The effect of wrongdoings by anthropomorphized brands on the relationship between brand and person

(1)

The effect of wrongdoings by

anthropomorphized brands on the

relationship between brand and person

Master Thesis

Author: Joni Kort (10670920)

University of Amsterdam, Faculty of Economics and Business July 2014

Under supervision of: Drs. Ing. A. Meulemans Second assessor: Prof. Dr. J. Tettero

(2)

UvA - Master Thesis - Joni Kort - 10670920 Page 2

Table of content

I. Executive Summary 4

II. Important abbreviations 6

III. Introduction 7

IV. Theoretical Framework 10

4.1 ‘Big Five’ 10

4.2 Brand personality 12

4.3 Brand Anthropomorphism 14

4.4 Brand Relationship Quality 15

4.5 Brand Wrongdoings 19

4.6 Brand versus Product 20

4.7 Hypotheses 20 4.7.1 Conceptual Model 21 V. Research Methods 25 5.1 The survey 25 5.2 Pre-test 29 VI. Findings 30 6.1 Pre-test 30

6.2 The actual test 31

6.3 Testing hypothesis I 33

6.4 Testing hypothesis II 43

VII. Conclusion 53

VIII. Limitations 55

(3)

X. Appendices 60

10.1 Appendix I - The TIPI 60

10.2 Appendix II - The advertisements 61

10.3 Appendix III - The 17-item BRI 62

10.4 Appendix IV - The news article 63

10.5 Appendix V - The TIPI syntax 64

10.6 Appendix IV – Findings 65 10.6.1 ‘Relaxing’ 65 10.6.2 ‘Erotic’ 69 10.6.3 Conscientiousness 73 10.6.4 Extraversion 77 10.6.5 Openness to Experience 81

10.6.6 Conscientiousness/ Emotional Stability 85

10.6.7 Emotional Stability 89

(4)

I. Executive summary

In the past five years of me being a business student, the field of marketing and the psychological aspect of it really gained my interest. The changes that appear in business, from selling to satisfying needs, result in more focus on the person behind the customer. Instead of dividing the market in large groups to sell your product; companies are trying to recognize markets’ needs first and adapting to that. This change brings a whole new range of interesting research fields.

In this research the focus is on the relationship between different types of advertising, personal characteristics and the effect of brand wrongdoings on that relationship. The aim of the research is to find out the effect of brand anthropomorphism on brand relationship quality in case of brand wrongdoing. Moreover, it will be researched if one’s personal character traits influences relationship quality between person and brand after wrongdoings. In other words: How are personal characteristics and brand anthropomorphism on the one hand and relationship quality on the other hand related? And in what direction does brand wrongdoing moderate this relationship?

Different theoretical models in the field of human personality, brand personality, brand anthropomorphism and brand wrongdoings are studied. This in the end leads to the following two hypotheses:

Hypothesis I a: Brand wrongdoing has a negative effect on the relationship between person and a brand.

Hypothesis I b: This effect is stronger for an anthropomorphized version of a brand. Hypothesis I c: There is a difference between ‘erotic’ and ‘relaxing’

(5)

UvA - Master Thesis - Joni Kort - 10670920 Page 5 Hypothesis II a: The effect on the relationship quality between person and brand after brand wrongdoing is influenced by the person’s character (Big Five).

Hypothesis II b: the effect on relationship quality is stronger for Conscientiousness than for Emotional stability (Neuroticism), which is stronger than Extraversion, which is stronger than Openness, which is stronger than Agreeableness.

An empirical study with use of a survey is conducted after which can be concluded that brand wrongdoings have a negative effect on the relationship between person and brand. However, anthropomorphism is not the reason per se that brands are punished more heavily for their wrongdoing. It appears that one kind of anthropomorphism, namely ‘erotic’, is more likable and gets punished less for the brand wrongdoing. Anthropomorphism in the theme of ‘relaxation’, does cause the brand being punished more heavily.

It also appears that difference in personal characteristics cause different reactions on brand wrongdoing. However, the ranking which was expected from reviewing (brand) personality theory cannot be completely confirmed. Indeed, conscientiousness persons are affected most by brand wrongdoings, and agreeable persons are, together with emotional stable people, least affected by brand wrongdoing in terms of relationship quality.

From this research more conclusions can be drawn. For example, it looks like Heineken’s customer base is not the most loyal, which can be concluded from the negative Net-Promoter Score. Heineken can expect their customers being vulnerable for competitors’ actions. Furthermore, from this research it was also visible what factors customers find important in expressing their opinion about Heineken. In both before and after wrongdoing condition, questions about communication with other customers weight high. Before wrongdoing, also product specific questions scored high weights, while after wrongdoing, company behavior and brand image appeared important.

(6)

II. Important abbreviations

TIPI = Ten Item Personality Inventory NPS = Net-Promoter Score

BRQ = Brand Relationship Quality BRI = Brand Relationship Index PCA = Principal Component Analysis KMO= Kaiser-Meyer-Olkin

(7)

III. Introduction

Doing business is not only about selling your product anymore (product focused), but is about satisfying customers’ needs. This shift also influenced marketing practices; there is a shift from product focused marketing to customer focused marketing (Homburg, Workman & Jensen, 2000). Customers are more seen as individuals with own values and needs. Because of the fact different persons react differently on their environment, and on the behavior of companies, they need to know their customers and how this customers stay or leave the company. This is achieved by delivering value for every individual customer.

Marketers’ task to build a loyal customer group is becoming more challenging since the introduction of the internet. Information is accessible for a very low cost; customers use each other’s opinions as references and companies cannot hide anything anymore.

Furthermore, the shift to consumer focused marketing pairs with the need for

personality research in the field of business and marketing. A person’s personality influences the buying decisions one make (Kassarjian, 1971), but how? And how does the customer react on a company’s behavior? This paradigm creates new opportunities for researchers and marketers and my interest is in the field of psychology crossing the path of economics, and how this is changing over the last few decades.

Personality can be deﬁned as ‘the intrinsic organization of an individual’s mental world that is stable over time and consistent over situations’ (Mulyanegara, Tsarenko & Anderson, 2009).One of the most extensive used models to group these different personalities in types is the “Big Five” framework, based on the “Five Factors Theory” developed by Costa and McCrea in the 1980s (Goldberg, 1990). Moreover, distinction by personality not only holds for human beings, also brand and products can be distinguished through different types of personality models. For example, brands can be divided in certain

(8)

UvA - Master Thesis - Joni Kort - 10670920 Page 8 brand personalities (Aaker, 1997) and anthropomorphism versus non-anthropomorphism (Puzakova, Kwak & Rocereto, 2013).

