or slim, they evaluate their self to be contrasted away from the comparison standard, in opposition to assimilated to it. It is this upward social
comparison that threatens the self-esteem, body satisfaction and other negative consequences of the female consumer because her self-judgments are so dissimilar to the comparison standard (Smeesters & Mandel, 2006). It has even been said that, when put in that comparisons, consumers can feel guilt, depression and even feelings of hopelessness, anger, anxiety and depression (Coulter, Zaltman, & Coulter, 2001). Only a brief exposure to media images of thin and highly attractive models can contribute to the decreased perception of woman’s own attractiveness (Ogden & Mundray, 1996).
It’s important to specify that, according to scholars, some individuals may experience positive affect also when engaging in upward comparison if their attributes ‘match closely with that of the upward comparison target’ (Watson, Wells & Hudson, 2011). In that case, the comparison provides the viewers with inspiration. Self-discrepancy theory implies that people feel negative affect when there is a difference between the self and the ideal self (Dittmar, 2005). It has been said that individuals with high levels of body image self-discrepancy engage more naturally in social comparisons and, therefore, experience negative outcomes related to mood, depression and self-esteem, when exposed to thin-ideal advertising. According to Watson, Wells, and Hudson (2011), it is possible that self-discrepancy moderates the relationship between thin-ideal advertisements and social comparison mechanisms.
Because they partly explain how media effects occur, social comparisons have been studied a numerous of time by scholars who were interested in the effects of highly
attractive/thin models in advertising on female consumers. They have often been investigated as being a mediator of the effects of such ads on body dissatisfaction and negative mood
self-concept development, which can be defined as the ‘totality of an individual’s thoughts or
feelings having reference to himself as a particular object’ (Rosenberg, 1979, p. 7) and comprises emotion and cognition since it is as much of a perception of a reflection, as well as an emotional response to that perception and reflection (MacKinnon & Heise, 2010). It has been said that this appearance comparison at least partially mediates the effects of the attractiveness of models in advertising. This comparison is established by two bases of self-referencing: criteria and normative. Critera is when a female consumer compare herself in terms or superiority or inferiority (better/worse) and normative is when she believes she is in harmony/agreement with others or disharmony/opposition with others (John DeLamater, 2013).
It’s been demonstrated that, in general, people tend to like others they perceive to be similar in various terms, such as demographics, culture, personality, attitudes, hobbies, social class, nationality, etc. (Freedman et al. 1974). The same principle applies for consumerism; it’s been shown that consumers prefer products or brands that match their perception of self-concept and use them as a vehicle for self-expression (Zinkhan & Hong, 1991). Thus, in advertising settings, the features of a brand’s advertisement should also match the viewer’s self. The concept of self-referencing supports this assumption. Self-referencing is defined as relating a stimulus to one’s self related knowledge structures (Escalas & Krishnamurthy, 1995). In other words, ‘it occurs when a consumer processes information by relating it to some aspect of their self, such as past experiences’ (Lee, Fernandez, Martin 2002, p. 181). Self-referencing also refers to self-evaluation, which is ‘the judgment of value, worth, or appropriateness of one’s abilities, opinions, or personal traits’ (Micu, & Coulter, 2012, p.81). According to Festinger (1954), self-referencing results in lateral comparison, which is the comparison with similar others such as peers. This lateral comparison with peers, or ‘average people’, shall not affect self-esteem
(Hanseung & Sohn, 2006). Thus, when a female consumer self-reference an average looking model in advertising, it should not affect her self-esteem. It has been shown by Krishnamurthy that female subjects, to the contrary of male consumers, do not like ambiguous stimulus and therefore do not engage in self-referencing when confronted to it. Reflecting these findings on self-referencing is the following hypothesize:
H2: When exposed to a stimulus advertisement, female consumers who engage in
self-referencing will experience more positive attitudes (Amodel, Aad, Aproduct, Abrand) and purchase intentions.
