Master thesis BA Marketing

(1)

Master thesis BA Marketing

Does beauty always sell?

The interaction effect of model attractiveness and product type

on product evaluations.

University of Groningen

Sandra de Weerd s1541544 Master thesis BA Marketing

Faculty of Economics and Business Sandra de Weerd s1541544 Master thesis BA Marketing

(2)

Abstract

The commonly used sentence “beauty sells” is the reason this research is done. In previous research, contradicting results are found. Some articles say that the more beautiful the model, the better the product evaluations. Others state that this depends on different factors like product type and fit with the product. In this thesis, the interaction effect of product type and model attractiveness on product evaluations is investigated. We expect that for attractiveness-relevant products, product evaluations will be better when using a highly attractive model and for attractiveness-irrelevant products, product evaluations will not be better when using a highly attractive model. Also the effect of the moderators “involvement” and “nature of the product” is researched. For nature of the product, two variables are researched, hedonic value and utilitarian value.

We expect that for high involvement/attractiveness-irrelevant products, product evaluations will be better when using a normally attractive model compared to a highly attractive model. For low involvement/attractiveness-irrelevant products and for both high and low

involvement/attractiveness-relevant products, we expect that product evaluations will be better when using a highly attractive model compared to a normally attractive model. For products with more hedonic values we expect that product evaluations will be better when using a highly attractive model compared to a normally attractive model. For products with more utilitarian values we expect that for attractiveness-irrelevant products, product evaluations will be better when using a normally attractive model compared to a highly attractive model and for

attractiveness-relevant products product evaluations will be better when using a highly attractive model compared to a normally attractive model.

(3)

purchase intentions. In this questionnaire also questions about how involved participants were with the product and about the nature of the product were asked.

The main outcome of this research is that the more attractive the model is, for both

attractiveness-relevant and attractiveness-irrelevant products, the better the product evaluations are. However, with low model attractiveness, product evaluations for attractiveness-irrelevant products are significantly higher than product evaluations for attractiveness-relevant products. For the effect of the moderator involvement, no significant influence is found. For hedonic value, no significant effect on the interaction effect of model attractiveness and product type on product evaluations is found. For utilitarian value it is found that when participants see an attractiveness-irrelevant product as high utilitarian, product evaluations are better when using a normally attractive model compared to a highly attractive model. When participants see an attractiveness-irrelevant product as low utilitarian, product evaluations are better when using a highly attractive model compared to a normally attractive model. When participants see an attractiveness-relevant product as high utilitarian, product evaluations are better when using a highly attractive model compared to a normally attractive model.

In further research it is important to reinvestigate these moderators by manipulating involvement in a different manner and research the nature of the product more extensively. Involvement can be manipulated in the setting of an experiment. I this way a causal effect of involvement on the interaction effect of product type and model attractiveness on product evaluations can be investigated. It is also important to look at control variables, like age, origin and social status, which can have an influence on the interaction effect of product type and model attractiveness on product evaluations. Previous research already found that age and area of residence can influence the effect of model attractiveness on product evaluations. It is found that women between 35 and 44, living in urban areas have higher self-esteem and so are more positive towards highly

(4)

So, for both attractiveness-relevant products and attractiveness-irrelevant products, it is better to use a highly attractive model compared to a normally attractive model. But when looking at the moderating influence of nature of the product, it is better to use a normally attractive model for an attractiveness-irrelevant/ high utilitarian product. This can be explained by the fact that for products which score high on utilitarian values, consumers look at product arguments instead of simple cues. For attractiveness-irrelevant products, the attractiveness of the model has nothing to do with the product and so model attractiveness cannot be seen as a product argument. So, when making an advertisement campaign it is important to find out if the product scores high on utilitarian values. If this is true, for attractiveness-irrelevant products it is better to use a normally attractive model and for attractiveness-relevant products it is better to use a highly attractive model.

For further research it is important to investigate the previous moderators more extensive, by looking at new moderators and by investigating the effect of not using a model at all. In this research it is found that in general for both attractiveness-relevant as attractiveness-irrelevant products, product evaluations are better when using a highly attractive model compared to a normally attractive model. However, it is found that for high utilitarian/ attractiveness-irrelevant products it is better to use normally attractive models compared to highly attractive models. While previous research shows that the use of highly attractive models can have a negative influence on women’s self-esteem (Richins, 1991), it could be interesting to find out if advertisements without a model in it can be equally or more effective compared to

(5)

1. Introduction

1.1 Motive for the research

Many advertisers make use of highly attractive, thin models; they are seen in all types of advertisements all over the world. But also many people remember the Dove-campaign with the “normal models” which received a lot of positive reactions. “Beauty sells” is a commonly used sentence, but is it always true? There is done a lot of research on the topic of the effect of model attractiveness on the evaluation of a product and the findings are contradicting. Some articles find that the more beautiful the source, the more positive the product evaluation (Kahle and Homer, 1985). Other articles find that instead of model attractiveness, mostly model expertise, the perceived ability of a source to make valid claims, is important for positive product evaluations (Bower and Landreth, 2001). For highly attractive models, this means that her attractiveness can be seen as model expertise in the case of attractiveness-relevant products. This is by some authors mentioned as the match-up hypothesis, which means that there has to be a fit between the model and the product (Till and Busler, 2000). When a model is highly attractive, this attractiveness (the beauty of the model) fits with the purpose of an attractiveness-relevant product (to make the consumer more beautiful).

(8)

2

attractive models would be better because consumers would not believe that a highly attractive model has the same (beauty) problems as normal people do.

According to Bower and Landreth (2001), further research is needed to understand the differences or advantages of including one type of model in an ad for problem-solving products over another. For example, the lack of difference in product evaluations between advertisements with normally attractive models and advertisements with highly attractive models, found in their research, is a reason to further investigate this topic. In this thesis, the role of involvement on this relationship will be investigated, which is also suggested by Till and Busler (2000). Involvement can, in this relationship, be defined as the personal significance that individuals ascribe to the product (Te’Eni-Harari et al. 2009). Under high involvement, consumers are persuaded by strong product arguments relating to the product’s features. Under low involvement, consumers are persuaded by peripheral cues in the advertisements, for example the attractiveness of the model in the advertisement (Te’Eni-Harari et al. 2009; Petty and Cacioppo, 1983). Previous research about this subject has contradicting findings. Häfner and Trampe (2009) found that for high involvement situations normally attractive models are more effective and for low involvement situations highly attractive models are more effective. Contrary, Kahle and Homer (1985) found that under both high- and low involvement, physical attractiveness of the model has a positive impact on product, brand and attitude recall. This is also found by Schaefer and Keillor (1997). They based these outcomes on the match-up hypothesis, which means the fit between the model and the product. This is especially true for attractiveness-relevant products and highly attractive models, because the beauty of the model fits with the goal of using the product.