Because of the differences within peoples characters, I assume they will react

differently on brands with different brand personalities or a type of brand anthropomorphism, and therefore develop different relationships with certain brands. In turn, people will react differently on brand wrongdoings en therefore their relationship with the company will be influenced differently. Moreover, gender, age, income, lifestyle etc. may all affect the relationship but for this thesis the interest is mainly in the character difference between people. For this research the five dimensions of the “Big Five Model” are used to determine the relationship between brand and person.

The aim of this study is to discern the effect of brand anthropomorphism on brand relationship quality. And if one’s personal character traits influences relationship quality between person and brand after wrongdoings,

This aim leads to the following research question:

How are personal character type and brand anthropomorphism on one hand and relationship quality on the other hand related? And in what direction does brand wrongdoing moderate this relationship?

To answer this question, theory will be studied and an empirical study will be conducted in a form of a survey.

In the next chapter, several theories and models will be explained on which a conceptual model and hypotheses are based. These theoretical models include the fields of

(9)

UvA - Master Thesis - Joni Kort - 10670920 Page 9 human personality, brand personality, brand anthropomorphism, brand relationship quality and brand wrongdoings.

In the method section it can be read how these hypotheses will be tested. Then the analysis and results will be provided. A conclusion and discussion will follow the results. Limitations and implications can be found at the end of this thesis.

(10)

IV. Theoretical Framework:

4.1 Human personality - The “Big Five”

The “Big Five” model was originally developed because researchers in personality psychology needed a technique/model to study specified domains of personality traits, rather than examining separately the thousands of particular attributes which makes a person an unique individual (Goldberg, 1990; John & Srivastava, p. 102, 1999). Nowadays, the models is extensively used in Marketing as well.

The “Big Five” serves as an integrative framework, that represents diverse systems of personality description in one common framework. The five dimensions represent personality in a very broad level of abstraction, rather than it implies that differences in personality can be reduced to only five traits (John & Srivastava, 1999).

The Five Factors are typically labelled as:

I. Intellect or Openness to experience; accompanied by traits as being imaginative, curious, original, intellectual, broad-minded, intelligent and artistically sensitive.

II. Conscientiousness; reflects traits like being careful, thorough, orderly, responsible, organized and dependable.

III. Extraversion or Surgency; mirrors traits like, being sociable, gregarious, talkative, energetic, active and assertive.

IV. Agreeableness; reflects traits like being courteous, flexible, good-natured, forgiving, tolerant, cooperative and trustful.

V. Emotional Stability versus Neuroticism; mirrors traits like, being anxious, depressed, angry, emotional, worried and insecure (Barrick & Mount, 1991).

(11)

UvA - Master Thesis - Joni Kort - 10670920 Page 11 through the “Big Five”. One of this methods is the Ten-Item Personality Inventory (TIPI) scale developed by Gosling, Rentfrow, and Swann (2003). The scale is based on the 240-item NEO Personality Inventory; Revise (NEO-PI-R) scale of Costa and McCrae (1992), which takes a respondent 45 minutes to complete the test. Gosling et al. (2003) developed the TIPI scale to shorten the measurement time of a persons’ personality extensively.

The TIPI (Appendix I) consists of 10 personality traits for which the respondents have to indicate in a 7 point Likert-scale to what extend the trait applies to them. The scale rates from 1(disagree strongly) to 7 (agree strongly). Examples of the traits are; anxious/ easily upset and, extraverted/enthusiastic.

Research in the field of human personality influencing economic behavior is done in several directions. Griskevicius and Kendrick (2013) started at the very beginning and questioned if buying behaviorreflects the operation of ancestral motivations and stated that a person’s preferences, behaviors, and decision processes can change dramatically in predictable

ways depending on which motive is currently active. The seven ancestral motivation Griskevicius and Kendrick (2013) take into account are:

I. Evading physical harm II. Avoiding disease III. Making friends IV. Attaining status V. Acquiring a mate VI. Keeping a mate VII. Caring for family

(12)

UvA - Master Thesis - Joni Kort - 10670920 Page 12 Other research shows that behavior is not only influenced by the ancestral motives, but also by characteristic differences between people. For example, Bosnjak, Galesic & Tuten (2007) show that persons with different scores on the “Big Five” test, show different

behavior with respect to online purchasing. In a more economic perspective, it is researched that personality influences both how a person deals with its financial situation (Donnely, Iyer & Howell, 2012) and marginal utility of one’s income ( Boyce & Wood, 2011). Besides character differences between persons, people’s self image also influences their buying behavior. Consumers do not only consume products for their functionality or utility, but also consume the symbolic meaning of products as portrayed in their images (Jamal & Goode, 2001). In their article, Jamal and Goode (2001) state that self-image congruity is both a very strong predictor of consumers’ brand preferences, and a good predictor of consumer

satisfaction.

So if it is the case that different persons react differently on brands, it should be taken into account that not all brands are the same. A distinction can be made between brands in terms of Brand Personality (Aaker, 1997) and Brand Anthropomorphism (Puzakova, Kwak, Rocereto, 2013).

4.2 Brand personality

The Brand Personality model is developed by Jennifer Aaker in the late ‘90’s. Brand personality refers to “the set of human characteristics associated with a brand” (Aaker, 1997).

Consumer personality impacts preference towards specific brand personality

(Mulyanegara, Tsarenko, Anderson, 2009). The authors find that the “Big Five” dimensions are significantly related to people’s preferences in certain dimensions of brand personality (Mulyanegara et al., 2009). Mulyanegara et al. (2009) provides the following example, a person who has a conscientious personality, generally demonstrates a preference towards a

(13)

UvA - Master Thesis - Joni Kort - 10670920 Page 13 ‘trusted’ brand. On the other hand, Extrovert persons are more motivated by ‘sociable’

brands.

According to Aaker (1997), brand- and human personality traits differ in how they are formed. As for the human personality traits; perceptions are inferred on the basis of the individual’s behaviors, physical characteristics, beliefs and attitudes and demographic characteristics. While on the other hand, perceptions of brand personality are formed and influenced by any contact between the consumer and the brand, either direct or indirect. This leads to an enormous amount of personality treats that individuals assign to the brands. In the development of the framework, Aaker (1997) divided 114 traits in five dimensions and determined which words best subsumed the five distinct personality dimensions:

I. Sincerity (down-to-earth, honest, wholesome, cheerful) II. Excitement (daring, spirited, imaginative, up-to-date) III. Competence (reliable, intelligent, successful)

IV. Sophistication (upper class, charming) V. Ruggedness (outdoorsy, though)

The evolution of consumer-brand relationships, in times of product/ brand

wrongdoings for example, are researched in respect of the Brand Personality model (Aaker, Fournier, Brasel, 2004). The model suggests that consumers react more negatively to brand wrongdoings when the brand is positioned as sincere than when positioned as exciting.