2.3 Self-Referencing as a mediating variable
In this study, we assume that model appearance will generate positive or negative effects on the consumer’s attitudes and intentions, but it is important to underline the psychological mechanism underlying these effects. It is expected that self-referencing should act as this cognitive processing strategy that will as a mediator. Indeed, a great deal of researchers, especially social cognition scholars, has studied self-referencing as mode of processing (e.g., Rogers, Kuiper, & Kirker, 1977). Escalas & Linville (1995) have shown that self-referencing can be a mediating variable and exposed that consumers mentally stimulating about a product
develop more favorable cognitive responses, i.e. more positive attitudes about this product and positive behavior intentions about trying and/or purchasing the product. According to these authors, this occurs because self-referencing focuses attention away from the attributes of the product and onto the self, decreasing ‘counterarguing and/or source derogation which both negatively impact the product attitudes’ (p. 30). The mental stimulation causes self-referencing which then produces positive affect that is transferred to the product. Moreover, the authors have seen that consumers who do not self-reference, either feel skeptical or disinterested emotions that
affect the product attitudes in a negative way. Other studies (e.g., Shavitt & Brock, 1984) have also stated that purchase intentions are increased when consumers make self-relevant judgments while exposed to an ad. Another study by Unnava, Burnkrant & Ahluwalia (2000) found that when usage experience and product involvement is low, the self does not have rich experiences associated with a targeted product and self-referencing does not contribute to enhanced
elaboration. Because of these reasons, it would be important to test communication strategies, like advertisements, to make sure consumers can relate to them.
2.4 Self-Referencing and ethnicity
There is a lack of ethnic models in advertising because advertisers think they would be linked to negative responses. It is evident that despite the increase of diversity in Western societies, most advertisers exclude ethic ‘minorities’ (sometimes not so minor!) from their marketing mix (Lee, Fernandez & Martin, 2002). This is baffling because it seems that
representing more racial diversity would be worthy for advertisers. In fact, Lee, Fernandez and Martin (2002) have found that ethnic minority models raised the attitudes and purchase
intentions of the same ethnicity without decreasing the attitudes and intentions of the majority group. Barry (2007) also came to the same conclusion in his study claiming that racial minority models don’t negatively affect the attitudes and intentions of the racial majority group. It seems that marketers can diversify the ethnicity of their models in their promotional materials and therefore improve the relationship between the brand and minority group without menacing the one with their majority group. Authors have discovered that self-referencing explains the positive affect that is generated when consumers view advertisements presenting models of the same ethnicity (Lee, Fernandez, Martin 2002). A model of the same ethnicity is a message that is
linked to a dimension central to an individual’s self. Self-referencing is activated, the message is better treated and, according to research, this process can generate more favorable ad evaluations since relating information to the self amplifies recall (Meyers-Levy & Peracchio, 1996). Overall, self-referencing enhances the remembrance of ads as well as their appreciation by consumers. The theory of distinctiveness explains why ethnicity is a dimension of the self (McGuire, 1984). More specifically, when an individual perceives ‘a complex stimulus, such as their self, they notice distinctive differences and characteristics which have greater informational value in discriminating themselves from others’ (McGuire, 1984). Lee, Fernandez and Martin (2002) have shown that when an ad refers to a salient dimension of a consumer’s self, like ethnicity, the consumer spontaneously self-references the ad, and this leads to more positive attitudes and purchase intentions. It seems that when consumers are involved in such self-referencing, they engage in similar social comparisons, which, occurs when one draws comparisons with hers who have similar attributes or qualities according to Festinger’s theory (Martin & Kennedy, 1994).
2.5 Self-referencing and viewer-source similarity
The source of the message – the model endorsing the product – is extremely important in contributing to the effectiveness of advertisements and success of brands. Acting as a conduct, the model shows the benefits of the product. Even more salient for the success of the ad is the source’s familiarity and similarity to the consumer (McGuire, 1984; McCracken, 1989). Past studies (e.g., Shavitt & Brock, 1984) have shown that individuals can’t be expected to engage in self-referencing solely when ‘instructed’ in the ad copy. In fact, an advertisement might be more efficient when individuals self-reference information by observing similarity judgments when viewing the ad. Therefore, it is even more relevant to include observable and
salient self-reference indicators for the consumers in the context of the ads (familiar setting, similar physical attributes of models, common activities presented, etc.). In fact, it has been shown that the more a consumer judges himself to be similar to the model of an ad, the more he can relate to it, i.e. engage in self-referencing (Kwai-Choi, Fernandez & Martin, 2002). When a consumer perceives these similarities (self-references), he views the model as being trustworthy (Deshpandé & Stayman, 1994). In sum, because self-referent judgments positively influence recall, consumer attitudes and purchase intentions, and that consumers self-reference salient dimension of their self such as physical attributes (age, race, body size), one can argue that ads would be more effective if the models would look less ‘idealized’ and more similar to the targeted consumers.