(9)

3

can be seen as utilitarian products and beauty enhancing products can be seen as hedonic products. In this thesis, there will not be made a difference in problem solving and beauty enhancing products, but there will be asked questions about the involvement level and the nature of the product (hedonic or utilitarian) which will later be used as moderators in the experiment. In this manner the role of involvement and nature of the product, on the interaction effect of product type and model attractiveness on product evaluations will be investigated, without choosing different products (high involvement products/low involvement products or hedonic products/utilitarian products). I have chosen this approach because I think that one product, for example foundation, can be seen as a problem solving product by one consumer and at the same time the product can be seen as beauty enhancing by another consumer. Cosmetic products are used as attractiveness- relevant products, to make sure there is a fit between attractiveness of the model and the chosen product.These products will be compared with products that have nothing to do with appearance, attractiveness-irrelevant products.

Investigation of this subject is also very important because research found that the use of highly attractive, thin models has a negative influence on women’s self-esteem about their physical appearance (Richins, 1991). It is important to find out if it is always necessary and positive to use highly attractive models or if the use of normally attractive women leads to the same or even better product evaluations.

1.2 Problem statement

There are some contradicting findings in literature about this topic and a lot of suggestions for further research. So, the goal of this thesis is to research the use of highly attractive versus normally attractive models for both attractiveness-relevant and attractiveness-irrelevant products, and in which way the interaction effect between model attractiveness and product type on product evaluations is influenced by involvement and the nature of a product. To find this out, the following problem statement has to be answered:

(10)

4

1.3 Overview of the research

(11)

5

2. Theoretical framework

A lot of research has been done to investigate the effect of model attractiveness on product evaluations. As said in the introduction of this thesis, the findings of several researches are contradicting. In this section these researches will be studied and the differences and similarities will be lined out. The role that involvement and nature of the product can be playing on this effect will also be investigated.

2.1 Model attractiveness and the match-up hypothesis

Highly attractive models are often used in advertisements and their effect on product evaluations is, according to several studies, positive. Kahle and Homer (1985), found that the more beautiful the model, the better the product evaluations. They investigated this effect based on social adaptation theory. Social adaptation theory argues that attitudes, values and other social cognitions are used by people to adapt to the environment. Persons seek equilibrium with the environment by assimilating new information into existing schemata while accommodating mental structures to incorporate new, discrepant information (Homer and Kahle, 1986). Individuals evaluate information in terms of its adaptive significance. If an individual decides a certain piece of information has achieved its potential in terms of facilitating adaptation (with one's environment), processing of that particular information will cease, and processing attention will be directed elsewhere (Homer and Kahle, 1986). For example for low involvement products, consumers only look at an advertisement for about two seconds before moving to the next source of information, since the advertisement is not advancing adaptation while for high involvement products consumers may spend more time reviewing the advertisement.

(12)

6

told that they get a sample of a disposable razor (high involvement) or a sample of toothpaste (low involvement) after finishing the questionnaire. Involvement was also manipulated by claiming that the product would be available in their town (high involvement) or on the other side of the country (low involvement). They found that under both high- and low involvement, physical attractiveness of the model has a positive impact on product, brand and attitude recall. According to Kahle and Homer (1985), a probable explanation for the lack of difference between high and low involvement situations, found in their research, invokes the matchup hypothesis and the social adaptation theory that physical attractiveness can also be a source of information. The match-up hypothesis suggests that endorsers are more effective when there is a "fit" between the endorser and the endorsed product (Till and Busler, 2000). One other explanation I want to address in my research is that Kahle and Homer (1985) did show advertisements of both toothpaste and disposable razors but they only asked questions about the disposable razors in their questionnaire. So, they manipulated involvement only by changing the free sample participants get or where the product can be bought. Participants can not indicate if they are involved with the product or not. In this thesis, involvement will be used as a moderator, so there will not be made a difference in high and low involvement products, but one product will be rated by consumers as high or low involvement, based on several questions. In this experiment, these questions will be used to see if the rating of a product as high or low involvement has an influence on the product evaluations of that product.

(13)

7

advertisement. In this thesis, attractiveness-irrelevant products will be compared with attractiveness-relevant products. Also the role of involvement and the nature of a product on the interaction effect of model attractiveness and product evaluations on product evaluations will be investigated for both product types, these moderators will be described later on in this section.

Till and Busler (2000) also did research in the topic of model attractiveness. In their research they did two experiments based on both physical attractiveness of the model and expertise of the model. The first study examines the extent to which physical attractiveness serves as a match-up factor and its impact on brand attitude and purchase intentions. A match-up factor is a factor, in this case physical appearance and in study two model expertise, which is important for the fit between the model and the product. For their first research they used an attractiveness relevant product (men’s cologne) and an attractiveness irrelevant product (pen). They found that for both products, using a highly attractive model led to better product evaluations. But they did not find an interaction effect between model fit and the product, so it did not matter if the model fitted with the product. Study two considers the effectiveness of expertise as a match-up factor.

In their second study they found a match-up effect based on expertise. This suggests that expertise, instead of attractiveness, is more appropriate for matching products with models. Model expertise was in this study manipulated by using an advertisement of a product (energy bar) with an athlete as model (which fits well with the product, because athletes use energy bars) and an advertisement of a product (energy bar) with an actor as model (does not fit well with the product). Till and Bustler (2000) found that product evaluations for the energy bar were better when using the athlete as a model. However, for attractiveness-relevant products, highly attractive models can be seen as experts because the main goal of using attractiveness relevant products is to become more beautiful. So, in this case the attractiveness of the model is also her expertise.

(14)

8

other studies already researched this subject (Kahle and Homer, 1985; Bower and Landreth, 2001; Häfner and Trampe, 2009).

Based on the findings above, the following hypotheses can be made:

H1: Advertisements with highly attractive models will lead to more positive product evaluations compared to advertisements with normally attractive models when the product is attractiveness- relevant.

H2: There will be no difference in product evaluations for advertisements with highly attractive models and advertisements with normally attractive models when the product is attractiveness- irrelevant.