A few years after development of the model, it is being criticized by Azoulaz and Kapferer (2003). They believe that Aaker’s five dimensions do not measure brand personality but only merge a number of brand identity dimensions.

(14)

UvA - Master Thesis - Joni Kort - 10670920 Page 14 4.3 Brand anthropomorphism

A part of brand personality and a way to distinguish brands among each other is the level of Brand Anthropomorphism; “the attribution of uniquely human characteristics and features to nonhuman creatures and beings, natural and supernatural phenomena, material states or objects, and even abstract concepts (Epley, Waytz, Cacioppo, 2007). Puzakova, Kwak and Rocereto (2013) simply define it as “the extent to which a brand in humanized”. In other words, brand anthropomorphism is the attribution of uniquely human characteristics and features to a brand.

Brand anthropomorphism is used extensively in the field of marketing to create favorable consumer reactions, such as more favorable attributions of brand personality (Puzakova, et al., 2013). Anthropomorphism of brands trigger consumers’ perceptions of brands as living entities with their own emotions, intentions, characteristics, human like motivations etc. (Puzakova et al., 2013). According to Aggarwal and McGill (2012) that is because either the consumer sees the human in nonhuman, or marketers suggest humanizing the product or brand. Anthropomorphism happens to make products more distinctive or memorable, to assign specific qualities where the brand stands for, and to make a brand more endearing and likable (Aggarwal, McGill, 2012).

According to Puzakova et al. (2013), prior research has focused on positive effects of brand anthropomorphism but they argue that “anthropomorphic positioning of a brand can have negative repercussions if the brand is perceived as responsible for its wrongdoings”. So it is important to understand if and how anthropomorphism of a brand can backfire when product failure occurs (Puzakova et al., 2013). In their article they found that humanized brands appear mindful, conscious, and capable of intentional actions and therefore are perceived as being responsible for performing negative actions. Because of this increased held responsibility of anthropomorphized brands for their actions, negative publicity caused

(15)

UvA - Master Thesis - Joni Kort - 10670920 Page 15 by product wrongdoing is likely to lead to less favorable consumer evaluations when a brand is anthropomorphized than when it is not (Puzakova et al., 2013).

For me, it is not clear in het existing literature if different versions of brand anthropomorphism elicit different reactions in terms of likability, either with or without brand wrongdoings. Thorough inspection of the literature did not cleared view for me.

4.4 Brand relationship quality

At the basis of a brand-person relationship, there is the interaction or exchange

between brand and consumer. For a relationship to exist, interdependence between partners is required, which is the partners collectively affect. It defines and redefines the relationship (Hinde 1979 in Fournier, 1998). A way to legitimize a brand as a partner is to highlight the ways in which brands are animated, humanized or personalized for example (Fournier, 1998). Moreover, persons differ not only in how they perceive brands, Fournier (1998) also suggests that there is a difference in how persons relate to brands. Aggarwal (2004) states that

sometimes, persons form relationships with brands in much the same way as in which they form relationship with each other in a social context. This is evidenced by the fact that persons often do not make the distinction between brands and manufacturers of the brands. They feel the company is the brand and the other way around. Especially when (potential) customers interact with a representative of a brand, they have a tendency to fall back on social relationships as a guide in their interactions. Moreover, even when the company only focusses on selling physical products, consumers may think of the brand as a living entity. This is due to the existence of ‘animism’, which is the belief that objects possess souls. This is recognized in the domain of products (Aggarwal, 2004). The underlying key in Aggarwal’s (2004) research is that when consumers form relationships with a brand, norms of

(16)

UvA - Master Thesis - Joni Kort - 10670920 Page 16 examined: exchange relationships and communal relationships. Which are relatively one in which benefits are given to others to get something back and, when benefits are given to show concern for other’s needs (Aggarwal, 2004). It is concluded that “an action that is in violence of a relationship norm leads to a poorer evaluation by the consumers relative to an action that is in conformity with the relationship norm”.

But why are relationships formed? According to Esch, Langner, Schmitt and Geus (2006), the goal of companies building relationships with consumers is the construction of strong current and future purchases. Their research indicates that to secure future purchases, “a familiar brand with a positive image must build a positive brand relationship with the consumer”. Moreover, the facets of such a relationship include both cognitive and affective components; brand satisfaction and brand trust that culminate in brand attachment, which is a key driver of future purchases.

The quality of a relationship in the end can be determined by different theoretical models; four will be described below. The first model is the Brand Relationship Quality (hereafter BRQ) model, developed by Fournier in 1998 (Smit, Bronner, Tolboom, 2007). It is a customer-based indicator of the strength and depth of a consumer-brand relationship.

Seven elements build up the BRQ:

I. Intimacy facet, which is the psychological closeness between the person and brand and the knowledge about the brand.

II. (personal) Commitment refers to the loyalty to the brand in terms of willingness and faithfulness to make small sacrifices.

III. Passion (passionate attachment), is the integration of the brand in its user’s daily life; it represents attachment and the feeling that something is missing if the brand is not used.

(17)

UvA - Master Thesis - Joni Kort - 10670920 Page 17 V. Self-concept connection, which reflects the extent to which the brand is part

of the self, part of the self-image of a person, and refers to the question whether the consumer and his or her brand have lots in common.

VI. Nostalgic connection, which means that the brand is part of the consumers’ history and related to particular memories.

VII. Brand participant quality, it refers to the qualities of the partner in the relationship, and whether the brand takes good care of the person, shows interest and is reliable (Smit et al., 2007).

Differences in relationship strength and depth are caused by differences are in satisfaction levels, perceptions of brand quality or the brand’s personality for example (Fournier 1998).

Building upon the BRQ model, the Brand Relationship Index (BRI) model is developed by Zhou (2006). The aim of the model is to “measure comprehensive brand relationship strength in a broad sense” (Zhou, 2006). Five factors of relationship quality are described;

I. Commitment/ connection; measures the extension of the relationship and the fit between the partners.

II. Attachment/ attention; measures kindness between consumers of the brand and the consumer’s intention to pay attention to promoter and brand.

III. Familiarity/ awareness; measures the knowledge the consumer has about product and the marketer’s important attributes.

IV. Trust/ respect; measures the respect the consumer received and the consumer’s trust in the marketer.

V. Association/ recognition; measures the consumer’s impressions on the brand symbols and logo (Zhou, 2006).

(18)

UvA - Master Thesis - Joni Kort - 10670920 Page 18 Zhou (2006) designed a questionnaire of 29 items that can be used to test relationship quality in a broad sense.