It’s been said that when a product is congruent with a consumer’s self-concept, he engages in self-referencing, which leads to more positive attitudes and purchase intentions (e.g., Zinkhan & Hong, 1991). We have also mentioned previously that viewers might also self-reference models in advertising. Thus, we can expect that viewer-source similarity, with various physical attributes such as race, age and body shape, would generate positive attitudes and higher purchase intentions. Past research proposed that self-referencing mediates the effect of a number of ad stimuli factors on attitudes and intentions, including a model’s attractiveness (Debevec & Kernan, 1987) and the various visual and verbal elements of ads (Debevec & Romeo, 1992). We can than hypothesize:
H3: For female consumers, self-referencing acts as a mediator between the effect of the
attractiveness of the model on attitude toward the model (Amodel), attitude toward the ad (Aad), attitude toward the product (Aproduct), attitudes toward the brand (Abrand) and purchase
H4: Viewer-source similarity, i.e. when female consumers share physical similarities (age, body
shape and race) with models in fashion advertisements, generates similarity judgments that moderates the relationship between the attractiveness of a model and self-referencing
2.6 Theoretical framework & present study
To fully assess the goals of this study, here is the original conceptual model with the hypotheses followed by more details about the present study.
Figure 1. Conceptual model with hypotheses
This study proposes that other salient dimensions of one self’s represented in an
advertisement such as age, gender and overall general appearance can also engage consumers in self-referencing. Because Lee, Fernandez and Martin (2002) have found that self-referencing leads to positive product and ad attitudes, but not necessarily to positive brand attitudes, the present study will test this. Also, past research (Unnava, Burnkrant & Ahluwalia, 2000), has
found that the subjects did not self-reference high involvement product categories more than low involvement product categories. This study will also assess this.
Past research has explored the effectiveness of larger/thinner or highly attractive/normal attractive models in advertising, but there has also been confusion pertaining the manipulation of weigh and attractiveness (e.g., Irving, 1990). To date, few studies have tested the effectiveness of advertisements showing non-manipulated, average looking but still attractive models. Moreover, little research has been done on how specific attributes of the appearance, such as the size, age and race of the models, can influence purchasing decisions in the fashion and beauty industries. Also, apart from the work of Ben Barry, few scholars have studied the possible positive effect on ad effectiveness of the actual –and not only perceived– physical similarity between the model in an ad and the targeted consumer. The approach used in this study is to examine if advertising effectiveness derives from the information-processing theory (Hamilton, 2007) and includes attention and behavioral intentions processes, which are attitudes toward the model, the ad, the product and the brand as well as purchase intentions. Moreover, unlike past research, this study grasps the importance of some ‘creativity controls’ that are crucial in fashion advertising. Just like Barry (2007) concluded, female consumers aspire to an image in fashion that is glamorous. They also assess a great deal of importance to the artistry and creativity of the fashion
advertisements. As Barry (2013) explained: ‘the underlying message is that fashion and beauty advertising continues to sell aspiration, but it is not a standardized model’s age, size or race that is primarily aspirational; it is the clothes, styling and creative direction of the shoot’ (p. 5).
RESEARCH METHOD
3.1 Design
The present study is investigating the effect of an instructional set on the dependent variables, attitudes and purchase intentions. Hence, the participants were randomly exposed to two conditions: either an image of an average size/attractive model that was not digitally altered or one image of a highly attractive and thin model that was also retouched. In addition, this study examined the correlational relationship between the form of processing (self-referencing) in which women engaged when exposed to ads and their attitudes and purchase intentions.
3.2 Ad models and ad stimulus
Two mock fashion ads were created specifically for this study: one with a model who fits the description of the targeted consumer (the ‘realistic’ average model) and another one with a very attractive and professional model. The very attractive woman was a professional model hired from the Dulcedo Model Agency based in Montreal. A managerial goal of the present study was to potentially guide a specific brand, BonLook, on its future choices of model in advertising campaigns. Therefore, the ‘average’ model used was physically reflecting the target market of the brand in age, race and body size; Caucasian, 22 years old and a BMI indicating a normal weight. As previously brought up, many scholars have outlined the importance of models to still be somewhat attractive. Although more diversity in race, body size and age seems to be equally or more effective in fashion advertising, the importance of the attractiveness of the models for advertising effectiveness remain, regardless of other features (e.g., Barry, 2007; Halliwell & Dittmar, 2004). Therefore, manipulation checks on the attractiveness of both models had to be measured. A pre-test measuring the attractiveness of the models were conducted with 25 young women.
