• No results found

The nature of the relationship between product gender perceptions and brand gender identity : Investigating the effect of a discrepancy between the gender of the product and the gender of the brand.

N/A
N/A
Protected

Academic year: 2021

Share "The nature of the relationship between product gender perceptions and brand gender identity : Investigating the effect of a discrepancy between the gender of the product and the gender of the brand."

Copied!
101
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Hannako Bakker - 10456139 Master Thesis - Final version 29 September 2014

Business Administration - Marketing track

Amsterdam Business School - University of Amsterdam

First supervisor: Prof. drs. ing. A.C. J. Meulemans Second supervisor: Prof. dr. J.H.J.P. Tettero

The nature of the relationship between

product gender perceptions and brand

gender identity

Investigating the effect of a discrepancy between

the gender of the product and the gender of the

(2)

2

Statement of Originality

This document is written by Hannako Bakker, who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in

creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Contents

1 Introduction ...8

1.1 Brand gender and product gender ...9

1.2 Problem statement ...9

1.3 Theoretical contribution ... 10

1.4 Managerial contribution ... 10

1.5 Research outline ... 11

2 Literature review ... 12

2.1 Differentiating between sex and gender ... 12

2.2 Gender identity ... 13

2.3 Theory on gender identity ... 14

2.3.1 Gender schema theory ... 15

2.3.2 Multifactorial gender identity theory ... 16

2.4 Gender identity in this study ... 17

2.5 Gender identity in consumer behavior and marketing research ... 17

2.6 Self-Congruity theory ... 18

2.7 Gender identity and the self concept ... 20

2.8 Brand personality and gender ... 21

2.8.1 Brand gender identity ... 22

2.9 Product personality and gender ... 23

2.9.1 Product gender perceptions ... 24

2.10 The relationship between brand and product associations ... 26

2.11 Brand gender identity and product gender perceptions ... 27

3 Conceptual Framework ... 29

4 Methodology ... 32

4.1 Research design ... 32

4.2 Required stimuli for treatment conditions ... 32

4.2.1 Product category ... 33 4.2.2 Brand ... 36 4.3 Measurement instrument ... 37 4.3.1 Independent variable ... 37 4.3.2 Interaction variables ... 39 4.3.3 Dependent variables ... 41 4.4 Demographics ... 42

(4)

4

4.5 Translation of the questionnaire ... 42

4.6 Sampling technique and procedure ... 42

4.7 Data analysis ... 43

4.7.1 Differences between variables/groups ... 43

4.7.2 Relationships between variables ... 44

5 Results ... 46

5.1 Data collection and missing values ... 46

5.2 Sample description ... 47

5.3 Reliability of measurement instrument ... 47

5.4 Description of the data ... 48

5.5 Manipulation check ... 50

5.5.1 Product gender perception treatment ... 50

5.5.2 Brand gender identity ... 51

5.6 Product gender and brand gender ... 51

5.6.1 Differences in perception of the product and brand gender ... 51

5.7 Relationships between the variables ... 53

5.8 Match: Masculine product gender and masculine brand gender ... 54

5.8.1 Purchase intention ... 54

5.8.2 Brand affect ... 55

5.9 Match: Feminine product gender and feminine brand gender ... 56

5.9.1 Brand Affect ... 56

5.10 Mismatch: masculine product gender, feminine brand gender ... 58

5.10.1 Purchase intention ... 58

5.10.2 Brand affect ... 59

5.11 Feminine product gender perception, masculine brand gender identity... 60

5.11.1 Brand affect ... 60 5.12 Gender identity ... 61 5.13 Demographics ... 62 5.13.1 Sex ... 62 5.14 Hypothesis summary ... 64 6 Discussion ... 64 6.1 Gender identity ... 64 6.2 Sex ... 65

(5)

5

6.3.1 Match: masculine product gender... 67

6.3.2 Match: feminine product gender ... 68

6.3.3 Mismatch: masculine product gender ... 69

6.3.4 Mismatch: feminine product gender ... 71

6.3.5 The insignificance of product gender for brand preference ... 71

6.3.6 The chicken or the egg: what came first? ... 72

7 Conclusion ... 73

7.1 The answer ... 73

7.2 Theoretical implications... 75

7.3 Managerial implications... 75

7.4 Limitations and avenues for future research ... 76

8 Appendix ... 89

8.1 Appendix A: The questionnaire ... 89

8.2 Appendix B. Translations of the questionnaire ... 96

(6)

6

Table 1. Masculine and feminine products according to Fugate&Phillips (2010) ... 33

Table 2. Means for each product on masculinity and femininity ... 35

Table 3. Non-significant difference between products in paired samples t-test ... 35

Table 4. Four conditions, resulting in a match or a mismatch ... 37

Table 5. Critical Mahalanobis distance for # of independent variables ... 45

Table 6. Internal consistency score ( per construct ... 47

Table 7. Means, Standard Deviations, Pearson correlations and Cronbach's () ... 49

Table 8.One-way between subjects ANOVA for effect of product gender in scenario on dependent variables ... 52

Table 9. Masculine product gender perception moderates relationship brand affect ... 57

Table 10. Masculine product gender perception moderates relationship ... 60

Table 11. Feminine product gender perception moderates relationship ... 60

Table 12. Non significant differences for gender identity... 61

Table 13. Sex moderates relationships for BA and BP ... 63

Table 14. The final result for the tested hypothesis in this chapter ... 64

(7)

7

Abstract

This study examines the relationship between the perceived gender of a product and the gender identity of a brand. To research this, the effect of a discrepancy between product gender perceptions and brand gender identity on purchase intention, brand affect and brand preference is studied. Data was gathered using a online questionnaire, where respondents were randomly divided over 4 scenarios. Two of these scenarios included a discrepancy between product gender perceptions and brand gender perceptions and two scenarios matched the product gender and brand gender. Based on statistical analysis of the data, hypotheses were tested. The study revealed that in case of a mismatch, brand gender identity will predict the product gender perceptions for that product. A mismatch is positively related to brand affect, but for purchase intention a match between product gender and brand gender works best. Limitations of the current study and future research ideas are discussed.

(8)

8

1 Introduction

If someone were to ask you "who are you?", it is likely that in one of the first sentences you would mention your gender. Gender is an important part of the self (Digman, 1990) and as such has been a target of interest for advertisers and marketers. Companies turn their efforts towards gendering their brands, hoping to attract more consumers. What is often forgotten in these efforts, is that product categories are already associated with a typical gender. And for those of you saying this phenomenon belongs in 1950's, think again. Fugate and Philips (2010) did a meta-analysis of the gender perception of products (like cars, beer, hairspray, shampoo) and found that, compared to earlier research, there were more

products with a clear gender association. This implicated that, without considering a brand, consumers are aware of the gender of a product category. These associations stand the test of time, for example cars are still considered a masculine product, despite marketers efforts to create cars for women. Sometimes a plan to sex-change a product backfires: Alphanail produced nail polish for men (Stein, 2013) . Their marketing strategy usesd extremely masculine personality traits to describe their product, a color nail polish is "burnin' rubber" for example. Even with their efforts to create a masculine brand of nail polish, the product manages to find its' way onto a top 5 of "men's products that are clearly women's products" (Cracked, 2012) . This example highlights the difficulties when trying to market cross gender brands/products. Researchers have turned their efforts to studying the effect of cross gendered brand extensions (Lee, 2006; Ulrich, 2013) , trying to find the key to successfully attracting the opposite gender. Unfortunately product gender associations are neglected in these endeavors. Other product category attributes, like symbolic or functional use of the product have been used in research on cross gendered line extensions (Lee, 2006) and had a significant effect on the evaluation of the line extension.