The two models describes below, test relationship quality by measuring the level of customer loyalty in specific. In 2003, Fredrick Reichheld developed the Net-Promoter Score (NPS). According to Reichheld (2003), there is only one question that has to be asked to test true loyalty and obtain a NPS; how likely are you to recommend the company to a friend or a colleague? The NPS measures loyalty and can predict company growth (Keiningham, Cooil, Wallin Andreassen & Aksoy, 2007). Important here is to clarify what loyalty is. “Loyalty is the willingness of someone – a customer, an employee, a friend – to make an investment or personal sacrifice in order to strengthen a relationship” (Reichheld, 2003). When loyal customers recommend a company to their friends, colleagues, or family, it is one of the best indicators of loyalty “because of the customer’s sacrifice, if you will, in making the

recommendation”. This is because one puts its own reputation on line by recommending a company. Further explanation about how to measure the Net-Promoter Score will be provided in the methods section of this thesis.

The fourth model is the Brand Resonance Model (Keller, 2013 p. 107-111). The model describes four building blocks or layers of development in a brand-person relationship.

I. Salience; which measures the awareness of a brand and how easily and often a brand is evoked under various circumstances and situations.

II. Brand meaning; which is divided in two paths; performance and imagery. The first is about how well the product or service meet the persons’ functional needs. The latter is about to what extend the brand satisfies the more psychological or social needs.

(19)

UvA - Master Thesis - Joni Kort - 10670920 Page 19 III. Response level; consisting of judgment of and feelings about the brand.

Judgments are the personal opinions about and evaluations of the brand of both the performance and imagery block. Feelings are the persons’ emotional responses and reactions to the brand.

IV. Resonance; which is the measure for the strength of the relationship in this model. Brand resonance describes the nature of the relationship and the extent to which persons feel that they are “in sync” with the brand (Keller, 2013). Brand resonance consists of two dimensions; intensity and activity. The two dimensions are divided into four categories of brand resonance: behavioral loyalty, attitudinal attachment, sense of community, and active engagement (Keller, 2013). The relationship quality in this case is measured by the level of brand resonance.

4.5 Brand Wrongdoings

A wrongdoing is defined as “illegal or dishonest behavior” in the Oxford advanced learner’s dictionary (Hornby, 1948 p. 1770). Brand wrongdoing in that sense is illegal or dishonest behavior by the brand. When brand wrongdoings are determined as intentional, they are perceived more negatively than actions categorized as accidental. Examples of wrongdoings are illegal actions like ‘fraud’ or ‘bribery’. On the other hand, examples of dishonest behavior are ‘use of child labor’ or ‘paying an unfair low wage’.

(20)

UvA - Master Thesis - Joni Kort - 10670920 Page 20 4.6 Brand versus Product

It is important to note that there is a difference between a brand and a product, although in the theory the concepts are often used reversible.

According to Keller (2013, p. 30) a brand is “a name, term, sign, symbol, or design, or a combination of them, intended to identify the goods and services of one seller of group of sellers and to differentiate them from those of competition”. A product on the other hand is “anything we can offer to a market for attention, acquisition, use of consumption that might satisfy a need or want” (Keller, 2013 p.31). In this thesis, an advertisement of the product is shown to the respondents while the wrongdoing is done by the company (the whole brand). Using the brand and the product reversible is possible in this thesis because the product and brand are often seen as one, carry the same name and the audience sees the product being the brand.

4.7 Hypotheses

This research includes two main directions; relationship strength with different types of anthropomorphism and relationship strength with different personalities.

There will be researched if and how different types of anthropomorphism affect the relationship between brand and person before and after brand wrongdoing. The second part of the study researches if certain character type of people react differently on brand wrongdoing, and therefore if relationship quality differs in strength, before and after brand wrongdoing.

(21)

UvA - Master Thesis - Joni Kort - 10670920 Page 21 This leads to the following conceptual model (figure 1):

Figure 1: The conceptual model

The conceptual model consist of two independent variables, brand anthropomorphism (three sorts) and personal character (Big five), and one dependent variable; relationship quality. Brand wrongdoings moderate the relationship.

From previous research it is known that product or brand wrongdoings have a negative effect on the relationship between a person and a brand (Puzakova et al. (2013). From the research of Puzakova et al. (2013) it is known that anthropomorphized brands are seen as living entities with own intentions, characteristics, and human like motivations.

Furthermore, anthropomorphism increases likeliness of a brand. However, in case of brand wrongdoing; it can be questioned if anthropomorphism will backfire. This is because Brand Anthropomorphism -‘normal’ -‘relaxing’ -‘erotic’ ‘Big Five’ - Openness - Conscientiousness - Extraversion - Agreeableness - Emotional Stability Relationship Quality Brand Wrongdoings

(22)

UvA - Master Thesis - Joni Kort - 10670920 Page 22 “anthropomorphic positioning of a brand can have negative repercussions if the brand is perceived as responsible for its wrongdoings” (Puzakova et al., 2013). This leads to hypothesis one.

Hypothesis I a: Brand wrongdoing has a negative effect on the relationship between person and a brand.

Hypothesis I b: This effect is stronger for an anthropomorphized version of a brand. Hypothesis I c: There is a difference between ‘erotic’ and ‘ relaxing’

anthropomorphized versions of the brand in terms of relationship change.

A person’s character does influence how people view and value things in life, and how a person reacts to certain happenings, either good or bad. Persons can be distinguished among each other by different psychological theories. For example; Puzakova et al. (2013) distinguished two types of implicit personality to research whether consumers’ attributes of responsibility to the humanized brand faced with negative media attention are qualified by the application of these two types of personality. According to this distinction, there are entity theorists (holistic thinkers) and incremental theorists. Respectively, people that believe

personality traits are fixed and expect a high degree of consistency in behavior, versus, people that believe personality is somewhat changeable and view behavior as varying, either over time or across situations (Dweck, Chiu, Hong, 1995). Entity theorists form less

favorable attitudes towards an anthropomorphized brand than a non-anthropomorphized when faced with negative brand information due to product wrongdoing. On the other hand, for incremental theorists, negative brand information caused by a product wrongdoing does not result in differences in attitudes toward either an anthropomorphized or a

(23)

UvA - Master Thesis - Joni Kort - 10670920 Page 23 As mentioned before, persons react differently on happenings in life. It is assumed that this is partly due to differences in human character, which leads to hypothesis II a.