The ‘average-looking’ model had a normal body shape (BMI of 22, between 19 and 25), which reflected the size of the average targeted consumer by the advertised brand. As for the professional model, she had a body size considered as underweight. This is even more interesting considering that studies have shown that most people do not perceive similarities with very thin models (BMI less than 18.5) or overweight models (BMI more than 25) (Yul et al., 2011). Many studies have also demonstrated that the use of an average-size model produces more positive attitudes and purchase intentions than very thin or plus-sized models (e.g., Sohn & Youn, 2013). No studies have found it to be more effective to use thin models over average sized models; in fact, in terms of advertising effectiveness, very thin models are either equally or less successful than average models (e.g., Dittmar & Howard, 2004; Lennon et al., 1999; Peck and Loken, 2004). Both models were around 22 years old (the average age of the brand’s target market) and both of them were Caucasian so is the case for 75% of the brand’s customer base.
A limitation of past research is the inclusion of ad copy instructing the viewers, for example, to focus on their body or wellness attitudes. Because such instructions could have a considerable effect on women’s perceptions and responses, the fictional advertisements did not include ad copy, nor did it reveal a logo or tagline of the brand.
A rigorous way of studying this matter was by manipulating the ‘physical attribute’ variable and controlling the other ones in an advertisement. Hence, in order to do so and for the research not to be biased, it was necessary to organize a photo-shoot with a professional team (i.e. fashion photographer, hair & makeup artist, stylist and model). This process was more realistic than comparing real models with digitally modified ones like past researchers did (Dittmar & Halliwell, 2005). All the features that can distort the respondent’s perceptions, such as differences in facial expressions or style were then controlled. Both models were wearing the
same clothes and were shot in the same environment (matching light, background, position, etc.) to avoid consumers being influenced by other variables. An interesting element of Barry’s research (2007) was the revealed importance of visual codes, particularly body position and facial expressions. It seems that ‘absent’ or sad expressions are unappealing to female consumers so are ‘sexualized’ or submissive positions and signs of weakness. Women want to see confident, smiling models that are represented in realistic activities. These criteria were taking into account while shooting the models and both of them had the same position and smiling expressions (see Appendix 1). Moreover, it seems that even though female consumers want to see models that look like them in fashion advertisements, they still favor arousal, creative ads that evoke fantasy and that were styled with a fashionable aesthetic. Also, Past studies assessing the matter have often compared traditionally attractive models with unattractive models instead of realistically attractive ones (e.g.: Kahle and Himer, 1985; Kamins 1990). Information regarding the physical attributes of both of the models were collected (height, etc.) to compare them with the
respondents.
This eyewear brand was selected because European consumers did not know it and because it retails a fashion product that is very relevant for this study for various reasons. First, past studies on viewer-source similarity have focused on advertisements of clothing and
cosmetics advertisements (e.g., Bower, 2001; Barry & Bell, 2007). Therefore, it is unclear if this strategy is as much effective for other product categories such as accessories. In fact, a gap from past research on the topic is that it has not tackled a product that is not related to the body and weight. Eyeglasses are considered a very fashion-oriented accessory, as well as an attractiveness-related and self-enhancement product (or else people would buy contact lenses). According to
Barry (2007), women appreciate viewer-source similarity not only because it makes them feel targeted/included but also because they can picture the advertised clothes on themselves. It will then be interesting to know if these results are as conclusive for other product categories such as accessories.
Also, Bower and Landreth (2001) suggested that highly attractive models are perceived to have greater source expertise (which leads to higher model attitudes) for enhancing products than normally attractive models so it is interesting to see if findings of this study will match this. According to the same authors, realistic models could also be more effective for problem-solving products such as eyeglasses. There were actually mixed findings about the effectiveness of highly vs normal attractive models depending of the product being beauty-enhancing or problem-solving. Eyeglasses fit both categories, which adds up to their relevance for their study. Also, eyeglasses represent a relatively high-involvement product since it is a fairly expensive and health-related accessory that is worn on the face i.e. a ‘riskier’ purchase. It has been said that self-referencing is more evident under high involvement (Burnkrant & Unnava, 1995).
Eyeglasses are also interesting because there are not as much differences between the brands as there are for clothing, which may influence the importance of viewer-source similarity or the process of self-referencing.