2.2 Involvement, nature of the product and the match-up hypothesis

2.2.1 Involvement

(15)

9

Contrary, research by Häfner and Trampe (2009) state that to get positive product evaluations, round, normally attractive models are more effective in high involvement situations and thin, highly attractive models are more effective for low involvement situations. They found this statement based on the reflective impulsive model (RIM). This model suggests that there are two ways of processing namely a reflective (explicit) way and an impulsive way. The reflective way is a more thoughtful way of processing information and can be seen as a high involvement process, while the implicit way is a more spontaneous way of processing information and can be seen as a low involvement process. This is also explained in the research of Petty and Cacioppo (1983) in which they distinguish between the central route of processing (explicit) and the peripheral route of processing (implicit).

In their research, Häfner and Trampe (2009) found that when people process information in a reflective way, normally attractive models are more effective than highly attractive models. For the explicit way of processing information they found the opposite. The likeability of the normally attractive models is an important factor on this effect. People’s first reaction is to like highly attractive models more than normally attractive models. But normally attractive models are in general more liked by consumers when they have or take more time to evaluate the advertisement, this leads to positive product evaluations. In their research, Häfner and Trampe (2009) only use one product (deodorant).

(16)

10

attractiveness-relevant and attractiveness-irrelevant products will be used, to make a more clear difference between product types and which model type fits best with these products. For both product types, the role of involvement will be researched by letting participants rate a product as high or low involvement based on several questions.

This explanation, that the physical attractiveness of the model can also be a product argument, is also supported by the study of Shavitt et al. (1994). They found that endorser attractiveness can be persuasive under both high and low involvement situations. In their research they used an advertisement for a restaurant. They found that when aspects like taste and aroma were important, the attractiveness of the endorser influences the evaluations about the restaurant under low but not under high involvement conditions. But when aspects about the image of the restaurant were important, the attractiveness of the endorser influences the evaluations of the restaurant under high but not under low involvement conditions. This indicates that under high involvement conditions the fit between the model and the product has a positive influence on product evaluations. The image of a restaurant can be seen as an attractiveness-relevant thing, because an image indicates how people are seen by others. The image of a restaurant however can also serve other purposes instead of attractiveness-relevant ones, for example an indicator about how the food tastes or an impression about the atmosphere etcetera. That is why in this thesis only cosmetic products will be used as attractiveness-relevant products. This is also suggested by Till and Busler (2000) in their discussion section.

(17)

11

evaluations tended to be based on the product arguments rather than on affect. Connecting this research to the thesis, highly attractive, thin models can be seen as stereotypes for attractiveness-relevant products, they match with the product. Contrary, normally attractive models are discrepant from expectations, because they do not match with the current beauty ideal and so they do not match with attractiveness-relevant products.

Just as Till and Busler (2000) did, also Bower and Landreth (2001) found that instead of model attractiveness, mostly model expertise is important for positive product evaluations. They mention the fit between a model and the product as source congruity, which is the same concept as the match-up hypothesis. In their research they focus, as recommended by Till and Busler (2000), on attractiveness-relevant products, products associated with appearance (cosmetic products). In this research, they distinguish between beauty-enhancing and problem-solving products, to find out if this explains the mixed-results in previous researches about the interaction effect of model attractiveness and attractiveness-relevant products. There is a difference between problem solving products and beauty enhancing products. Problem solving products are used to solve problems related to beauty and appearance, for example skin problems like acne and wrinkles. Beauty enhancing products are used to enhance appearance, to make someone look more beautiful, for example mascara and lipstick.

(18)

12

irrelevant) will be rated by the participants as high involvement or low involvement, because participants can see one product in a different manner.

But as found in previous research (Kahle and Homer, 1985; Kamins, 1990; Till and Busler, 2000) highly attractive models can not only be seen as stereotypes for attractiveness-relevant products (under low involvement situations), but also as product arguments (under high involvement situations). Looking at the research of Bower and Landreth (2001), it might be possible that the lack of difference on product evaluations for problem solving products between normally and highly attractive models they found can be explained by the role of involvement. The problem solving products they used are attractiveness-relevant. So, in high involvement situations, consumers look at product arguments instead of simple cues. Because the products that are used are attractiveness-relevant, the highly attractive model can also be seen as a product argument (Shavitt et al. 1994). While highly attractive models can only be seen as a product argument for attractiveness-relevant products in high involvement situations (Kahle and Homer, 1985; Kamins, 1990; Till and Busler, 2000), this must be different for attractiveness-irrelevant products.

Based on these findings the following hypothesis can be made:

H3: When people score low on involvement, highly attractive models always lead to higher product evaluations compared to normally attractive models, irrespective of attractiveness relevance of the product. When people score high on involvement, highly attractive models lead to higher product evaluations compared to normally attractive models, when the product is attractiveness-relevant and normally attractive models lead to higher product evaluations compared to highly attractive models, when the product is attractiveness-irrelevant.

2.2.2 Nature of the product

(19)

13

be seen as hedonic products. Utilitarian product benefits refer to the functional and practical benefits of consumption and hedonics refers to their aesthetic, experiential, and enjoyment-related benefits (Alex and Joseph, 2012). Also on this variable, products can be seen by participants in a different manner, or can have both hedonic as utilitarian values (Batra and Ahtola, 1990).

There is a difference in how advertisements for hedonic or utilitarian products are being viewed. The hedonic ad appeal involves building a personality for the product, from which consumers may project images, such as the values and lifestyles of product users. In clear contrast, the utilitarian ad appeal explicitly addresses product attributes or focuses on product quality claims (Chang, 2004). For attractiveness-relevant products, the beauty of the advertised model can be seen as a product argument in which the perceived beauty of the model has a fit with the product and so can be seen as a product quality claim like “this model is very beautiful so when she is using this product, I have to use it too to become just as beautiful as she is”. This is not true for attractiveness-irrelevant products because beauty has nothing to do with the product. In the context of this thesis we expect that the use of highly attractive models leads to better product evaluations for hedonic products (both attractiveness-relevant as attractiveness-irrelevant products). For utilitarian products, we expect that using highly attractive models lead to better product evaluations for attractiveness-relevant products and that using normally attractive models leads to better product evaluations for attractiveness-irrelevant products.

Based on these findings the following hypothesis can be made:

H4: When a product is hedonic, highly attractive models always lead to higher product

(20)

14

2.3 Conceptual model

(21)

15

3. Research Methodology

In this section, the method of research will be explained. The goal of this thesis is to find out if there is an interaction effect of model attractiveness and product type on product evaluations and how involvement and nature of the product moderate the effect of this interaction.