Hypothesis II a: The effect on the relationship quality between person and brand after brand wrongdoing is influenced by the person’s character (Big Five).

In this thesis, personal characters are measured with and distinguished by the ‘Big Five’ model. Research is being done about the overlapping between the ‘Big Five’ model and Aaker’s five brand personality dimensions (Sweeney, Brandon, 2006). They argue that Competence corresponds with Conscientiousness, Excitement to Extraversion and that Sincerity somewhat relates to Agreeableness. Openness and Emotional stability do not directly relate to one of the five brand personality dimensions.

Extraversion and Agreeableness are the interpersonal dimensions; together they are the dominant dimensions underlying human personality (Sweeney, Brandon, 2006). Because of these differences between the five dimensions, it is expected that brand wrongdoings do not lead to the same effect on relationship quality between brand and person.

Analyzing the different traits reflecting the five factors, it can be expected that due to traits like orderly and, responsible; Conscientiousness persons will be most influenced by the wrongdoing. Furthermore, due to traits like anxiousness, angry, and emotional; Emotional Stability (Neuroticism) will be second most influence by brand wrongdoing. Extraversion, with traits like energetic, talkative, and social will be third most influenced. Both Openness and Agreeableness humans are expected to be least influenced by brand wrongdoings. This is because of their accompanying traits which are respectively imaginative, curious, broad-minded and, flexible, forgiving and, tolerant. This expectation leads to hypothesis II b.

(24)

UvA - Master Thesis - Joni Kort - 10670920 Page 24 Hypothesis II b: the effect on relationship quality is stronger for Conscientiousness than for Emotional stability (Neuroticism), which is stronger than Extraversion, which is stronger than Openness, which is stronger than Agreeableness.

(25)

V. Research Methods:

This descriptive research includes three variables; Brand Anthropomorphism (normal, erotic and relaxing), Character Type (“Big 5” dimensions) and Relationship Quality (Net-Promoter Score and Brand Relationship Index). Brand wrongdoings act as a moderator on this relationship.

The data used for this descriptive research is primary data collected through a survey. The brand used in this research is Heineken. By choosing this brand for this research, some requirements apply to the sample including the respondents being over the age of 18, since than they have proper knowledge and experience with the beer industry. Ideally the sample should consist of people to which Heineken at least somewhat appeals. Therefore, the sample consists of (university) students and recently graduates, in the age of 20-30. The respondents will be approached online, via my social media network and to friends via direct mailing. Since the survey will be online, the sample should have access to, and knowledge of the use of, internet.

In the most ideal situation, the participants’ character types are divided so that all of the “Big five” dimensions are represented even. However, this cannot be known upfront, therefore the sample should be rather large, to increase reliability and validity of the research. At least 150 - 175 respondents are needed.

5.1 The survey

The survey conducted consists of several parts, from which the “Big Five” test is the first. The Ten-Item Personality Inventory (TIPI) scale developed by Gosling, Rentfrow, and Swann (2003) is used to determine the respondents’ character type.

(26)

UvA - Master Thesis - Joni Kort - 10670920 Page 26 In part two of the survey, the brand is introduced to the participant and the

relationship with the brand before wrongdoing is determined.

A choice that has to be made here is if a made-up brand will be used, which eliminates the bias of differences in knowledge by participants, or an anthropomorphized version of a well-known brand will be used. A reasonable doubt of using a non-existing brand is the fact that nobody reaches the higher state of brand relationship quality and therefore the test will not be executed optimally. Therefore, it is chosen to use an anthropomorphized version of a well-known, and used, brand by the (Dutch) young adult, student population; Heineken.

An online print advertisement of Heineken beer is shown to the participants. Three different advertisements are used in this survey (Appendix II), from which a respondent is shown one. The first advertisement is a ‘normal’ non-anthropomorphized ad. The second is an anthropomorphized ad with an ‘erotic’ theme, and the third ad is an anthropomorphized ad again, but with a ‘relaxing’ theme. This is to test the difference between

non-anthropomorphized and different anthropomorphism versions of the brand and test hypothesis I.

The ad will be followed by a survey consisting of two parts. The first part is testing the Net-Promoter Score (Reichheld, 2003), the second part is testing the Brand Relationship Index (Zhou, 2006).

As mentioned before; the Net-Promoter Score (NPS) measures true loyalty of the customers by only one question. Respondents score the following question on an 11-point scale: how likely are you to recommend Heineken beer to a friend or colleague? The items range from 1 (not at all likely) to 10 (extremely likely). Respondents scoring Heineken a 9 or 10 are the promoters; they are loyal and enthusiastic about the brand and will recommend

(27)

UvA - Master Thesis - Joni Kort - 10670920 Page 27 Heineken to others. The passive customers are the respondents scoring a 7 or 8, are satisfied but not enthusiastic, they are vulnerable to competitive offerings. At last, respondents scoring Heineken a 0 to 6 are unhappy about the brand and can damage the brand through negative word-of-mouth (Reichheld, 2003). The Net-Promoter Score is measured by taking the percentage of customers who are promoters and subtract the percentage who are detractors.

NPS = % promoters (9 and 10) - % detractors (0 to 6) NPS = ((𝑥1

𝑛) 𝑥100) − (( 𝑥2

𝑛) 𝑥100)

x1 = number of 9 and 10, x2 = number of 0 to 6, n = number of participants

To test hypothesis I, the data will be divided by advertisement, after which the before and after wrongdoing NPS will be calculated and compared.

To test hypothesis II, the data will be divided by character type. Again, the before and after wrongdoing NPS will be calculated and compared.

Relationship quality is measured with the Brand Relationship Index (Zhou, 2006), which acknowledges five factors for relationship quality. Zhou developed a questionnaire of 29 items which he calls the Broad Brand Relationship Scale (Zhou, 2006). This scale

measures the relationship quality. For this thesis the questionnaire is reduced to 17 items, to shorten the time in which a respondent can complete the survey. Questions where removed which measure almost the same as other questions, in the end remaining with 17, one of a kind questions. The 17 items are rated on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). A few examples of the 17 items are; “If I buy product Y next time, I would like to buy brand X again” and “I know the development history of company X”. In the case of this thesis, product Y is beer and brand X is Heineken (The full 17 item questionnaire can be found in appendix III).

(28)

UvA - Master Thesis - Joni Kort - 10670920 Page 28 The advantage of using an existing questionnaire is the proven validity and reliability by an expert in the field and reducing time of pre-testing the survey before it actually

measures what has to be measured.

After gaining data to determine the brand relationship before wrongdoing, a news article is shown about the wrongdoing by Heineken (Appendix IV). Hereafter, the brand relationship is measured again through the 17-item BRI questionnaire and NPS. The aim is to discover a potential difference in before-and after wrongdoing condition, for the different advertisements and character types.