(9)

9

1.1 Brand gender and product gender

In the wake of Aakers' (1997) paper on brand personality associations, Grohmann (2009) discovered that consumers also have brand gender associations. She confirmed that a brand gender congruent with the consumers gender achieved positive results for purchase

intention, brand loyalty and brand affect. One of the experiments in her paper also found that the gender of a spokesperson promoting a brand, determined (among other factors) the

brands gender identity. The next experiment concluded that a brand gender congruent to that of the consumer lead to positive results, as stated above. Now it would be easy to assume, based on this, it is best for a brand to be promoted by a spokesperson with a gender congruent to the gender of the consumers that it targets. However, in the product gender perceptions research area, opposite results were found. They found two important things, first that the gender of a spokesperson provides a solid cue about the gender of the product (Debevec & Iyer, 1986b)and second that a masculine product achieves better results when promoted by a women and vice versa (Debevec & Iyer, 1986a) . So consumers like a brand gender that is congruent with their own, yet prefer a product gender that is incongruent. Also, product gender perceptions are based on the human associated with a certain product category. So instead of measuring the effect of spokesperson gender on brand gender, did Grohmann (2009) actually measure the effect of spokesperson gender on product gender, which in turn determined brand gender?

1.2 Problem statement

To answer the previous question we need to know what the nature is of the relationship between product gender and brand gender. Current studies in brand gendering tend to neglect the effect of product gender perceptions, meaning there is hardly any information

(10)

10

about the connection between brand gender and product gender. Companies are continuing their efforts to reach the "other" gender with their products, mainly via branding. Cars are branded as feminine and, as seen before, nail polish as masculine. To study discover what the result is of a discrepancy between product gender perceptions and brand gender identities and achieve an initial understanding of the relationship between the two constructs a research question was formulated:

How does a discrepancy between brand gender identity and product gender perceptions have an effect on, brand preference (BP), brand affect (BA) and purchase intentions(PI)?

1.3 Theoretical contribution

The significance of both the gender of a brand and product gender perceptions for consumer decision making has found evidence in academic literature. Researchers studying brand gender and product, often mention the need for researching the link between the concepts (Azar, 2013; Darpy & Azar, 2009; Fugate & Phillips, 2010; Grohmann, 2009), yet no serious effort has been made yet. Brand gender identity is now a proven concept and has recently received academic attention (Azar, 2013), a solid knowledge of the relationship with product gender perceptions will help researchers interpret the results. Better understanding the effect of a discrepancy will lead to a better understanding of the relationship between product gender and brand gender, inspire future studies and finally combine the two related research areas.

1.4 Managerial contribution

Especially in the grooming industry, men are becoming an increasing target for marketing efforts (McNeill & Douglas, 2011). At one time men were the consumers that bought the shampoo and shower gel in one, but cosmetics are not accepting that any longer. Day

(11)

11

cream, hair products and even make up are targeting the male consumer. A better

understanding in the effect of product gender perceptions and brand gender on purchase intention and brand outcomes will increase the effectiveness of their campaigns. Knowledge about the effect of product gender perceptions, specifically when they are incongruent with the brand gender, will help marketing managers tailor their campaigns

1.5 Research outline

In this study, first available theory about consumer gender, brand gender and product gender perceptions will be discussed. Based on the available theory hypotheses will be developed that, when tested, will provide the answer to the research question. A survey was developed, to test for the response to a discrepancy between brand gender and product

gender perception. This will be compared to the response a congruent brand gender and product gender perception elicit. Both will be measured using scenarios involving a fictional brand, to avoid previous associations with existing brands confounding the results. After analyzing the results, they will be discussed in the light of the literature on the subject and finally a conclusion can be drawn about the research question.

(12)

12

2 Literature review

2.1 Differentiating between sex and gender

To accurately provide an overview of the literature concerning gender identity, we first need to make a distinction between the concepts of sex and gender. Psychology and consumer behavior research tend to use the terms interchangeably , adding to a confusion about the definitions of sex and gender (Fischer & Arnold, 1994). Sex refers to the biological and physiological aspects of being male or female. This includes chromosomes and DNA that, within the current biological framework, indicate that an individual is either a male or a female. This definition of sex is supported by other research on gender and has remained relatively stable over time (Fischer & Arnold, 1994; Palan, 2001; Sherif, 1982; Ulrich, 2013; Unger, 1979; West & Zimmerman, 1987). Gender, however, has many different definitions. These may vary with subject, area of research or even the preference of the researcher writing the paper. Unger (1979) introduced the concept of gender in the mainstream psychology (Eagly, Beall, & Sternberg, 2005), defined as: "those nonphysiological

components of sex that are culturally regarded as appropriate to males or to females". A seminal work on gender by West and Zimmerman (1987) viewed gender as an

accomplishment. An activity of managing proper conduct in light of normative conceptions of appropriate behavior and attitudes. In the consumer behavior domain gender has been defined as "the psychological features associated with sex" (Fischer & Arnold, 1994) and "gender is the cultural definition of behavior defined as appropriate to the sexes in a given society at a given time. Gender is a set of cultural roles "(Lerner, 1986 in Palan, 2001, p. 2). We chose to adopt a definition and distinction that is inclusive of this variety of descriptions of gender. The definition is based on one used in the book by Otnes and Zayer (2012). They

(13)

13

state that while sex is designated based on biology, gender is socially and psychologically constructed (Otnes & Zayer, 2012 p.434). We altered the definition slightly, to make it more inclusive, based on the previous definitions:

Gender is defined as the socially, psychologically and culturally constructed features of a person associated with belonging to a certain sex.

2.2 Gender identity

Now that we have defined gender and distinguished it from sex, the next step is to go from gender to gender identity. Although researchers often mention both gender and gender identity, little is explained about the relationship the concepts have to each other. Palan (2001) explained this relationship as:

"As children continue to become culturally socialized, they add to their belief system regarding gender, developing cognitive networks of associations to biological sex. One type of association that children learn is the culturally defined personality traits linked to being male (masculine traits) or female (feminine traits). The extent to which an individual identifies or thinks of him- or herself as masculine or feminine is what we now refer to, in consumer behavior research, as gender identity."