3.3 Data collection and procedure
The participants were told the questionnaire was on diversity in fashion advertising, which is a topic that is very much discussed in the media nowadays. Even though ‘fashion diversity’ can somewhat be an indicative of the purpose of the study, it still does not reveal the full intentions of the study. Respondents completed a survey of 20 questions through the
it should take a maximum of 10 minutes and an average time of 7-8 minutes. They were told it would take 7 minutes and statistics showed after survey closure that this was actually the average time completion. They were shown a first page with information and instructions about the questionnaire, such as the procedure, the incentives (raffle of Amazon vouchers) and contact details. It was also specified that it be completed on a voluntary basis, that there was no right or wrong answer and that the data would be kept anonymous. I knew many people would answer the questionnaire on their mobiles; therefore the width of the survey was reduced and the answer choices were adapted for such mode of completion. Only the 2 photographs were too wide but then the respondents were advised to scroll to the right. Because many questions were assessing personal data (e.g., weight), the participants were not forced to complete the questions. Unlike past research, one goal of the present study was to not only rely on student samples. Therefore, a wider range of women was solicited to participate in the survey. The data was gathered by snowball sampling, through e-mail and social networks and went extremely well, especially considering that only women could fill it, with a response rate of more than 80% and 380 people who took interest in the research and opened the survey. Respondents were mostly from Western Europe and Canada since they are similar markets. Besides, literature on women’s evaluations of models in advertising indicated that the attitudes are the same for Northern America and Western Europe. For example, Asian, Mexican or African consumers do not feel the same about the topic so they were not solicited. Each response was linked to an IP address showing the location of the participants so this could be verified. Out of the total of 380 respondents, 15 who identified themselves as men were excluded since the study only focuses on women’s attitudes and
behaviors. Moreover, even though it was instructed that the survey was only targeting women, it is obvious now that it was prudent to include a gender question to avoid bias. Also, 25 women
who failed the ‘instruction question’ were also not considered. This question was asking
respondents to ignore the five stated choices and skip to the next question (see question 4.9 of the questionnaire in the Appendix). Although it is possible that some respondents might have
answered ‘quickly’ only to this question while paying attention to the rest of the questionnaire, we can assume, that in general, women who ticked an answer were not taking the time to take the survey properly. Because an interesting amount of data was collected, the questionnaires that were partially completed were also removed from the data set. An imputation using Hotdeck was considered for the missing values but the idea of manipulating ‘real’ data from complete
questionnaires was not only stimulating but also obviously more reliable.
3.4 Sample
Out of the total of 380 respondents, 299 complete questionnaires were used for this study and they were almost equally divided between the 2 conditions (N=152 for realistic attractive models identified as RAM and N=147 for traditionally attractive models identified as TAM in further results).
3.5 Measurement scales
To ensure reliability, all the scales were adapted from previous research and none were my own constructions. This section will describe the scales that were used to build the survey. Note that there are more scales in the questionnaire that were included for additional research.
3.5.1 Manipulation check
Both model were assessed as ‘attractive’, more specifically an average of 4 for the RAM and 5 for the TAM on a 5 point scale. The item for the manipulation check was part of the scale for the model attitudes and respondents had to rate the model on a 5 point Matrix table scale
from ‘not attractive’ to ‘very attractive’. This is crucial because extensive literature had
suggested that attractiveness is a major determinant of advertising effectiveness, regardless of the other attributes of a model. It also confirms our pretest on attractiveness. Here are the graphs showing the attractiveness.
Figure 2. Attractiveness of the realistically attractive model
3.5.2 Independent variable: Attractiveness
Originally, the independent variable was a binary variable (the TAM or RAM model) because it was a condition. For the results to be more accurate and ‘robust’, we needed to use a continuous variable. Hence, the ‘attractiveness’ variable was used to test the model. The item for attractiveness was “The model in the ad looks very good” and respondents had to choose an answer on a range from ‘strongly disagree’ to ‘strongly agree’. The variable for attractiveness is formed from only one item, which is acceptable since it is the independent variable and the question is straightforward.
3.5.3 Dependent variables
The dependent variables were the attitudinal attitudes, i.e. the model attitudes, the ad attitudes, the product attitudes, the brand attitudes and the purchase intentions.
Attitudes toward the model (Amodel) were assessed on five, five-point scales and were querying on how competent, likable, persuasive, attractive and believable were the woman in the ad (Martin, Less & Yang, 2004). As seen in the following table, the Cronbach’s Alpha is ‘good’ since it’s above 0.7 so the scale is considered reliable.
Table 1
Reliability of the model attitudes scale
Cronbach's Alpha N of Items
,741 5
It was particularly important to measure the product attitudes since they are a great indicator for the eyewear brand represented in this study. Respondents had to rate the product on
6 points scale on 7 specific qualities such as quality, worthiness and taste. Reliability is shown below.
Table 2
Reliability of the product attitudes scale
Cronbach's Alpha N of Items
,881 7
As for the attitudes towards the ad, they were measured by this 6 points scale
(exceptionally unfavorable - exceptionally favorable): “How would you describe your reaction to the advertisement?” The brand attitudes were evaluated by the item ‘This ad is much better than other ads for products in this product category”. Participants had to choose from ‘strongly disagree’ to ‘strongly agree’. As for the purchase intentions, they were assessed by the following question: ‘Please give your first feelings or impressions toward using the product’. These scales were formed of only one item since they are straightforward.