The change for the Brand Relationship Index is measured through comparing the means of the most important questions in the questionnaire in the before and after wrongdoing conditions. First, a Principal Component Analysis (PCA) is performed to discover underlying structures in the 17 questions, and weight the importance of the components and questions. Questions are considered to be important if they have a

component loading of >0.6. The PCA is performed with the direct Oblimin Rotation Method. The criteria for how many components to extract are based on three steps. First, the components should have Eigenvalues >1.0. Second, the ‘elbow’ in the scree-plot indicates the amount of components extracted. At last, the component matrix should indicate at least values > 0.6.

Items with a loading > 0.6 in the pattern matrix are indicated as important. The mean of the items will be measured, weighted against their loading to determine one brand

relationship quality value.

(29)

UvA - Master Thesis - Joni Kort - 10670920 Page 29 Thereafter, a Paired Sample t-Test is performed to compare the value of the before and after wrongdoing condition, and determine the relationship quality change.

To test hypothesis I, the data is split by advertisement, for which the means will be determined and compared.

To test hypothesis II, the data is split by character type, for which the means will be determined and compared.

5.2 Pre-test

Before conducting the actual research, a pre-test is will be conducted to test if the relationship after wrongdoing is worse, or more affected, than when first establishing a ‘positive’ relationship before wrongdoing.

A sample of ten test persons will be shown the ‘normal’ ad. This is because the possible effect of the different anthropomorphized versions and reactions on it is not addressed. Half of the respondents are shown the ad, and directly the news article about the wrongdoing and then the NPS and BRI questionnaire. The other five persons is shown the ad, answers the NPS and BRI, shown the news article and will answer the NPS and BRI

(30)

UvA - Master Thesis - Joni Kort - 10670920 Page 30 VI.

Findings

6.1 pre-test

The aim of the pre-test is to find out if the relationship between Heineken and customers is perceived more negative when the news article is directly shown after the ad, before first have the opportunity to think ‘positive’ about the brand.

Due to the small sample size for the pre-test, it is not possible to do an analysis with SPSS. Therefore the scores are measured by hand.

Due to errors in the survey, only five out of 10 participants completed the pre-test, from which two persons directly viewed the news article after the advertisement and three people were tested as in the actual test, were first the before condition relationship is determined before being confronted with the wrongdoing.

The three participants scored an average Net-Promoter Score of 7 for the ‘before’ wrongdoing condition. After seeing the news article, the average Net-Promoter Score is only a 5.33. Both the participant which saw the news article directly after the ad scored Heineken a 7. This would indicate that the results derived from the later research would be more radical than reality. However, when calculating the means for the BRI (table 1), not a significant difference can be seen for the ‘after’ wrongdoing condition for both groups.

(31)

UvA - Master Thesis - Joni Kort - 10670920 Page 31 6.2 The actual test

Before analyzing the data with SPSS, the data will be cleaned. This is done in three steps: checking for, and handling of missing values, handling counter-indicative items and checking for reliability.

Checking for, and handling of missing values: After running a frequency test (Analyze> descriptive statistics > frequencies), it can be seen that not all 175 respondents finished the survey. Missing values are ranging from 0.6% for Age and Gender, 10.9% for the ten item personality index, 11.4% first net promoter score, 14.9% for the first 17 item brand relationship index, 16.6% for the second net promoter score and, 19,4% for the second 17 item brand relationship index set of questions.

Some respondents quit the survey after filling out the age and gender; so no answers for personality type, net promoter score or brand relationship are provided. I deleted these respondents from the data set. In total I reduced the data set from 175 to 149 respondents. These removed respondents are totally useless because if the missing values would be replaced with the mean for example, the data set is totally biased because the potential difference between character types is not visible.

A second frequency test is run to identify the percentage of missing values. Now, no missing values for age and gender. Also no missing values for the TIPI, the first NPS, and the first BRI question set, 2.0% for the second NPS and, 5.4 % for the second BRI question set. So every item has missing values under 10% now. Further handling of missing values is done by replacing the missing values with the mean of the specific item set (transform> replacing missing values > mean). The missing values are pasted in the right cells of the data set in SPSS. Last, a frequencies test is ran again to check if all missing values are replaced, which appears to be the case.

(32)

UvA - Master Thesis - Joni Kort - 10670920 Page 32 Handling counter-indicative items: The recoding of the TIPI items will be according to a method developed by Justin S. Cheng (n.d) from the Loyala University Chicago (see appendix V). The TIPI test is structured in a way that the five dimensions are displayed by the following items (R denotes reverse-scored items): extraversion 1,6R; agreeableness 2R,7; conscientiousness 3, 8R; emotional stability 4R, 9; openness to experience 5, 10R.

Therefore item 6, 2, 8, 4, and 10 should be recoded via the following method: recode a 7 with a 1, a 6 with a 2, a 5 with a 3, etc.

The syntax provided by Justin S. Cheng (n.d.) can be pasted in my own syntax and ran for this data set to determine the respondents’ character types in the end.

The 17 item brand relationship index questionnaire does not contain counter indicative items, so recoding items (transform > recode into same variables) won’t be necessary.

Checking for reliability: For the first analysis only the Brand Relationship Index (BRI) before wrongdoing is taken into account. To compute the reliability, I ran a Cronbach’s Alpha test (analyze > scale > reliability analysis). The Cronbach’s Alpha for this 17 item BRI is 0.882, which indicates a high level of internal consistency for the BRI scale. From the Item-total Statistics it can be seen that deletion of the 16th_{and 17}th_{item results in a minor} improvement of the Cronbach’s Alpha of respectively 0.883 and 0.892. Next, the Cronbach’s Alpha test is conducted for the BRI after wrongdoing. The value for the test is in this case 0.890, which is indicates an even higher level of internal consistency for this scale. From the Item-total Statistics table it can be seen that only deletion of the 17th item of the BRI-II results in a minor improvement of the Cronbach’s Alpha of 0.901. It is chosen not to delete the items because it is such a minor improvement.

(33)

UvA - Master Thesis - Joni Kort - 10670920 Page 33 6.3 Testing hypothesis I:

To be able to answer hypothesis Ia/b/c, the data set is divided in three groups splitting the three advertisements (Data > split file). For every group, the change in NPS and BRI is determined.