(Palan, 2001, p. 3)

This is one of the few articles where the semantic and theoretical relationship between gender and gender identity is specified. The theories on gender and gender identity will be discussed in the next chapter, for now we will focus on the semantics. The definition Palan (2001) used, is mentioned more often. It was originally developed by Fischer and Arnold (1994), and is typically used to define gender identity in consumer behavior research. Earlier

(14)

14

research has used similar concepts to gender identity, but under different labels. Some of the more famous ones include "sex- typing" (Bem, 1981) or "sex-role stereotypes" (Spence, Helmreich, & Stapp, 1975), but in marketing research it seems a trend to use "gender identity"(Palan, 2001; Schertzer, Laufer, Silvera, & McBride, 2008; Ye & Robertson, 2012). For the purpose of this research we will adopt the definition as given by Palan (2001) and Fischer and Arnold (1994):

"The extent to which an individual identifies with masculine or feminine personality traits."

When researching gender identity, two other, related but not similar terms appear: gender roles and gender role attitudes (Fischer & Arnold, 1994; Palan, 2001; Ulrich, 2013). Although sometimes used interchangeably, they are not the same. Gender roles refer to the behaviors that are culturally related to masculinity, or femininity that an individual chooses to adopt (Palan, 2001). In Western cultures, for a person who identifies with a feminine gender identity this may include becoming a nurse or taking care of someone. Gender role attitudes refer to the beliefs about which roles, rights and responsibilities are appropriate for men and women.

For this research we work with gender identity. The next paragraph will further elaborate on theory about gender identity and its value in consumer behavior and marketing research.

2.3 Theory on gender identity

There are two competing theories on the nature of gender identity. The first is "gender schema theory", developed by Bem (1981) and the second is the "multifactorial gender identity theory" conceived by Spence (1993). In this chapter we will discuss these theories and provide an argumentation for the view we take on gender identity.

(15)

15

2.3.1 Gender schema theory

Before the 1970's gender identity was viewed as having the same dichotomy as biological sex. It represented a bipolar continuum, with masculinity and femininity as opposite ends (Palan, 2001; Vantieghem, Vermeersch, & Van Houtte, 2014). Bem (1974) developed the first view on gender identity as consisting of two separate orthogonal dimensions: masculinity and femininity. Both dimensions can exist within an individual within varying degrees. Based on the existence of masculine and feminine traits within an individual, the population can be divided into four groups: high masculine/high feminine, high masculine/low feminine, low masculine/high feminine, low masculine/low feminine. People who display both a high degree of masculinity as well as a high degree of femininity are considered psychologically androgynous. Undifferentiated is a second atypical outcome, where an individual identifies neither with masculinity or femininity (Bem, 1974; Schertzer et al., 2008).

According to Bem (1981), not only do individuals display and acquire features consistent with their gender identity, it also serves as an organizing framework. Gender identity is said to function as a principle through which individuals process information about themselves and the world around them. Bem (1981) refers to this framework or principle as a gender schema. The extent to which this gender schema influences information processing depends on whether or not someone is "sex-typed". Sex-typing is the psychological process by which a male or female child, learns to be masculine or feminine (Bem, 1985). A sex-typed

individual, also referred to as gender schematic, will be more likely to frame the

environment and the self based on his or her gender identity, than a non-sex typed person (Palan, 2001; Schertzer et al., 2008). The gender schema theory is considered a unifactorial theory of gender identity, because according to Bem (RW.ERROR - Unable to find

(16)

16

reference:64; 1974; 1981), gender identity and gender related phenomena can be entirely measured by measuring one factor: personality traits. A famous critique on this vision came from Spence(1981; 1993), who proposed a multifactorial gender identity theory as

alternative.

2.3.2 Multifactorial gender identity theory

Applying the gender identity construct as proposed by Bem, did not always lead to

significant results in studies. While some researchers blamed the measurement instruments, a growing group believed it was because of the way gender identity was defined and

operationalized (Palan, 2001; Vantieghem et al., 2014). Among them was Spence (1993), who devised a theory that a variation of gender-related phenomena contribute to gender identity, including, but not limited to, masculine and feminine personality traits. Not only are there several phenomena underlying the concept of gender identity, each of these factors can be present in different people in varying degrees. Some of these factors might be associated with each other, while some of them are totally unrelated. The level of certain gender-related factors in an individual is also situation specific, with certain attributes being salient only in situations where they are likely to impact behavior(Palan, 2001; Spence, 1993). Gender-schema theory certainly captures one important contribution to the total gender identity construct, but is still just one of several and does not account for situational differences. The underlying factors of gender identity fall into four categories: (1) an

individual's global sense of maleness or femaleness; (2) gendered personality traits; (3) gender-related interests, role behaviors and attitudes; (4) sexual orientation (Palan, 2001; Spence, 1985). If only one of these underlying phenomena is measured, the results can only be predictive of situations in which that factor is salient.

(17)

17

2.4 Gender identity in this study

One thing we noticed in researching gender identity, is that in the multifactorial gender identity theory, the conceptual boundaries between gender and gender identity seemed to be blurring. The generally accepted definition of gender identity as "the extent to which an individual identifies with either masculine or feminine personality traits", is violated by the assumptions multifactorialists have about gender identity. They state that it is a combination

of related behaviors, attitudes, role behaviors and interests. This plethora of gender-related constructs, seems more similar to the definition of gender, combining the social, cultural and psychological consequences of belonging to a certain sex category.

This does not mean we solely support the unifactorial gender-schema theorists: claiming

that gender, on an individual level, is measured just by identification with personality traits is too simplistic. Therefore we choose to use an approach from a recent study on gender and consumer behavior. Ulrich (2013) used the multifactorial gender identity theory, as a general theory on individual gender. She states that the four factors, mentioned earlier, are

contributions to the general construct of individual gender, with gender identity as one of these underlying contributors. So in this study we will accept that only one contributing factor of gender is measured, not individual gender as a whole: it is limited to gender identity, defined as an individual’s identification with masculine or feminine gender identities.

2.5 Gender identity in consumer behavior and marketing research

Marketers use sex as an important segmentation variable and gender identity has been linked to several brand-consumption and product-consumption related constructs(Fischer & Arnold, 1994). We highlighted some of the significant relationships found in marketing and

(18)

18

consumer behavior studies. Fischer and Arnold (1994) linked gender identity to Christmas shopping behavior. The researchers defined and measured gender identity as the

identification with desired masculine or feminine personality traits, showing more similarity with the gender-schema theory than the multifactorial theory. Support for their approach comes from the fact that (as Bem (1974) proposed) their model showed that gender identity was indeed measured by two separate dimensions: masculine personality traits and

feminine personality traits. They also found evidence for the conceptual difference between the constructs sex, gender identity and gender role attitudes. The main result of the study was that individuals with a feminine gender identity (this can be both male and female respondents) had greater psychological involvement with and got more enjoyment out of the Christmas shopping.

Palan (2001) reviewed 38 studies in the period 1963 - 2001 researching the effect of gender identity on a consumer behavior variable. Masculine and feminine personality traits were linked to purchasing decisions, perceptions of brand and product gender and information processing abilities. Gender identity can also have a significant positive impact on brand variables, like brand attitude and purchase intention. The researcher concludes by stating that, if properly measured and operationalized, gender identity can be a valuable construct in the consumer behavior research area.