3.5.4 Mediator: Self-Referencing
Considering the purpose of this study, it was more relevant to test the actual occurrence of self-referencing. This could be done by using measures that tap in this perceived similarity between the consumer and model. As Debevec and Iyer (1988) mentioned, this method should allow a more powerful test of the self-reference effect on the attitudes of the consumer and its meditational properties. Self-referencing was tested on a 5 point-scales adapted from previous research (e.g., Burnkrant & Unnava 1995; Debevec & Iyer, 1988; Debevec & Romeo, 1992; Meyers-Levy & Peracchio, 1996). Specifically, the items, anchored by strongly agree/strongly
disagree were: “I can easily picture myself using the advertised product”, “I can easily form similarity judgments between myself and the advertising model”, “I can easily relate myself to the advertising model”, “The ad seemed to be written with me in mind”, “The ad made me think about my own experiences with a similar product”, “The ad seems to relate to me personally”. In order to be sure that the construct was reliable, a Cronbach’s alpha test was conducted. We know that the Cronbach’s Alpha need to be above .60 to be reliable and 0,90 and above to be excellent, which is the case (see table 3). Moreover, table 4 shows that the scale would be less reliable if items were deleted.
Table 3
Reliability of the self-referencing scale
Cronbach's Alpha Cronbach's Alpha Based on Standardized Items
N of Items
,905 ,905 6
Table 4
Reliability of the self-referencing scale for each item
Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item Total Correlation Cronbach's Alpha if Item Deleted I can easily picture myself using the
advertised product
13.86 28.450 .637 .903
I can easily form similarity judgments between myself and the model
14.23 28.079 .727 .890
I can easily relate myself to the model 14.55 26.470 .785 .881
The ad seems to be written for me 15.12 26.312 .833 .874
The ad made me think about my own experiences with a similar product
15.22 27.943 .681 .897
The ad seems to relate to me personally
3.5.5 Moderator variables
Various questions assessing demographics particularities were important not only to better understand the particularities of the sample but also to test viewer-source similarity (age, height, weight, race, etc.). To assess viewer-source similarity, race, age and body shape were chosen since they were the most salient attributes and also because extensive past literature has suggested that relative variance in age, body shape and racial background should not jeopardize advertising effectiveness in Western countries (e.g., Sohn & Youn, 2013)
Age
Because the questionnaire was targeting a younger audience, most of these 299 female respondents were between 18 and 34, which is the age of BonLook’s target market; specifically, 40% were between 25 and 29 years old, 33% between 18 and 24 years old and 22% between 30 and 34 years old. Below you can see the chart for the age of the whole sample as well as a graph showing that there is not much differences between the 2 conditions.
Figure 4. Age of the sample
Figure 5. Age difference between the 2 conditions
1% 33% 40% 22% 4% 0% Age Under 18 18 to 24 24 to 29 30 to 34 35 to 39 50 years old and up
Body shape
Respondents were asked to write their height and weight in the survey so it would be possible to know their body shape. Calculating the BMI, which is obtained by dividing the weight by the height2, assesses the body shape. The respondents had an average BMI of 22, which is considered ‘normal weight’ and it reflected the size of the realistically attractive model. The traditionally attractive model had a BMI considered as ‘underweight’ according to her measurements on her online profile page of the modeling agency that represents her. Women with a BMI < 18.5 are considered underweight, women with a BMI ranging from 18.5 to 25 are considered ‘normal-sized’, women with a BMI ranging from 25 to 30 are considered overweight and women with a BMI > 25 are considered obese.
Figure 6. Body shape of the sample
Figure 7. Body shape difference between the 2 conditions
12%
72%
13% 3%
Body shape (BMI)
Underweight Normal Overweight Obese
Racial background
As for the racial background, the following graph and chart below clearly show that the respondents were mainly white.
Figure 8. Racial background of the sample
Figure 9. Racial background difference between the 2 conditions
RESULTS
4.1 Overview
First, IBM SPSS Statistics was used for descriptive statistics and to test the existing relationships between the variables. Second, the IBM SPSS Amos software was employed to define more exactitude pertaining the relationships between variables and to test the mediating and moderating effects of certain variables. Amos was a good tool to test the model since it allows conducting path analysis – or structural equation modeling (SEM). SEM helped to assess the relative importance of the different predictor variables of the model on the outcomes of the
88% 3% 2% 3% 0% 4% Race White/Caucasian Black/African American Hispanic or Latino Asian Native American or American Indian Other
dependent variables. Because the model that needed to be tested in this study had a mediator and moderator, we needed to test multiple regressions focusing on causality. Amos also has the bootstrapping tool, which is resourceful when dealing with small samples, and to allow for errors in measurements in the statistical model. First thing to do was to setup the bootstrap and then organize the model.