The Net-Promoter Score is calculated with the following equation: NPS = ((𝑥1

𝑛) 𝑥100) − (( 𝑥2

𝑛) 𝑥100)

The Brand Relationship Index is measured through measuring the mean for the most important items in the 17-item questionnaire. A Principal Component Analysis (PCA) is conducted for all of the three advertisements in the before wrongdoing condition, to indicate the most important components (loading ≥ 0.6). The mean of these loadings will be

measured, where after a PCA for the ‘after wrongdoing’ condition is performed, and measuring a mean for those most important items. The two means will be compared via a paired sample t-test test.

The data set is still sorted by advertisement, to be able to see the difference between groups. A PCA (analyze > dimension reduction > factor) is ran to determine the weight of the factors. The Kaiser-Meyer-Olkin (KMO) measure of the ‘normal’ ad is .614 (table 2), which means the sample is suitable for performing a PCA, since the value is higher than .6. Also both the KMO measures of the ‘erotic’ and ‘relaxing’ ads are sufficient for performing a PCA, respectively being .864 (table 3) and .815 (table 4). From the tables it can also be seen that For all three the advertisements, the value of Bartlett’s Test of Sphericity is .000, which is less than .05, which indicates again that the data is suitable for performing a PCA.

(34)

table 2 table 3 table 4

In the after wrongdoing condition the data is also suitable for performing a PCA; the outcome of the KMO test being .708, .758, .854, for respectively the ‘normal’, ‘relaxing’, and ‘erotic’ ad. Bartlett’s Test of Sphericity is .000 for all three the groups (see table 5, 6 and 7).

table 5 table 6 table 7

Further analysis will be done by advertisement, starting with the ‘normal’ advertisement.

‘normal’ advertisement, non-anthropomorphized

NPS; the first group represents the ‘normal’ ad, which is not anthropomorphized. A total number of 42 respondents where provided with this ad. Before wrongdoing, 6

respondents scored a 9 or 10, 18 respondents a 7 or 8 and 18 respondents a 6 or lower. This results in a NPS of -28.58; ((6/42)x100) – ((18/42)x100).

After wrongdoing only 2 respondents scored a 9 or 10, 11 respondents a 7 or 8 and 30 respondents a 6 or lower, which results in a NPS of -66.67 ((2/42)x100) – ((30/42)x100).

The change of NPS in the before and after condition is 66.67 – 28.58 = 38.09, to the negative side.

(35)

UvA - Master Thesis - Joni Kort - 10670920 Page 35 BRI: from the PCA it can be seen that the first 5 component have initial eigenvalues above 1.0, which together explain 67.721% of the variance (table 8). This could indicate to extract five components. However, interpreting the scree plot (figure 2) and component matrix (table 9), extraction of only three components will be sufficient.

table 8

<< figure 2

table 9>>

Run with only 3 components: the total variance explained by these three components is 52.826%, which is lower than before but still sufficient (table 10).

(36)

UvA - Master Thesis - Joni Kort - 10670920 Page 36 From the pattern matrix (table 11) it can be seen that seven item have a loading of > 0.6.

The components can be labelled according to the following themes:

-component I: communication between customers -component II: product specifics

-component III: company behavior

The next step is to determine a value for relationship quality. This is done be weighing the items means against their contribution through the following equation .

∑ 𝒘𝒆𝒊𝒈𝒉𝒕 𝒙 𝒎𝒆𝒂𝒏 𝒕𝒐𝒕𝒂𝒍 𝒘𝒆𝒊𝒈𝒉𝒕

The means are extracted in SPSS(analyze > compare means > means) and can be seen in table 12.

((.694 x 3.26) + (.855 x 1.45) + (.836 x 1.76) + (.837 x 1.71) + (.727 x 1.71) + (.639 x 2.86) + (.781 x 3.67)) / 5.369 = 2.299

<<table 11

(37)

UvA - Master Thesis - Joni Kort - 10670920 Page 37 After wrongdoing: a PCA is ran again, and the same analysis is conducted, only this time with the items from the ‘after wrongdoing survey’ (indicated with II- before the items). It can be seen that five components explain 72.538% of the total variance (table 13). From the Scree plot ( figure 3) and component matrix (table 14) it can be seen that extracting three components is sufficient.

table 13

<<figure 3

table 14 >>

The PCA is run again with only three variables extracted, still a majority of variance is explained by the three components: 58.986% (table 15).

(38)

UvA - Master Thesis - Joni Kort - 10670920 Page 38 The pattern matrix (table 16) indicates that 10 items have loadings >0.6 for this three components.

The components represent different themes in the questionnaire:

-component I: communication between customers -component II: product quality

-component III: information specifics

The next step is to determine a value for relationship quality. This is done be weighing the items means against their contribution through the following equation .

∑ 𝒘𝒆𝒊𝒈𝒉𝒕 𝒙 𝒎𝒆𝒂𝒏 𝒕𝒐𝒕𝒂𝒍 𝒘𝒆𝒊𝒈𝒉𝒕

The means are extracted and can be seen below in table 17.

((.700x1.90) + (.840x1.48) + (.865x2.93) + ( .668x1.95) + (.862x1.45) + (.821x1.60) + (.879x1.45) + (.747x1.74) + (.656x2.85) + (.810x3.43)) / 7.848 = 2.064

<< table 16

(39)

UvA - Master Thesis - Joni Kort - 10670920 Page 39 The remaining part of the analysis is done in the same way as the above analysis of the

‘normal’ advertisement. However, for the sake of readability of this thesis, the calculations, tables and figures on which the analysis and conclusions are based, can be found in appendix IV.

‘Relaxing’ advertisement, anthropomorphized

All tables, figures and calculations performed for analysis of the ‘relaxing’ advertisement can be found in appendix 10.6.1.

This group consists of 54 individuals. The NPS before wrongdoing is -22.22 ((11/54)x100) – ((23/54)x100); 11 people scored a 9 or 10, 20 respondents a 7 or 8 and 23 respondents a 6 or lower. After wrongdoing, only 3 people scored a 9 or 10, 14 respondents a 7 or 8 and 37 respondents a 6 or lower. This results is a NPS of -62.95; ((3/54)x100) – ((37/54)x100). The change in NPS for this group is; 62.95 – 22.22 = 40.73, to the negative side.

The BRI in the before wrongdoing condition is measured via a PCA. From the PCA it can be seen that extraction of three components is sufficient. The components can be labeled according the following themes respectively; communication between customers, the brand Heineken, and the brand image.

After wrongdoing, extraction of three components is again sufficient, which represent the following themes respectively; communication between customers, image of the company, and knowledge of the company.

The relationship quality value before wrongdoing is calculated as 2.35. For the after wrongdoing condition, the relationship quality value has dropped to 2.25.