2.6 Self-Congruity theory

"For marketers, the issue of gender is critically important. Consciously or unconsciously, it is probably the first segmentation variable that marketers apply. This is a rational response since, from a consumer perspective, men and women want to know that the product is “theirs”."

(19)

19

(Milner & Fodness, 1996)

The reason gender identity influences consumer behavior can be explained with self-congruity theory. Self-self-congruity is an abbreviated notation of self-concept/product-image congruity. According to self-congruity theory consumer behavior is, at least in part,

determined by a psychological process where the image of a product (or brand) is compared to the self-concept. The self-concept is defined as the totality of the individual's thoughts and feelings having reference to himself as an object. An individual's self-concept is expected to affect behavior in such a way that his or her perception of self can be enhanced through the consumption of goods as symbols (Earl & Kemp, 2002). This can result in either a high congruity outcome, where the product image is perceived as a match with the self-concept, or a low self-congruity outcome, where the product image is perceived as

incongruent with the self-concept. A match in this case is associated with positive effects on several outcome variables, like purchase intention or brand loyalty. This is called the "self-congruity effect". (Sirgy, 1982; 1997). The motivation for this self-"self-congruity effect lies in the cognitive-consistency theories, which assume that consumers strive for consistency in their beliefs and behaviors. This can be applied to consume behavior in the sense that consumers prefer to buy products and brands that are congruent with how they perceive themselves (Malär, Krohmer, Hoyer, & Nyffenegger, 2011; Sirgy et al., 1997). The better the match, the higher the chance that an individual will purchase something, or hold favorable attitudes. In this section we will shortly discuss gender, brands and products in the light of self-congruity theory.

(20)

20

2.7 Gender identity and the self concept

Gender plays a big role in consumers thinking about others and thinking about the self. It is therefore not surprising that gender is a central part of the consumers self-concept (Cross & Markus, 1993; Hende & Mugge, 2014). Gender identity is also mentioned separately as playing a big role in the self-concept of the consumer, reinforcing the importance of gendered personality traits to the overall self-concept (Patterson & Hogg, 2004).

In self-concept and self-congruity research gender identity is often referred to as sex-role self-concept. This is a different label for the same concept as gender identity. Palan (2001) and Arnold and Fischer (1994) already advised against the use of “sex-role”, because of the misuse of the term for different concepts in marketing and consumer behavior research. The

word "sex-role self-concept" is a variation on this same theme. Studies that employed this term generally used the same definition and measurement instruments associated with gender identity. This is highlighted by a review of sex-role self concept and the measures in marketing by Stern (1988). In her article on sex-role self-concept, she mentions the

conceptualization of femininity and masculinity as separate constructs, which both contribute to the sex-role self-concept construct. This conceptualization by Bem (1974) is commonly used to define and explain gender identity, as is the evolvement from one bipolar construct, to separate orthogonal constructs. Sex-role self-concept and gender identity have been used for interchangeably for the same concept. For the sake of consistency and given the previously discussed issues with "sex role", we continue with gender identity. This means we use the concept of “gender identity congruity” (Feiereisen, Broderick, & Douglas, 2009) to refer to the possible congruity between the representation of masculinity or femininity in a brand or product and the gender identity of an individual. How brands or product can be attributed a gender identity will be discussed in the next section.

(21)

21

2.8 Brand personality and gender

Approximately ten percent of the variance in consumer behaviors, attitudes and intentions can be explained by the self-congruity effect (Aguirre-Rodriguez, Bosnjak, & Sirgy, 2012). This means that organizations are motivated to create brands and products that congruent with their target consumer. Anthropomorphism refers to the tendency of people to attribute human characteristics to non-human objects, like products or brands (Aaker, 1997; Eisend &

Stokburger-Sauer, 2013). Since self-congruity tells us that consumer prefer brands or products that are consistent with their self-concept, anthropomorphism has a positive effect on brand outcomes like loyalty and consumer evaluations (Kressmann et al., 2006; Parker, 2009). To reap the benefits of the self-congruity effect, marketers actively try to bestow human attributes onto their brands and products. This is a two way street, since consumers naturally attribute human characteristics to brands and products (Aaker & Fournier, 1995; Dolich, 1969; Eisend & Stokburger-Sauer, 2013). Assigning human

characteristics to brands has led to the development of the brand personality, this refers to the set of human characteristics associated with a brand. A brand personality can be created by indirect or direct communication with the consumer. Antecedents of brand personality include advertising, consumer demographics like age and education and endorsers of the brand (Aaker, 1997; Eisend & Stokburger-Sauer, 2013). Brand personality is composed of five dimensions: sincerity, ruggedness, excitement, sophistication and competence (Aaker, 1997). The greater the congruity between the personality traits associated with the brand and the self of the consumer, the more the brand can benefit. Via the self-congruity effect, brand personality is positively associated with brand outcomes like brand image, brand attitude and purchase intention(Eisend & Stokburger-Sauer, 2013; Valette-Florence, Guizani, & Merunka, 2011).

(22)

22

2.8.1 Brand gender identity

Levy (1959) was one of the first to start the discussion on the symbolic meaning of

consumption and therefore of brands. He made a point of saying that one of the first human attributes people see in inanimate objects is a gender. Using symbols to distinguish

masculine from feminine is eminent in all cultures and times. Just as gender identity plays an important role in the human personality (Digman, 1990), brand gender identity exists

alongside brand personality. Relative to brand personality, brand gender identity refers to the masculine and feminine personality trait associations one has with the brand. Grohmann (2009) researched the idea of a brand having a gender identity as separate from the brand personality traits. The study found proof of the existence of a brand gender identity or brand gender personality. The construct had the same orthogonal nature as gender identity in humans: so a brand can also be masculine, feminine or even androgynous (Grohmann, 2009; Lieven, Herrmann, Landwehr, & Van Tilburg, 2011). Grohmann (2009) developed a scale, to measure the two dimensions of brand gender identity: brand masculinity and brand

femininity. Each dimension is measured by rating a brand on several personality traits that are either masculine (e.g. assertive) or feminine (e.g. gentle) with a Likert-type scale. The measure has evolved in literature, creating separate scales for measuring either brand masculinity or brand femininity(Azar, 2013; Darpy & Azar, 2009). Grohmann finds positive brand outcomes (e.g. brand affect, brand loyalty, purchase intention) due to the self-congruity effect that is created when the gender identity of a brand is consistent with the self-concept of a consumer. The positive effect of brand gender congruity on brand equity has also found empirical support (Lieven et al., 2011).

(23)

23

2.9 Product personality and gender

In discussing product gender or product personality, we distinguish between a product category or class and an individual product. Several research papers on product personality have been written from the industrial design perspective, to determine whether aesthetic factors of a product can create a personality (for example: Govers & Mugge, 2004). Product personality in that sense refers to the set of personality characteristics people use to

describe a specific product and to discriminate it from other products in the same product category (Govers & Mugge, 2004; Govers & Schoormans, 2005). These type of papers refer to a product, one specific product, a computer for example. In this study and more

commonly in consumer behavior research, when the term product is used, it refers to the product category or class in general; all computers. It does not refer to any one specific product.