To do this, the following research question will be answered:

What is the interaction effect of model attractiveness and product type on product evaluations? And how do involvement and the nature of a product moderate the effect of this interaction?

3.1 Pre-test

Before doing this research, a pre-test has been done to find out if the models I think are attractive were also perceived as attractive by the participants of the experiment and if the products I think are attractiveness relevant were also perceived as attractiveness-relevant by the participants of the experiment. The goal of this pre-test is to select appropriate models and products for the actual research. This has been done by showing the selected advertisements to a sample of 25 people within the target group; this is enough to see the sample as representative for the target group. The target group existed of women of sixteen years and older and were selected by randomly choosing women in my environment to participate in the pre-test by showing them the questionnaire. The participants in the pre-test were not participating in the main experiment. The participants, 25 women of sixteen years and older, had to answer questions about the models and the products on a 7 point Likert-scale (appendix 1.1 and 1.2).

(22)

16

Model 1 (M = 6.46), 3 (M = 6.58) and 10 (M = 6.42) have the highest scores, so they can all be used in the main experiment (appendix 1.3). First, their means are compared to each other and none of them is significantly different, model 1 and 3 (t (24) = -1.953, p= 0.063), model 1 and 10 (t (24) = 0.235, p= 0.816) and model 3 and 10 (t (24) = 0.937, p= 0.358). This means that all three models can be used in the experiment. But, model 10 is recognized by fewer participants, so this one will be used in the experiment to make sure this will not influence the results of the experiment.

Model 2 (M = 4.01), 6 (M = 4.85) and 9 (M = 4.19) have the most average scores, so they can be used in the main experiment (appendix 1.3). Also their means are compared to each other and the mean of model 6 is significantly different than the means of model 2 (t (24) = -2.864, p= 0.009) and 9 (t (24) = 2.789, p= 0.010). So, model 6 cannot be used in the main experiment. Between model 2 and 9, no significant difference is found (t (24) = -0.685, p= 0.500), so they can both be used in the experiment. Now, the means of model 2 and 9 are compared with the mean of model 10, to see if the means are significantly different. This is true for both model 2 (t (24) = 8.034, p= 0.000) as model 9 (t (24) = 7.451, p= 0.000), so they can both be used in the main experiment. Model 2 is eventually chosen because this advertisement is most similar to the advertisement of model 10 (same pose, hair colour, picture quality).

(23)

17

Also 10 products have been chosen, 5 products which are, in my opinion, attractiveness-relevant and 5 products that are attractiveness-irrelevant. As recommended by Till and Bustler (2000) the attractiveness-relevant products are all cosmetic products. The attractiveness-irrelevant products have nothing to do with status, as recommended by Shavitt and Swan (2001). After answering the questions the products with the highest scores on attractiveness-relevance (perfume,

foundation, nail polish and shampoo) and the products with the lowest scores on attractiveness- relevance (toilet paper, DVD-player, microwave, coffee maker) were compared with each other in paired sample t-tests to see if there means are significantly different. This was true for all of the pairs.

There is also looked at the results of the other questions because it is desired that the 2 products which are used in the main experiment have similar results on these questions, so these factors don’t have an influence on the outcomes of the main experiment. So, after comparing the

products with each other the product perfume is chosen as attractiveness-relevant product and the product coffee-maker is chosen as the attractiveness-irrelevant product. This is done because these 2 products do not differ much in price and the differences on the factors hedonic/ utilitarian and usage are smaller than for most of the products.

(24)

18

3.2 Main experiment

The main experiment existed of 100 people which filled in a questionnaire. The study consisted of a 2 (model condition: highly attractive/normally attractive) x 2 (product type: attractiveness-relevant/ attractiveness-irrelevant) design (fig 2.) The participants were randomly assigned into one of the four conditions; so each of these conditions existed of approximately 25 participants. Once participants were assigned to one of the four conditions; a short introduction to explain the experiment was given, here was explained that the participants would view an advertisement and that several questions were asked about this advertisement; which they had to look at like they normally did look at an advertisement. After this, one of the four advertisements was shown to the participant. For example, a participant in condition one of the four conditions, was shown an advertisement for an attractiveness-relevant product with a highly attractive model in it. Finally, the questionnaire with questions about the product, the advertisement and the purchase intentions was shown to the participants. In this questionnaire also questions about involvement and the nature of the product were asked, so these factors can be used as moderators in the experiment. The advertisements can be found in appendix 2.1 and the questionnaire can be found in appendix 2.2.

3.3 Independent variables

3.3.1 Product type

The chosen attractiveness-relevant product is a cosmetic product as suggested by Till and Busler (2000), namely perfume. Also advertisements with an attractiveness-irrelevant product have been shown to the participants, namely a coffeemaker. For this product category a product which have nothing to do with status/attractiveness is used, this was also suggested by Shavitt and Swan (2001), to make sure that this would not influence the outcomes. The products which are used were selected based on the pre-test described in chapter 3.1.

3.3.2 Model attractiveness

(25)

19

for perfume “the new smell sensation” and for the coffeemaker “the new taste sensation”. The used advertisements can be found in appendix 2.1.

Involvement is used as a moderator in this experiment. All participants have answered five questions about involvement in the questionnaire. For the questions about involvement a factor analyses and a reliability test are done, to see if these questions can be combined as one factor. These outcomes of the factor analysis are shown in appendix 2.3. The factor analysis shows that all four questions load over .40 on the factor, which means that all four questions can be

combined as one factor called involvement (Floyd and Widaman, 1995). Cronbach’s Alpha is 0.750, which means that this test also shows that the four questions can be combined as one factor called “involvement”.

So, there are not used different high and low involvement products, because participants can see one product in a different manner. So, one product can be rated as high involvement by one participant and as low involvement by another participant.

3.3.4 Nature of the product

Also the nature of the product is used as a moderator in this experiment. Participants in the experiment have answered if they see the product as hedonic or as utilitarian on a dichotomous scale. (I use this product for fun/ I use this product because I need it.) Products can also be rated as both utilitarian and hedonic.

3.4 Dependent variable

The dependent variable in this research was the product evaluation. This was measured by asking participants questions about their purchase intentions (f.e. “How likely is it that you would buy the product instead of another similar one?”) and the product (f.e. “what is your feeling about the product?”). By using the questions about involvement and hedonic/utilitarian as moderators, the influence of these factors on product evaluations will also be measured in the experiment. These questions were rated by the participants on a 7-point Likert-Scale ranging. The complete

questionnaire can be seen in appendix 2.2.