Here’s a summary of the conducted analysis that needed to be performed:
1. Independent and positive effect of model attractiveness on consumer attitudes and intentions
2. Independent and positive effect of model attractiveness on self-referencing 3. Independent and positive effect of self-referencing on consumer attitudes and
purchase intentions
4. Mediating effect of self-referencing on the relationship between model attractiveness and consumer attitudes and purchase intentions
5. Moderating effect of viewer-source similarity (age, body shape, race) on the relationship between attractiveness of the model and self-referencing
4.2 Correlation and significance
First, as mentioned previously, because the independent variable was a binary variable (the TAM or RAM condition), the ‘attractiveness’ variable was used as independent. It was relevant to use this variable because a discrepancy in attractiveness was distinguished between the 2 models. Yet, in order to test our first hypothesis, we needed to see if the RAM model led to more positive attitudes and purchase intentions than the TAM model. Although the correlation is very low (because of the binary variable), we saw that the RAM model led to more positive model attitudes, but not purchase intentions. Our first hypothesis 1 was partly confirmed. This
correlation signifies that, for this sample, the more attractive a model is, the highest are the purchase intentions.
Figure 10. Attractiveness on purchase intentions
However, since it was decided to use the attractiveness variable as independent, I still tested if there was a direct effect of this variable on the attitudes and purchase intentions. The correlations table on the next page shows correlations between the independent, mediating and depending variables.
Table 5
Correlations depending of condition
Table 6
Correlations Coefficients between the main variables Attractive-ness Model Attitudes Ad Attitudes Product Attitudes Brand Attitudes Self- Referencing Purchase Intentions Attractive-ness Pearson Correlation 1 ,460** ,272** ,412** ,294** ,258** ,341** Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,903 ,293 ,267 ,419 ,357 ,254 ,406 Model Attitudes Pearson Correlation ,460** 1 ,560** ,639** ,508** ,502** ,444** Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,293 ,448 ,387 ,458 ,435 ,348 ,373 Ad Attitudes Pearson Correlation ,272** ,560** 1 ,577** ,519** ,506** ,503** Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,267 ,387 1,069 ,639 ,686 ,542 ,651 Product Attitude Pearson Correlation ,412** ,639** ,577** 1 ,573** ,609** ,635** Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,419 ,458 ,639 1,145 ,785 ,674 ,851 Brand Attitudes Pearson Correlation ,294** ,508** ,519** ,573** 1 ,700** ,501** Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,357 ,435 ,686 ,785 1,634 ,926 ,803 Self Ref. Pearson Correlation ,258** ,502** ,506** ,609** ,700** 1 ,581** Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,254 ,348 ,542 ,674 ,926 1,071 ,754 Purchase Intentions Pearson Correlation ,341** ,444** ,503** ,635** ,501** ,581** 1 Sig. (2-tailed) ,000 ,000 ,000 ,000 ,000 ,000 Covariance ,406 ,373 ,651 ,851 ,803 ,754 1,569
Note. **Correlation is significant at the 0.01 level (2-tailed). N = 299
Variables RAM Model attitudes 0,0058 Product attitudes 0,0347 Brand attitudes 0,1162 Ad attitudes 0,0497 Purchase intentions -0,0465
As predicted, we see that attractiveness had a positive relationship with all the attitudes and purchase intentions, as well as self-referencing, the mediator variable of the model. The p value is significant for all relationships and Pearson correlation is particularly important between self-referencing and brand attitudes since we know that if r = +.70, it is sign of a very strong positive relationship. The weakest relationships are between the ad attitudes and attractiveness as well as the brand attitudes and attractiveness, although they are all positive. Thus, we can confirm the following:
- There is an independent and positive effect of model attractiveness on consumer attitudes and intentions
- There is an independent and positive effect of model attractiveness on self-referencing - There is an independent and positive effect of self-referencing on consumer attitudes and
purchase intentions
As additional information, here are the means, modes and standard deviation for the main variables. Table 7 Descriptive Statistics Ad Attitudes Purchase Intentions Brand Attitudes Model Attitudes Product Attitudes Self Referencing Mean 3.73 3.27 2.73 3.5478 4.1581 3.1409
Mode 4 (‘Favorable’) 4 (‘Somewhat
Likely’) 3 (‘Somewhat Disagree’) 3.60 a 4.29 3.50 (‘Somewhat Disagree’) Std. Deviation 1.034 1.253 1.278 .66953 1.07015 .93522
Note. Multiple modes exist. The smallest value is shown. N = 299
4.3 Mediator and moderator effects
A mediation between variables requires relationships (correlation) between the mediator and all variables that are implicated. We had proven independent and positive relationships
between these variables, i.e. between attractiveness and self-referencing, attractiveness and attitudes (Amodel, Aproduct, Aad, Abrand) and intentions, as well as self-referencing and attitudes and intentions. However, mediation was still not proven. To do so we needed to do a path analysis in Amos was to test multiple regressions. It allows to predicting an outcome variable from one or more predictor variables. Here are the standardized correlations for the regressions that can lead us to conclusions regarding the mediator.