The notion that products can possess an image, associations other than the purely functional ones, has first been proposed by Levy (1959) in his work on symbolic consumption. Sirgy (1982) states that a product image is determined by the stereotype of the user. Not every product will create personality associations, they have to meet three requirements to elicit personality associations: the product has to be consumed visibly, there has to be variability in the product category and the product has to "personalizable". A product is considered personalizable, when the use of a product can be attributed to a generalized image of the stereotypical user (Aguirre-Rodriguez et al., 2012; Sirgy, 1982; Sirgy et al., 1997). The image of a product seems to be determined mostly by the humans that are associated with it. Although there is no research on the personality of product categories as such, the gendering of products has received some attention.

(24)

24

2.9.1 Product gender perceptions

When discussing the gender identity of products, the term product gender perceptions is typically employed. This deviates from the previous habit of using gender identity, but with good reason. Where gender identity in people and brands is measured on two separate, multi-item scales, product gender perceptions are determined by just one question: "to what extent do you perceive this product as masculine/feminine?". This question has been validated and used in different cultures and in different timeframes. (Fugate & Phillips, 2010; Milner & Fodness, 1996; Milner, Speece, & Anderson, 1990).The reason this measure and operationalization is different is because product gender perceptions are based on the humans associated with the product. The product gender is based on the gender identity of the consumer associated with the product class (Patterson & Hogg, 2004; Sirgy, 1982). And their gender identity can be either masculine or feminine. Therefore it is assumed that consumers do not associate the underlying gendered personality traits with products, but just the gender identity they represent. This can also be seen in the way androgynous products are discussed. Like brands and people, product gender is not one-dimensional, but rather is measured by two dimensions: masculinity and femininity. Therefore, products can also be androgynous or undifferentiated. But, unlike brands and people, this does not mean they are identified with both masculine and feminine personality traits. Androgynous products may be those which are "used" equally by men and women while undifferentiated products may simply be those which have not yet established a gender image (Debevec & Iyer, 1986b). This different conceptualization and operationalization of the construct, lead us to the decision to keep the term "product gender perceptions", rather than adapting it to "product gender identity".

(25)

25

Product gender perceptions revolve around the stereotypical user. To assign a product a gender, it has to be associated with feminine or masculine stereotypical personalities and traits (Alreck, 1994). Consumers observe cues about the gender of the product in their environment, these include the gender of promoters, gender of users of the product and the gender of people seen buying the product. These cues can be developed via advertising and promotion, priming product gender perceptions (Milner & Fodness, 1996; Patterson & Hogg, 2004). As stated before, the assigning of human characteristics to products and brands is a two way street, so gender identity and biological sex are a significant influence on product gender perceptions (Patterson & Hogg, 2004). Consumers try to recognize themselves in products and naturally attribute human characteristics to products in the process of anthropomorphism. Product gender perceptions tend to remain relatively stable across time. Fugate and Phillips (2010) replicated a 1996 Milner and Fodness study on the perceptions of product gender and found that there were hardly any shifts in the view of masculinity and femininity (Fugate & Phillips, 2010; Sirgy, 1982). Interestingly, the research area of gendered products and brands is void of studies connecting product gender

congruence to outcome variables. It focuses more on the antecedents and nature of the product gender, rather than the consequences. When starting to discuss the symbolic meaning of products, or the non-functional usage of a product the debate tends to quickly focus on brands. Earlier research often used brand image and product image

interchangeably, adding to the lack of research specifically on the image of a product

category (Malhotra, 1988; Sirgy, 1982; Sirgy, 1985) .We believe a possible explanation of this void is that marketers put so much effort towards differentiating their one product and brand from other products in the category that little research is done on the associations of the product category as a whole.

(26)

26

2.10 The relationship between brand and product associations

So both product and brands can have images, personalities and genders. But what

differentiates a product from a brand? According to Keller, anything can be branded, but for it to be a brand, it has to be consciously and strategically managed as such. Industries or organizations can choose to consciously manage the image of a product category as well, but that does not turn it into a brand. "A brand is more than a product, because it can have

dimensions that differentiate it in some way from other products designed to satisfy the same need" (Keller, 2013, p.31). This is a differentiating aspect, since the personality or gender associations with the product class computers, cannot help differentiate one computer from another. Brands do have that ability.

There is very limited research into the relationship between product and brand associations. Patterson (2004) mentions the success of assigning gender neutral products, a specific gender by giving the product a masculine or feminine brand name. This could indicate that a product gender could be based on, or created by a brand gender in some way. For the opposite view there is also evidence. Lamb and Low (2000) studied brand associations and indicated that it was a faulty assumption that brand associations were similar across product categories (e.g. the brand Philips in the television market, compared to the audio market), rather they were organized per category. So the product category influenced which brand associations were created. Grohmann (2009) showed that spokespeople in advertisements significantly influenced perceptions of brand gender. As stated before, endorses are also an important source of product gender perceptions.

It is a bit of a chicken and the egg story, what came first? It seems logical to assume that product gender perceptions and other product associations were there first, since without a

(27)

27

product to brand, there is no brand. And most brands, when they enter a product category, they enter a network of associations that is already related to the product category.

2.11 Brand gender identity and product gender perceptions

Due to the knowledge vacuum in this area, researchers in the brand and product gendering arena have wondered what would happen when there is a discrepancy between the gender of a product and that of a brand (Azar, 2013; Darpy & Azar, 2009; Fugate & Phillips, 2010; Grohmann, 2009). Research into cross gendered brand extensions showed that product type (symbolic vs. functional) had a significant effect on the perception of fit of cross gender brand extensions (Lee, 2006). This analysis, however, did not take product gender into account.

No research yet has tried to find the effects of this discrepancy and the possible interactions effects or relationship between brand gender identity and product gender perceptions. Therefore there is also no research on the effect such a discrepancy would have on outcome variables, such as brand preference or purchase intention. Researching this effect would also contribute to the knowledge about the relationship between product category associations and brand associations and hopefully stimulate future research.

Therefore the research question we will try to address in this paper is:

How does a discrepancy between brand gender identity and product gender perceptions have an effect on, brand preference (BP), brand affect (BA) and purchase intentions(PI)?

The three brand outcome variables are chosen based on the previous research by Grohmann (2009), who finds a positive effect of brand gender identity congruence on all of these factors. The two brand constructs (brand affect and brand preference) are grouped under

(28)

28

"brand outcomes" in the hypotheses and the model. This research aims to determine how and if this relationship changes in the situation of a conflicting brand gender identity and product gender perception. Other (brand) outcomes have been connected to brand gender identity, but due to the scope of this research we focus on these. The research on product gender perception typically focuses on purchase intention as the sole outcome of interest (Fugate & Phillips, 2010).

(29)

29

3 Conceptual Framework

Based on existent literature, hypotheses can be developed regarding the nature of the relationship between brand gender identity and product gender perceptions

Although there is no literature available on the gender specific topic, there is research suggesting that consumers organize brand associations or images by product category (Low & Lamb Jr, 2000). This product based association network, leads to the conclusion that brand associations are fundamentally product driven. So they are, linked to a certain product category. Brand personality and the related brand gender identity are also considered associations, or images available in the consumers mind. Based on the former reasoning, we hypothesize that the brand gender identity is at least in part driven by the product gender perceptions.