(26)

20

one factor, called product evaluations. In appendix 2.4, the outcomes of the factor analysis are shown. Factor analysis shows that all five questions load over .40 on the factor, which means that all five questions can be combined as one factor called product evaluations (Floyd and Widaman, 1995). Also, the Cronbach’s Alpha of 0.874 (which is higher than 0.6), shows that the five questions can be combined together as one factor called “product evaluations”.

Based on the hypotheses mentioned in chapter 2, the following outcomes were expected:

Product type

Model

condition

Attractiveness-relevant product Attractiveness-irrelevant product High involve ment Low involve ment Utilitar ian Hedonic High involv ement Low involv ement Utilita rian Hedonic Highly attractiv e model + + + + - + - + Normall y attractiv e model - - - - + - + -

(27)

21

4. Results

4.1 Descriptives

The main experiment existed of 100 female participants in the age of sixteen years and older. These women participated in the study by filling in a questionnaire with one of the four advertisement conditions. These advertisements can be found in appendix 2.1. Only women are used as participants, because beauty products are particularly used by women, also little research is done about male models. Participants were randomly chosen out of the target group (women of sixteen years and older) and were friends, family, co-workers or selected by randomly asking them to fill in the questionnaire in the shopping mall. All 100 participants were assigned to one of the four conditions and the answers on the questionnaire were put into a SPSS data file. Participants who saw the normal attractive model was given a 1 and participants who saw the highly attractive model was given a 2. The same was done for product type, participants who saw the relevant product was given a 1 and participants who saw the attractiveness-irrelevant product was given a 2.

4.2 The interaction effect of model attractiveness and product type on product

evaluations

After putting the questions, as described in chapter 3.3 and 3.4, into 2 factors, “product

evaluations” and “involvement”, several SPSS tests are performed to test the hypotheses based on the outcomes of the questionnaire.

First a factorial 2-way ANOVA is done to test the interaction effect of product type and model attractiveness on product evaluations. In this factorial 2-way ANOVA, the variables product type and model attractiveness were used as independent variables and the variable product evaluations is used as the dependent variable. For model condition no significant result is found (F (1, 96) = 2,499, p = 0.117). Also for product type no significant result is found (F (1, 96) = 2.700, p = 0.104. Finally, for the interaction between model attractiveness and product type also no

significant result is found (F (1, 96) = 0.688, p = 0.409). This would mean that the attractiveness of a model, product type and it’s interaction with each other are not influencing product

(28)

22

But, because there are not found significant results in the test above, the question about how participants rate the model attractiveness is reviewed and although the means of the highly attractive model and the normal attractive model are significantly different (t (98) = -3.413, p= 0.001) this question was added to the ANOVA as a covariate to see if this question has an influence on the interaction effect of model attractiveness and product type on product

evaluations. This is done because there are participants who rated the normal model as highly attractive (7 on a 7-point Likert scale) and there are also participants who rated the highly attractive model as unattractive (2 on a 7-point Likert scale). These ratings can influence the outcomes of the ANOVA.

In the next factorial 2-way ANOVA, the variables product type and model attractiveness were used as independent variables and the variable product evaluations is used as the dependent variablejust as in the test above, but now the question about the model attractiveness (question 7 in the questionnaire) is used as a covariate, to see if this question is moderating the outcomes of the previous test. For model condition a significant result is found now (F (1, 95) = 13.110, p= 0.00). Also for the covariate “rating of the model” a significant result is found (F (1, 95) = 33.552, p= 0.00). For product type no significant result is found (F (1, 95) = 2.405, p = 0.124). Also, for the interaction between model attractiveness and product type no significant result is found (F (1, 95) = 1.285, p = 0.260). This means that the attractiveness of a model and the rating of the model are influencing product evaluations. But product type and the interaction between model attractiveness and product type are not influencing product evaluations.

Because the covariate “rating of the model” has a significant influence, also a regression is performed. In this regression product type and rating of the model are used as independent variables and the variable product evaluation is used as the dependent variable. Because rating of the model is a continuous variable, first this variable is mean-centered.

(29)

23

better the product evaluations are. The effect of product type is not significant (B= -0.359, t (99) = -1.530, p= 0.129).

In the second step of the regression, the interaction between product type and rating of the model is added. To see the interaction between product type and rating of the model, a dummy variable of the variable product type is made, with attractiveness-relevant product = 1 and attractiveness-irrelevant product = 0. With the centered means of the model ratings and the dummy-variable of product type, a new interaction variable is made. This interaction variable is added into the regression as an independent variable. The second step of the regression also shows a significant effect (r²= 0.232, F (3, 96) = 9.648, p = 0.000). This regression shows, just like in step one, that rating of the model is significantly influencing product evaluations B= 0.232, t (99) = 2.112, p= 0.037. The regression also shows that the interaction between product type and rating of the model is significantly influencing product evaluations B= 0.334, t (99) = 2.026, p= 0.045. Product type is not significantly influencing product evaluations B= - 0.356, t (99) = -1.541, p= 0.127.

To see if the interaction effect of product type and rating of the model on product evaluations differs for high and low ratings of the model, a simple slope analysis is done.First, the standard deviation (SD) is added to the centered mean (for low ratings) and subtracted of the centered mean (for high ratings). So, two new variables, “high ratings of the model” and “low ratings of the model”, are made in SPSS. After doing this, the previous regression is repeated twice, once with the high rating of the model and product type as independent variables and once with the low rating of the model and product type as independent variables.

For the high model ratings, the regression shows that product type is not significantly influencing the product evaluations B= 0.046, t (99) = 2.112, p= 0.772. This means that for high model ratings, the product evaluations do not significantly differ per product type.

For the low model ratings, the regression shows that product type is significantly influencing the product evaluations B= -0.830, t (99) = -2.532, p= 0.013. This means that for low model ratings, the product evaluations are higher for attractiveness-irrelevant products compared to

(30)

24

To see if the interaction effect of product type and rating of the model on product evaluations differs for attractiveness-relevant and attractiveness-irrelevant products, a simple effect analysis is done. This analysis shows that for attractiveness-relevant products, product evaluations are better when using a highly attractive model (F (1, 96) = 7.484, p = 0.007). For attractiveness-irrelevant products also highly attractive models lead to better product evaluations (F (1, 96) = 5.349, p = 0.023).