Table 8
Standardized Regression Weights and P value
Variables Estimate P
Self-Referencing <--- Attractiveness .260 *** Model Attitudes <--- Self-Referencing .411 *** Model Attitudes <--- Attractiveness .354 *** Ad Attitudes <--- Self-Referencing .301 *** Ad Attitudes <--- Model Attitudes .405 *** Ad Attitudes <--- Attractiveness .008 .877 Product Attitudes <--- Attractiveness .135 .002 Product Attitudes <--- Ad Attitudes .214 *** Product Attitudes <--- Model Attitudes .298 *** Product Attitudes <--- Self-Referencing .316 *** Purchase Intentions <--- Attractiveness .104 .029 Purchase Intentions <--- Self-Referencing .284 *** Purchase Intentions <--- Ad Attitudes .154 .005 Purchase Intentions <--- Model Attitudes .073 .222 Purchase Intentions <--- Product Attitudes .377 ***
As confirmed in the r correlations previously shown, there is a positive (0.260) and significant (p < .01) relationship between attractiveness and self-referencing (0.260). Also as mentioned before, there is a positive (0.104) and significant relationship between attractiveness and purchase intentions (0.029). The same goes for the relationship between attractiveness and model attitudes (0.354). There is also a relationship (0.411) between self-referencing and model attitudes as well as between model attitudes and ad attitudes (0.405). However, to the opposite of what was found out in the first correlations test in SPSS, the p value (0.877) was not significant for the relationship between attractiveness and ad attitudes. It is the same for the link between ad attitudes and purchase intentions. The relationship between product attitudes and purchase intentions were however significant (p < .01) and positive (0.377). These findings confirmed the mediating effect of the self-referencing, but also a new mediating effect of attitudinal attitudes on purchase intentions. Attractiveness had a direct effect on the purchase intentions that was
partially mediated by self-referencing. Hypothesis 2 and 3 are partly confirmed (more justification is presented in the Discussion section).
Next, it was required to verify the moderation effect of the viewer-source similarity. The literature review led us to believe that similarity can moderate the relationship between the attractiveness of a model and self-referencing. Because we wanted to test more than one physical attribute, i.e. age, body shape and race, we needed to test them separately since they represent 3 different variables. Regrouping them would make the variable too restrictive. First we calculated the interaction term, which is generated by multiplying the predictor by the moderator, i.e. age and attractiveness. The standardized regression coefficients indicated that age had a negative and somewhat significant (p value close to 0.05) influence on self-referencing. As we knew, a
variable and a dependent variable. Therefore, when age is high, self-referencing is low. Age also had a direct effect on self-referencing.
Table 9
Standardized Regression Weights and P value for Age
Variables Estimate P
Self-Referencing <--- Age & Attractiveness -.104 .062 Self-Referencing <--- Age .123 .026
The Chi-square was low (7.753) and probability level of the Chi-square was close to 5 (0.458) so we could conclude that there was no better possibility of model. Other indicators confirmed that the model could not be improved, such as the CFI that was at the maximal value (1.0) and the RMSEA at the lowest value (0.0). The same procedure was conducted for the body shape. Body shape is represented by the body-mass index (BMI) that was calculated by obtaining the height and weight of the participants.
Table 10
Standardized Regression Weights and P value for BMI
Variables Estimate P
Self-Referencing <--- BMI & Attractiveness -.071 .202 Self-Referencing <--- BMI -.194 *** Note. *** = p < .01
The body shape of respondents has a direct negative (-.194) and significant effect (p < .01) on the self-referencing, i.e. that the highest the BMI is, the lowest the self-referencing is.