Hypothesis 1a:

Product gender perceptions have a significant positive effect on purchase intention (PI).

Hypothesis 1b:

The effect of product gender perceptions on purchase intention is moderated by brand gender identity.

This means that if there is congruity between the product gender perceptions and the brand gender identity, there will be a stronger positive effect on purchase intention, compared to when there is no match.

A product category alone can have no direct effect on brand outcomes, since there is no impact or any effort to differentiate between products in the same category. Without the

(30)

30

benefit of strategically managing and communicating a brand, it is unlikely associations about it are formed.

Hypothesis 2a:

The effect of product gender perceptions on brand affect is mediated by brand gender identity

Hypothesis 2b:

The effect of product gender perceptions on brand preference is mediated by brand gender identity

As specified in the literature review, the positive effects of product and brand gendering can be attributed to self-congruity. To account for this effect we incorporate gender identity in the framework. A conscious choice was made not to use sex, as research has already

indicated the relationship between brand gender identity and brand equity is not moderated by biological sex (Lieven et al., 2011). Due to the conceptual similarities between gender identity, brand gender identity, and product gender perceptions, we believe that gender identity might have such an interaction effect.

Hypothesis 3:

The effect of product gender perceptions on purchase intention is moderated by gender identity.

Hypothesis 4a:

The effect of brand gender identity on brand affect is moderated by gender identity

(31)

31

The effect of brand gender identity on brand preference is moderated by gender identity

H3 H1a H1b H4 H2

Figure 1. Conceptual model Product Gender Perceptions [MPGP/FPGP] Purchase Intention [PI] Gender Identity [MGI/FGI] Brand Gender Identity [MBGI/FBGI] Brand Outcomes [BA, BP]

(32)

32

4 Methodology

4.1 Research design

The purpose of this research is to compare two levels of two independent variables. To make inferences about the relationship these variables have with the dependent variables we used a 2x2 between-subjects experimental design; product gender perception

(masculine/feminine) x brand gender identity (masculine/feminine). Since gender identity in this model is not manipulated, it is not considered as part of the experimental model. The gender identity is "a given" of a respondent and is not influenced by the researchers. A second argument to not include it here is that gender identity functions solely as a

moderator variable. This means we hypothesize it will influence the relationship, but in itself has no effect on the dependent variables. This differentiates it from the hypothesized

mediator brand gender identity, since a mediator does affect the dependent variables.

A between-subjects design was chosen to avoid the negative effects that are associated with a within-subjects design, such as carry-over effects. An inherent disadvantage of the

between-subjects design is that by assigning different people to different groups, previously existing differences between the groups may explain variance in the dependent variables rather than the manipulation of the independent variable (Dooley, 2001). This risk can be managed by randomly assigning respondents to groups, as was done in this study. However, it remains a limitation (Kantowitz, Roediger III, & Elmes, 2014).

4.2 Required stimuli for treatment conditions

To manipulate the product gender perception and the brand gender identity variable, we needed to develop stimuli that would prime the correct product gender perceptions and brand gender identity. This paragraph explains which product categories were chosen to

(33)

33

prime the product gender perceptions and why. In the second part we explain how two brands were developed to communicate the brand gender identities.

4.2.1 Product category

We needed two products, one masculine and one feminine product. A main requirement was that the two products differed significantly from each other with respect to gender perception. On the other hand the products had to differ equally from the midpoint of the scale. So product A has to be as masculine as product B is feminine. This is to avoid any chance that differences occur due to one product gender being perceived much stronger than the other.

Table 1. Masculine and feminine products according to Fugate&Phillips (2010)

Masculine products Feminine products

Beer Bath soap

Car Wine

SUV Digital camera

Coffee Facial tissue

Athletic shoes Food processor

Lawnmower Frozen vegetables

Potato chips Hairspray

To pick a suitable product category, we used a list of product categories that was previously used and tested for product gender perceptions by Fugate and Phillips (2010). The list of products and whether they were feminine or masculine is presented in Table 1. Perceptions of product gender are based primarily on the humans associated with the product, so they differ per culture. The products Fugate and Phillips (2010) used were validated in the US, but the results might differ substantially in the Netherlands, where this study is conducted. Also, we needed to choose two products that would score equally on the masculinity and

(34)

34

To choose the right stimulus product, a pre-test survey was developed and distributed via email and Facebook to 35 respondents. Of the total 35 participants, 32 completed the survey. The pre-test was designed to replicate Fugate and Phillips's (2010) research, so the same method and procedure were applied.

Respondents were asked to rate each of the products on a femininity scale (1 =highly

feminine, 2 = feminine, 3 = not at all feminine) and a masculinity scale (same as previous but with masculinity). This was a slight deviation from the original research, where there were four scale points (0 = not at all feminine/3 highly feminine). This was done intentionally, since symmetrical scales allow ordinal data to be processed as interval data (Carifio & Perla, 2007; Jaccard & Wan, 1996, p. 2; Leon, Brown Sr, Ruch Sr, & Johnson Sr, 2003, p. 66). This pre-test also gave us the chance to see this measurement instrument in practice. It is a controversial instrument, since it only has one item to measure each of dimensions, yet it is commonly used in product gender perception research (Fugate & Phillips, 2010; Milner & Fodness, 1996; Milner et al., 1990).

Results of the pre-test

In Table 2 the means for the products on both the masculinity dimension and the femininity dimension are presented. The results are interpreted by using paired sample t-tests for each product score, to see if there is a significant difference (p < ,05) between the femininity and the masculinity score. If that difference is significant, than the dimension with the lowest mean (closest to 1) determines the product gender. In Table 2 all products that were masculine or feminine are flagged with an asterisk. The other products can be either androgynous or undifferentiated, but it is not relevant to discuss that here.

(35)

35

Table 2. Means for each product on masculinity and femininity

* Significant difference in scores masculine and feminine construct (p<.05)

To find the suitable stimuli, a paired sample t-test was performed to find two products that were equally masculine/feminine. This meant that their means on the masculine/feminine scale did not differ significantly. The two products that came forward were the food processor (M = 2,61/1,88, SD= ,47/,61) and the lawnmower (M= 1,78/2,78, SD=,49/,42). Table 3. shows the non significant difference in the scores, as was a result of the paired samples t-test.

Table 3. Non-significant difference between products in paired samples t-test

To conclude, for the product gender perception manipulation we used the lawnmower as a masculine product and the food processor as a feminine product.

Product (N=32) Masculinity score Femininity score

Beer 1.75* 2.63 Car 1.69* 2.44 SUV 1.78* 2.66 Coffee 2.25 2.25 Athletic shoes 2.25 2.19 Lawnmower 1.78* 2.78 Potato chips 2.22 2.44 Bath soap 2.91 1.34* Wine 2.55 1.81* Digital camera 2.44 2.31 Facial tissue 3.00 1.25* Food processor 2.69 1.88* Frozen vegetables 2.56 2.44 Hairspray 3.00 1.19* T Df p Pair 1

Food processor (masculine) - Lawnmower (feminine)

-1.139 31 .263

Pair 2

Lawnmower (masculine) - Food processor (feminine)

(36)

36

4.2.2 Brand

The main requirement for a brand to serve as a suitable stimulus, was that it properly communicated either a masculine or a feminine brand gender identity. Given the close proximity of the brand gender identity concept to other brand associations like brand personality, the decision was made to not use an existing brand. The prevents earlier associative networks from influencing the dependent variables.