So, we can say that the use of highly attractive models is better for product evaluations for both attractiveness-relevant products and attractiveness-irrelevant products, but when using low attractive models product evaluations are better for attractiveness-irrelevant products. In the next section some moderators are researched which can influence the interaction between model attractiveness and product type on product evaluations.

4.3 The effect of the moderators on the interaction effect of model attractiveness

and product type on product evaluations

After researching the interaction effect of model attractiveness and product type on product evaluations, now the influence of the moderators, mentioned in chapter 3.3, on this interaction are investigated.

4.3.1.1 3-way ANOVA with model condition

First, a factorial 3-way ANOVA is performed to see if the moderator involvement has a significant influence on the interaction effect of model attractiveness and product type on

(31)

25

The ANOVA shows that all variables and interactions, product type (F (1, 92) = 2.678, p = 0.105). model condition (F (1, 92) = 2.291, p = 0.134), involvement(F (1, 92) = 0.207, p = 0.650), the interaction between product type and involvement (F (1, 92) = 0.545, p = 0.462), the interaction between product type and model condition(F (1, 92) = 0.642, p = 0.425), the

interaction between involvement and model condition(F (1, 92) = 0.091, p = 0.763) and the interaction between involvement, model condition and product type (F (1, 92) = 0.110, p = 0.741), have no significant effect on product evaluations.

4.3.1.2 3-way ANOVA with rating of the model

Second, a factorial 3-way ANOVA with the independent variable rating of the model is done. Because involvement and rating of the model are continuous variables, a split-median is made for both variables. Low involved participants are rated a one and high involved participants are rated a two. Also, participants who rated the model as not very attractive are rated a one and participants who rated the model as highly attractive are rated a two. After doing the median splits, both variables are inserted, together with product type, as independent variables in the ANOVA, in which product evaluations is entered as the dependent variable.

The ANOVA shows that rating of the model has a significant effect on product evaluations (F (1, 92) = 11.495, p = 0.001). This means that products are evaluated more positively for high ratings of the model (M = 4.992, SD =1.2596) than for low ratings of the model (M = 4.071, SD

=1.1750). The ANOVA shows that the other variables and interactions, product type (F (1, 92) = 1.581, p = 0.212), involvement(F (1, 92) = 0.016, p = 0.899), the interaction between product type and involvement (F (1, 92) = 1.113, p = 0.294), the interaction between product type and rating of the model(F (1, 92) = 0.009, p = 0.926), the interaction between involvement and rating of the model(F (1, 92) = 0.709, p = 0.402) and the interaction between involvement, rating of the model and product type (F (1, 92) = 0.229, p = 0.633), have no significant effect on product evaluations.

(32)

26

4.3.2 Hedonic nature of the product

First, a factorial 3-way ANOVA is performed to see if the moderator hedonic nature of the product has a significant influence on the interaction effect of model attractiveness and product type on product evaluations. Because the thesis started with using the model condition (normal attractive ft. highly attractive model) as independent variable instead of using the question about model ratings, first an ANOVA with this variable is done. In this ANOVA product type, model condition and the split-median of hedonic nature of the product are used as independent variables and product evaluations is used as dependent variable. For the split-median of hedonic nature of the product, participants who rate the product as low hedonic are rated a one and participants who rate the product as high hedonic are rated a two. The ANOVA shows that the interaction between product type and hedonic nature of the product has a significant effect on product evaluations (F (1, 92) = 13.109, p = 0.000). The ANOVA shows that the other variables and interactions, model condition (F (1, 92) = 2.988, p = 0.087), hedonic nature of the product (F (1, 92) = 2.810, p = 0.097), product type (F (1, 92) = 2.615, p = 0.109), the interaction between product type and model condition(F (1, 92) = 1.350, p = 0.248), the interaction between hedonic nature of the product and model condition(F (1, 92) = 0.560, p = 0.456) and the interaction between hedonic nature of the product, model condition and product type (F (1, 92) = 2.870, p = 0.094), have no significant effect on product evaluations.

(33)

27

ratings of the model (M =4.992 SD =1.2596), than for low ratings of the model (M =4.071 SD =1.1750). The ANOVA shows that the other variables and interactions, product type (F (1, 92) = 1.205, p = 0.275), hedonic nature of the product(F (1, 92) = 1.135, p = 0.289), the interaction between product type and rating of the model(F (1, 92) = 0.111, p = 0.740), the interaction between hedonic nature of the product and rating of the model(F (1, 92) = 0.068, p = 0.795) and the interaction between hedonic nature of the product, rating of the model and product type (F (1, 92) = 1.431, p = 0.235), have no significant effect on product evaluations.

Both outcomes show that the interaction between hedonic nature of the product and product type has a significant influence on product evaluations. To understand how this interaction looks like simple effect analyses are done. First, an analysis is done to see the effect of hedonic nature of the product within attractiveness-relevant and attractiveness-irrelevant product types. This analysis shows that for attractiveness-relevant products, product evaluations are significantly different (F (1, 96) = 15.272, p = 0.00). This means that product evaluations are better when participants rate the product as high hedonic (M =4.968, SD =1.1412) instead of low hedonic (M = 3.656, SD =1.1882). For attractiveness-irrelevant products product evaluations do not

significantly differ (F (1, 96) = 3.696, p = 0.057). Second, an analysis is done to see the effect of product type for low and high ratings on hedonic nature of the product. For low ratings of hedonic, product evaluations significantly differ (F (1, 96) = 17.263, p = 0.000). This means that product evaluations are better when using an attractiveness-irrelevant product (M =4.964, SD =1.0252) instead of an attractiveness-relevant product (M =3.656 SD =1.1882). For high ratings of hedonic, product evaluations do not significantly differ (F (1, 96) = 3.376, p = 0.069). But the outcomes show that hedonic nature of the product has no effect on the interaction effect of product type and model attractiveness on product evaluations.