We developed the feminine and masculine brand based on literature. First we used the theory of Grohmann (2009) to attribute characteristics to the brand, that are indicative of a feminine or masculine brand gender identity. Second, an article by Wu, Klink and Gruo (2013) described how the phonetic features of a brand name can influence brand gender identity. They found that so called back vowels (the o sound in bronze) create a masculine gender identity, while front vowels (the i in fish) attribute to a feminine brand gender identity.

Based on this information we developed two brands: Krog (masculine brand) and Frisé (feminine brand). Each brand was accompanied by a text where the gendered information about the brand was provided, priming the correct brand gender identity. An example of such a scenario priming a feminine brand gender identity with the masculine product is:

"I would like to introduce you to a new brand of lawnmowers: Frisé. Please read this information about the brand Frisé carefully.

Frisé produces high quality lawnmowers. As a brand of lawnmowers, Frisé is delicate and refined. Frisé is now available at local stores in the Netherlands. Frisé tends to be warm in their communication towards customers and consumers.

(37)

37

The next questions are about the brand Frisé."

With the developed stimuli, four of these "scenarios" were created and participants were randomly assigned to one of the conditions. Two of the conditions were a "match", so gender of product and brand were congruent and two resulted in a "mismatch".

Table 4. Four conditions, resulting in a match or a mismatch FEMININE product/FEMININE

brand

Match FEMININE product/MASCULINE

brand

Mismatch

MASCULINE product/MASCULINE brand

Match MASCULINE product/FEMININE

brand

Mismatch

4.3 Measurement instrument

To measure the different variables in the theoretical model, we used existing measures. This saves on time and allows comparing the results of this study to previous research, since differences that appear can then be attributed to differences in the sample and the design, rather than the measurement instrument (Dooley, 2001). Two of the dependent variables were measured on a semantic differential, rather than a Likert-type scale. This choice was made for two reasons, first of all this type of scale has been validated in recent research (Grohmann, 2009) and second adding semantic differential scales to the questionnaire will diminish acquiescence bias without diminishing the psychometric quality (Friborg,

Martinussen, & Rosenvinge, 2006). All the scales used to measure the different variables were found and tested in the scientific literature on the subject.

4.3.1 Independent variable

Product gender perceptions

As explained before, product gender perceptions can be divided in two dimensions:

(38)

38

are asked to rank the product on the given dimension. We adjusted the scale after the pre-test that was discussed earlier. Respondents indicated that they found it confusing that the lowest score (1 out of 3) on the scale correlated with the highest ranking on

femininity/masculinity. In the measurement instrument this was changed, so that the highest score correlates with the highest perception of femininity/masculinity. An example of this one item scale was: "I would indicate this product as: not masculine at all/highly masculine".

The use of single -item measures has received major criticism, mainly for being unreliable or their reliability being difficult to assess with common measures of reliability. Other critiques focus on the lack of scope or precision of a single-item scale (Gliem & Gliem, 2003).

However, the single item measure is commonly used in the research on product gender perceptions and has delivered significant results (Fugate & Phillips, 2010; Milner & Fodness, 1996) . To manage the disadvantages of the single-item measure, we decided to use a semantic differential measure in the questionnaire. Rossiter (2007; 2002; 2008) is a researcher that is in favor of the use of single-item measures in marketing research, if the object being measured can be described as "concrete and singular" (Bergkvist & Rossiter, 2007). However, his research criticizes the use of the Likert-scale when working with single-item measures. According to this scientist the neutral midpoint is what creates the "fuzzy" results and should be avoided. In marketing research he advocates using a semantic

differential scale when working with single-item measures (Rossiter, 2008). The change to a semantic differential scale is further substantiated by remarks respondents of the pretest made, stating that the distance between "not at all masculine(feminine)" and the midpoint "feminine(masculine)" seemed larger than the distance between the midpoint and "highly feminine(masculine)". We decided against creating a bipolar semantic differential scale, with

(39)

39

feminine and masculine at opposite ends. This is because it would violate the innate nature of the product gender perception construct. Like gender identity and brand gender identity, the construct consists of a combination of the two separate dimensions of masculinity and femininity. This means that the single-item measure of the masculinity dimension in this research looked like this:

4.3.2 Interaction variables

Brand gender identity

The hypothesized mediator-variable brand gender identity was measured with a two-dimensional scale measuring feminine and masculine brand gender identity. This scale was

developed and validated by Grohmann (2009), proving highly reliable with  = .90 for the

feminine brand gender identity dimension and .masculine brand gender identity dimension. The scale has already been validated in an Italian context, proving it is reliable and suitable for cross-cultural usage (Ugolini, Cassia, & Vigolo, 2014). A total of 12 items were measured on a 7-point Likert-type scale, where 1 equals strongly disagree and 7 resembles strongly agree. Respondents were asked to indicate to what extent they would describe a brand as displaying a certain gendered personality trait. An example of an item measuring feminine brand gender identity was "tender" and for a masculine brand gender identity it was "aggressive".

(40)

40

Gender identity

As we discussed in the literature review, several scales were developed to measure gender identity, the most notable being the Personal Attributes Questionnaire (PAQ) (Spence, Helmreich, & Stapp, 1974) and the Bem Sex Role Inventory (BSRI)(Bem, 1974). The BSRI was developed further by Barak and Stern (1986) who created a shorter version, consisting of twenty items (half of the forty items the original BSRI consisted of). Researchers validating both versions found that the short form was more reliable, especially when measuring the femininity dimension. Where the long version scored a Crohnbach's alpha between .75 and .78, the BSRI short-form generated an alpha between .84 and .87 for the femininity

dimension. The shorter version also fitted gender identity data better, compared to the long form (Martin & Ramanaiah, 1988). A recent cross-cultural validation of the short-form in Mexico (extremely masculine culture), Norway (extremely feminine culture) and the US (baseline), further decreased the number of items to sixteen. Items in this study were eliminated based on their contribution to the scale reliability, loading on the desired factor, cross-loading on the undesired factor and their contribution to the overall fit (Schertzer et al., 2008). This two-dimensional scale, consisting of sixteen items (eight items for each dimension of gender identity) is referred to as the Gender Trait Index (GTI). This was the measurement instrument was used here. It had the advantage that it had been validated quite recently and cross-culturally, making it the best option available.

The GTI consists of two dimensions, feminine gender identity and masculine gender identity. The respondent was asked to indicate on a 7-point Likert scale to what extent a certain trait is true for him or her. Traits include for example "assertive" or "tender".