4.3.3 Utilitarian nature of the product

(34)

28

variable, a split-median is made for this variable. Participants who rate the product as low

utilitarian are rated a one and participants who rate the product as high utilitarian are rated a two. Because the thesis started with using the model condition (normal attractive ft. highly attractive model) as independent variable instead of using the question about model ratings, first an ANOVA with this variable is done. In this ANOVA product type, model condition and the split-median of utilitarian nature of the product are used as independent variables and product

evaluations is used as dependent variable. The ANOVA shows that the interaction between product type, model condition and utilitarian nature of the product has a significant effect on product evaluations (F (1, 92) = 6.630, p = 0.012). The ANOVA shows that the other variables and interactions, model condition (F (1, 92) = 0.441, p = 0.508), utilitarian nature of the product (F (1, 92) = 1.101, p = 0.297), product type (F (1, 92) = 0.185, p = 0.668), the interaction

between product type and model condition(F (1, 92) = 1.305, p = 0.256), the interaction between utilitarian nature of the product and model condition(F (1, 92) = 0.123, p = 0.727) and the interaction between utilitarian nature of the product and product type (F (1, 92) = 2.449, p = 0.121), have no significant effect on product evaluations.

(35)

29

between utilitarian nature of the product and rating of the model(F (1, 92) = 0.721, p = 0.398) and the interaction between utilitarian nature of the product, rating of the model and product type (F (1, 92) = 0.000, p = 1.000), have no significant effect on product evaluations.

The outcomes of these ANOVA’S show contradicting results. When including rating of the model as independent variable no significant effects of utilitarian nature of the product were found. But, when looking at the other ANOVA, which included model condition as independent variable, it is found that the interaction between utilitarian nature of the product, model condition and product type has a significant influence on product evaluations. This would mean that

utilitarian nature of the product has an influence on the interaction effect of model condition and product type on product evaluations. To understand how this interaction looks like, simple effect analyses are done.

First, a split file for utilitarian nature of the product is made to compare high and low ratings on utilitarian nature of the product with each other. Then, an analysis is done to see the effect of model condition within attractiveness-relevant and attractiveness-irrelevant product types on product evaluations. This effect is measured for both low and high utilitarian ratings. The outcomes of the simple effect analysis show that only for high utilitarian ratings and the use of an attractiveness-irrelevant product, model condition has a significant effect on product

evaluations (F (1, 39) = 8.564, p = 0.006). This means that for high utilitarian rated/ attractiveness-irrelevant products, product evaluations are better when using a normally attractive model (M =5.200 SD =0.9452) instead of a highly attractive model (M =3.891 SD =1.7050). For low utilitarian rated/ attractiveness-relevant products (F (1, 53) = 1.320, p = 0.256), low utilitarian rated/attractiveness-irrelevant products (F (1, 53) = 0.246, p = 0.622), and high utilitarian rated/ attractiveness-relevant products (F (1, 39) = 1.295, p = 0.262), no

significant effects of model condition are found.

Finally, a simple effect analysis is done to see the effect of product type for highly and normally attractive models on product evaluations. This effect is also measured for both low and high utilitarian ratings. The outcomes of the simple effect analysis show that for low utilitarian ratings and the use of a highly attractive model, product type has a significant effect on product

(36)

30

=4.843 SD =0.9419) compared to attractiveness-relevant products (M =3.895 SD =1.5324). For low utilitarian rating and the use of a normally attractive model, product type has no significant effect on product evaluations (F (1, 53) = 0.067, p = 0.797).

(37)

31

5. Discussion

5.1 Conclusion

At the beginning of this thesis, the question “does beauty always sell” was asked. Previous research done in this field, showed contradicting results. Some articles found that the more beautiful the source, the more positive the product evaluations (Kahle and Homer, 1985). Other articles found that instead of model attractiveness, other characteristics of the model are more important (Bower and Landreth, 2001). In this thesis, the question “does beauty always sell?” is answered by investigating the effect of model attractiveness on product evaluations, based on different product types (an relevant product (perfume) and an attractiveness-irrelevant product (coffeemaker)). We expected that highly attractive models, only lead to higher product evaluations for attractiveness-relevant products. The effect of the interaction between model attractiveness and product on product evaluations is researched. Also, the moderating influence of involvement, the personal significance that individuals ascribe to a product, and nature of the product (hedonic or utilitarian) on this interaction effect are studied. To investigate this, the research question “What is the interaction effect of model attractiveness and product type on product evaluations? And how do involvement and nature of the product moderate this effect” is used.

(38)

32

for attractiveness-irrelevant products are significantly higher than for attractiveness-relevant products. Linking these findings to the match-up hypothesis, we can say that highly attractive models fit with attractiveness-relevant products. Here the beauty of the model fits with the goal of the relevant product (make someone more beautiful). For attractiveness-irrelevant products, the goal of the product does not fit with the beauty of a model. This can explain why for low model attractiveness, product evaluations for attractiveness-irrelevant products are better compared to attractiveness-relevant products.

So, in general it is better to use highly attractive models. For attractiveness-relevant products it is always better to use highly attractive models but for attractiveness-irrelevant products the

advantage of using highly attractive models is less than for attractiveness-relevant products. This can be explained by the fact that for attractiveness-relevant products, model attractiveness can also be seen as a product argument. So, when model attractiveness is low, for attractiveness-relevant products the model does not fit with the product and this will lead to lower product evaluations. For attractiveness-irrelevant products, model attractiveness is not a product argument. So, when the model attractiveness is low, for attractiveness-irrelevant products the influence on product evaluations is lower. Further research has to investigate this broader by, for example, comparing highly and normally attractive models with advertisements without models in it. While previous research shows that the use of highly attractive models can have a negative influence on women’s self-esteem (Richins, 1991), it could be interesting to find out if

advertisements without a model in it can be equally or more effective compared to

advertisements with models in it. So, it could be interesting to compare advertisements with models in it, to advertisements with text and other images in it, for example with slogans and claims about the quality of the product.

(39)

33

For nature of the product, two separate variables are researched, hedonic nature of the product and utilitarian nature of the product. We expected that for hedonic products, highly attractive models led to better product evaluations for both product types. Hedonic nature of the product has no significant influence on the interaction effect of model attractiveness and product type on product evaluations. But, it can be said that product evaluations for attractiveness-relevant products are higher when they are seen by participants as high hedonic. When participants see a product as not/low hedonic, product evaluations are better for attractiveness-irrelevant products. Model attractiveness has no influence on this effect.

For utilitarian nature of the product we expected that normally attractive models are better for irrelevant products and highly attractive models are better for attractiveness-relevant products. The outcomes show a significant influence on the interaction effect of model attractiveness and product type on product evaluations. When participants see an attractiveness-irrelevant product as high utilitarian, product evaluations are better when using a normally attractive model compared to a highly attractive model. When participants see an attractiveness-irrelevant product as low utilitarian, product evaluations are better when using a highly attractive model compared to a normally attractive model. For attractiveness-relevant products with a highly attractive advertising model, product evaluations are better when participants see the product as high utilitarian. This can be explained by the fact that for high utilitarian products people are persuaded by product arguments (Alex and Joseph, 2012). For attractiveness-relevant products, the beauty of the advertised model can be seen as a product argument in which the perceived beauty of the model has a fit with the product and so can be seen as a product argument like “this model is very beautiful so when she is using this product, I have to use it too to become just as beautiful as she is”. This is not true for attractiveness-irrelevant products because beauty has nothing to do with the product.

5.2 Limitations and future research

(40)

34

different, the highly attractive model was seen as more attractive than the normally attractive model, the interaction effect of model attractiveness and product type on product evaluations was not significant. However, in the questionnaire also a question about model attractiveness was asked and this question showed a significant interaction effect with product type on product evaluations. That is why in the rest of the thesis also the continuous variable rating of the model attractiveness is used.

In the research is found that involvement has no influence on the interaction effect of model attractiveness and product type on product evaluations. In this thesis, involvement is measured based on four questions asked in the questionnaire. For further research it is important to measure involvement in a different manner. For example by doing an experiment with one low involvement condition and one high involvement condition, in this manner participants are not aware of the role that involvement plays in the research and a causal relationship can be investigated. One example of doing such an experiment is by dividing participants into two groups, with one condition where a new product is launched in the US and one condition where a new product is launched in the Netherlands and where participants get a free sample if they participate in the experiment.

(41)

35

Further, it is important to look at other moderators which can be influencing the interaction effect of model attractiveness and product type on product evaluations. In this study not only students are used, but the majority of the participants were students between 18 and 25. In future research it can be very useful to also look at the influence of age, social class, origin etcetera of the participants. For example, research by Drolet, Williams and Lau-Gesk (2007) found, in a research about hedonic and utilitarian product category types, that elderly consumers (age 65 plus) had more favorable attitudes toward affective (vs. rational) advertisements, regardless of product category type. In contrast, young adult consumers (age 18–25) favored affective ads only for hedonic products. They favored rational ads for utilitarian products. These findings can play a role in the further investigation of the moderating role of nature of the product. Also, Antioco, Smeesters and le Boedec (2012) found that age and also place of residence (rural or urban) can moderate the effect of model attractiveness on product evaluations. They found that participants between 35 and 44, living in urban areas, have the highest self-esteem and so react more positive on highly attractive models than other target groups do. Therefore, it is important to look at different variables which can have an influence on the outcomes of further research. For example by investigating the variable age by making two groups (younger people and older people) and study the differences between those groups. This can also be done for other moderators like social class, living situation etc.

(42)

36

For now, when starting a new advertisement campaign, it is recommended to always use highly attractive models for attractiveness-relevant products. For attractiveness-irrelevant products it depends on the nature of the product whether it is better to use a normally or a highly attractive model. For utilitarian/attractiveness-irrelevant products, it is better to use a normally attractive model compare to a highly attractive model, but when attractiveness-irrelevant products are seen as low utilitarian, it is better to use a highly attractive model compared to a normally attractive model. There are also other moderators which can influence the outcomes of this research and can have implications on what kind of models are best to use. So, investigating the effect of not using a model at all and further investigating the effect of using normally attractive models (by researching other moderators and the used moderators more extensively) can lead to lowed advertisement costs and better product evaluations. Also important; it is good for women’s self-esteem and satisfaction about themselves.

(43)

37

6. References

Alex, Joji N. and Ashwin Joseph (2012), “Hedonic Versus Utilitarian Values: The Relative Importance of Real and Ideal Self to Brand Personality and Its Influence on Emotional Brand Attachment.” XIMB Journal of Management, Sep., 77-90.

Antioco, M, Dirk Smeesters and Aline le Boedec (2012), “Take Your Pick: Kate Moss or the Girl Next Door? The Effectiveness of Cosmetics Advertising.” Journal of Advertising Research. March., 15-30.

Batra, R. and Ahtola, O.T. (1990), “Measuring the Hedonic and Utilitarian Sources of Consumer Attitudes,” Marketing Letters 2 (April): 159-170.

Bower, Amanda B, and Stacy Landreth (2001), "Is Beauty Best? Highly Versus Normally Attractive Models in Advertising." Journal of Advertising. 1 (Spring), 1-12.

Chang, Chingching (2004), “How Mood and Ad-self-congruency Affect the Relative Influence of Hedonic Ad Appeals and Utilitarian Ad Appeals on Product Evaluations.” Advances in Consumer Research. 31, 721-727.

Drolet, Aimee, Patti Williams and Loraine Lau-Gesk (2007), “Age-related differences in responses to affective vs. rational ads for hedonic vs. utilitarian products.” Market Lett, 18, 211-221.

Fennis, Bob M, and Wolfgang Stroebe (2010), The Psychology of Advertising, 1st ed., Hove: Psychology Press.

(44)

38

Häfner, Michael, and Debra Trampe (2009), "When thinking is beneficial and when it is not: The effects of thin and round advertising models." Journal of Consumer Psychology, 19, 619–628.

Halliwell, Emma, and Helga Dittmar (2004), “Does size matter? The impact of model’s body size on womens’s body-focused anxiety and advertising effectiveness.” Journal of Social and Clinical Psychology, 23 (1), 104-122.

Homer, Pamela M, and Lynn R. Kahle (1986), “A Social Adaptation Explanation of the Effects of Surrealism on Advertising.” Journal of Advertising, 15 (2), 50-60.

Kahle, Lynn R, and Pamela M. Homer (1985), "Physical Attractiveness of the Celebrity Endorser: A Social Adaptation Perspective." Journal of Consumer Research 11 (March), 954-961.

Kamins, Michael A. (1990), “An Investigation into the "Match-Up" Hypothesis in Celebrity Advertising: When Beauty May be Only Skin Deep.” Journal of Advertising, 19 (1), 4-13.

Lynch, James, and Drue Schuler (1994), “The Matchup Effect of Spokesperson and Product Congruency: A Schema Theory Interpretation.” Psychology & Marketing, 11 (5), 417-445.

Ohanian, R. (1990), “Constructions and Validation of a Scale to Measure Celebrity Endorsers’ Perceived Expertise, Trustworthiness, and Attractiveness.” Journal of Advertising, 19 (3), 39-52.

Peck, Joann, and Barbara Loken (2004), “When Will Larger-Sized Female Models in

Master thesis BA Marketing