(41)

41

4.3.3 Dependent variables

For this study we measured three outcome variables, of which two were brand outcome variables. The measurement instruments were all based on the instruments used by Grohmann (2009) in her research on the effects of brand gender identity.

Brand affect

Brand affect was defined as a brand's potential to elicit a positive emotional response in the average consumer as a result of its use (Chaudhuri & Holbrook, 2001). Brand affect was measured with three items on a 7-point Likert scale (1 = strongly disagree/7= strongly agree), an example of an item is "This brand will make me happy".

Brand preference

When a consumer has "brand preference", it means he or she will choose the brand over competitive offerings in the same product category (Michman, Mazze, & Greco, 2003). It was measured with a semantic differential scale, where the respondent was asked to indicate their degree of preference relative to other brands in the same category on two bipolar, 7-point scales (1=very poor/7=very good).

Purchase intention

The final dependent variable was purchase intention. Behavioral intention tends to capture the motivation factors of behavior. The stronger the intention to engage in a behavior, the more likely it is that the behavior is actually performed. Purchase intention was measured with a semantic differential scale, where respondents indicated the likelihood of them buying the product of the primed brand on a 7-point scale (1=unlikely/7=likely).

(42)

42

4.4 Demographics

The respondents participating in this study were asked about their age, sex and education level, to exclude any possible influence these demographics had on the variables in the research and to accurately describe the sample.

4.5 Translation of the questionnaire

Since the questionnaire was distributed in Dutch, it was important to develop a translated equivalent of the original construct measures as presented above. To achieve this a

technique called "parallel translation" was employed. This is process where multiple people translate the same questionnaire and the translations are compared to reach the optimal wording (Usunier, 1998, p.52). In total, three translations of the questionnaire were developed. First, a preliminary translation was done by someone who had an academic background and would understand the relevance of the items. Second the questionnaire was checked by two bilingual people, one reviewed the preliminary translation and the other translated the original to Dutch again (Del Greco, Walop, & Eastridge, 1987). The three resulting questionnaires were compared and differences in translation were reviewed. The resulting Dutch questionnaire is Appendix A and the three translations are in Appendix B.

4.6 Sampling technique and procedure

The final questionnaire was developed using software by Qualtrics (www.qualtrics.com), which was made available to us by the University of Amsterdam. The target sample in this study was defined as Dutch consumers between the age of 15 and 65. A second requirement of our target sample was that they had to be Dutch. Gender identity and perceptions of gender are culturally determined (Palan, 2001), so mixing respondents from different cultures could influence the results. The questionnaire was spread via social media, which

(43)

43

meant we were dependent on the availability of respondents. This method is considered non-probability convenience sampling. Since subjects select themselves for participation in the study, this may lead to a biased sample and poor representation of the population (Gravetter & Forzano, 2011). Fugate and Phillips (2010) used a convenience sample of students for their study on product gender perception and argued that it was a suitable sampling technique because gender identification is a broadly based phenomenon and is likely to be registered in most population frames. Relying solely on a web-based

questionnaire had a clear time advantage, however it had the disadvantage of missing those people in a population that do not have access to internet or social media. The minimum sample size to run effective statistical analysis on is 30 (Saunders, Lewis, & Thornhill, 2009, p.218), so for each of the scenarios we needed at least 30 respondents: a total of 120 respondents was the minimum sample size.

All respondents to the questionnaire were greeted with a introduction message explaining they should read carefully, guaranteeing their anonymity and thanking them for

participating. After completing the questionnaire they were given the option to comment on the questionnaire and thanked for their participation again.

4.7 Data analysis

To test the hypotheses mentioned in the third chapter data analysis, with the Statistical Package for the Social Sciences (SPSS), version 20, was performed.

4.7.1 Differences between variables/groups

To test for differences between the variables independent-samples t-test were performed when there were only two groups, for more groups the (two-way between-subjects) ANOVA and ANCOVA were used. The ANOVA and all variations on it hold certain requirements for

(44)

44

the data (Pallant, 2010, p.197). For example, the dependent variable has to be measured on a interval/ratio level. Officially Likert and semantic differential scales (conservative stance) are ordinal scales, but if they are symmetrical and computed on a scale-level it is accepted practice to analyze them as interval-level measures (Brown Sr, Ruch Sr, & Johnson Sr, 2003, p.66; Carifio & Perla, 2007; Jaccard & Wan, 1996, p. 2). A second assumption discussed here is the normality assumption. There was no explicit testing for violations of the normal distributions, since sufficient samples size should prevent any violations creating problems. The final assumption is homogeneity of variance, which means that the variability of the scores for the groups in each scenario is similar, since the groups come from a population with equal variances. To test this, each ANOVA /t-test was be accompanied by the "Levenes test for equality of variances", which indicates if the difference in variances within a given sample is significant (p<.05). If this was significant, the assumption was violated. In that case, either the "equal variances not assumed" option was used, or the value for the robust

Brown-Forsythe test of equality of means was calculated. In the results chapter, these assumption checks were only reported if a violation occurred.

4.7.2 Relationships between variables

To check whether variables explain a significant portion of the variance in a specific dependent variable and if they predict the results in this variable (multiple) regression analysis was used. The entry of variables was forced entry and never hierarchal or stepwise. When performing a regression analysis, certain assumptions about the data must hold, for example the independent variables cannot be too correlated (multicollinearity) or perfectly correlated (singularity). To avoid these issues with the data, independent variables that had a Pearson's product moment correlation of r=.800 or over, were not included in the same analysis. Furthermore for each of the regression analyses performed here the

(45)

45

multicollinearity statistics (tolerance and variation inflation factor (VIF)) and the

Mahalanobis distance (checks for outliers, the maximum of this statistic cannot be higher than a certain critical value that is dependent on the number of independent variables, check Table 5 for the critical values) were included in the output of the analysis. Only if the assumptions were violated, it was reported.

Table 5. Critical Mahalanobis distance for # of independent variables

# of independent variables Critical value # of independent variables Critical value # of independent variables Critical value 2 13.82 4 18.47 6 22.46 3 16.27 5 20.52 7 24.32

Referenties

GERELATEERDE DOCUMENTEN

With the collapse of the diamond market, the number of blacks employed declined from 6 666 in 1928/1929 to 811 in 1932 and workers began to stream back to the

Compared to the South African RCA index (Figure 4.1), it is clear that Argentina has a revealed comparative advantage for the entire period for all the sunflower seed products

In this study, four milli-scale soft robots are designed (Inchworm, Turtle, Quadruped and Millipede) and their actuation under external magnetic fields is investigated with

In a previous study, we showed that healthy people were able to control an active trunk support using four different control interfaces (based on joystick, force on feet, force

To estimate the potential effect of different light colours on the pollinator’s contribution to variation in female reproductive output, we calculated the per flower

Kaplan-Meier curve showing time to discontinuation of levodopa-carbidopa intestinal gel (LCIG) infusion. The vertical tick marks denote censored observations. The horizontal

Multinational Hotel Group Development and Urbanization: A Study of Market Entry Mode in the second and third tier Cities of

H2D: Consumer attitude (consumer evaluation, purchase intention and willingness to pay a price premium) towards the brand extension will be more positive for low