(40)

UvA - Master Thesis - Joni Kort - 10670920 Page 40 ‘erotic’ advertisement, anthropomorphized

All tables, figures and calculations for the ‘erotic’ advertisement can be found in appendix 10.6.2.

This group consists of 53 respondents which were shown the ‘erotic’ advertisement. Before wrongdoing, 8 respondents scored a 9 or 10, 23 respondents a 7 or 8 and 22

respondents a 6 or lower, this results in a NPS of -26.41 ((8/53)x100) – ((22/53)x100). After wrongdoing, only 3 people scored Heineken a 9 or 10, 14 respondents scored a 7 or 8 and as many as 36 respondents scored a 6 or lower, resulting in a NPS of -62.25; ((3/53)x100) – ((36/53)x100). The change in NPS for this third group is; 62.25 – 26.41 = 35.84, to the negative side again.

The BRI is calculated again by performing a PCA. This PCA shows extracting two components is sufficient since together they explain majority of the total variance. The two components can be labelled as; communication between customers, and product specifics respectively.

After wrongdoing, extraction of three components is needed to explain majority of the total variance. These components respectively represent the following themes;

communication between customers, company behavior, and product specifics.

The relationship quality value for the ‘erotic’ advertisement drops from 2.41 in the before wrongdoing condition to 2.35 in the after wrongdoing condition.

By comparing the results for the three advertisements, it can be seen that overall, the relationship quality value decreased. This can be concluded from both the NPS and BRI analysis.

(41)

UvA - Master Thesis - Joni Kort - 10670920 Page 41 To check if the change is significant, a Paired Sample t-Test is performed. Two new variables are created in the data set to perform the test on (see figure 4).

Figure 4 table 18

table 19

The Paired Samples Test indicates that there is an almost significant change between the before wrongdoing condition (m=2.353, s=0.055) and the after wrongdoing condition (m=2.233, s=.126), t (2.882) = 2, p ≥ 0.05 ( p=0.051) (see table 18 and 19).

although not completely significant, brand wrongdoings have a negative effect on the relationship between person and brand. This means hypothesis Ia can be confirmed.

This research backs up the research of Puzakova et al. (2013) that anthropomorphized brands are more likable than non-anthropomorphized versions of the brand. This can be seen in the higher relationship quality values for the ‘relaxing’ and ‘erotic’ ads. The ad with the erotic theme is the most likable. Also, the NPS for both the ‘relaxing’ and ‘erotic’ ads are less negative, thus higher. However, this research does not completely confirm the expectation that anthropomorphized brand are held more responsible for their wrongdoings, so hypothesis Ib is not confirmed. This can be seen by measuring the changes in both the NPS and BRI. The NPS suggest the change for the ‘relaxing’ ad is the biggest to the negative side (40.73),

(42)

UvA - Master Thesis - Joni Kort - 10670920 Page 42 followed by the ‘normal’ ad (38.09). The ‘erotic’ ad is least held responsible for the

wrongdoing (35.84). From the BRI it can be concluded that relationship quality for the normal ad changed from 2.30 to 2.10, which is a decrease of 0.20. the ‘relaxing’ ad’s value changed from 2.35 to 2.25, which is a decrease of 0.10. At last, the ‘erotic’ ad’s value changed from 2.41 to 2.35, which is a decrease of 0.06.

Overall, this means that the erotic ad is the most likable and least held responsible for brand wrongdoing. This is different for the ‘relaxing’ ad, which is clearly held more

responsible for wrongdoing. This confirms that the relationship change is different for the both anthropomorphism ads, which means hypothesis Ic can be confirmed.

(43)

UvA - Master Thesis - Joni Kort - 10670920 Page 43 6.4 Testing hypothesis II

To test hypothesis II, the data will be split according character type (Data > split file > compare groups). With the TIPI method, it can be seen on which of the five dimensions of the “Big Five” test the participant scores the highest, thus which character type a participant has. An extra variable is created which states the character type for each participant, called ‘character type’, on which the data is split.

measuring every participant’s character type appeared to be not as easy as thought, since some participants score even on several of the dimensions and mostly the same

combination of two dimensions; conscientiousness and emotional stability. Therefore, a sixth category is added, which represents this combination. After splitting the file, it appears that not all groups are of equal size. Conscientiousness is the most dominant character type, with 41 participants; 34 participants score highest on extraversion; 31 on openness to experience; 19 on emotional stability; 13 on the combination of conscientiousness/ emotional stability; and only 11 participants score highest on agreeableness.

For all groups separately, the NPS and BRI are calculated just as in the former part of this thesis. First the NPS will be calculated with this equation:

NPS = ((𝑥1

𝑛) 𝑥100) − (( 𝑥2

𝑛) 𝑥100)

Thereafter, the BRI will be measured with a PCA. This is tested for suitability first by performing a Kaiser-Meyer-Olkin test (KMO) a Bartlett’s Test of Sphericity. For the two smallest groups, agreeableness and conscientiousness/emotional stability combination, SPSS errors the test, which is due to the small size of the groups. For the other four groups, the Bartlett’s Test of Sphericity is significant, since all of them scoring .000. the KMO for conscientiousness is .627, extraversion is .790 and openness to experience is 0.613. The KMO of emotional stability is only .270, which is < .5, this can be due to the small size of the group (n=19) (see table 20, 21, 22, and 23).

(44)

table 20 table 21

table 22 table 23

Also for the after wrongdoing condition, the data is checked for suitability of performing a PCA. Again, for the groups agreeableness and the combination conscientiousness/ emotional stability no results for the KMO and Bartlett’s Test of Sphericity are displayed, due to the small size of the sample. The other four groups score a .000 significance on the Bartlett’s test. For the KMO, conscientiousness has a value of .568, Extraversion scores .767 and openness to experience a .665, which are all sufficient.

Emotional stability scores a .271, which is actually too small (<.5) (table 24, 25, 26 and 27). However, since the Bartlett’s test is significant, I continue the PCA.

table 24 table 25

(45)

UvA - Master Thesis - Joni Kort - 10670920 Page 45 Further results will be discussed per group, starting with the largest group. As for testing hypothesis I, the tables, figures and calculations for testing hypothesis II can be found in appendix VI.

Conscientiousness:

Tables, figures and calculations on which the analysis for this group is based, can be found in appendix 10.6.3.

Consisting of 41 participants (n=41), this group is the largest of the six groups. The NPS before wrongdoing is -17.08 ((8/41)x100)-((15/41)x100). Eight people scored a 9 or 10, 18 people scored a 7 or 8, and 15 people scored a 6 or lower.

After wrongdoing, only two scored a 9 or 10, 11 participants a 7 or 8, and 28 a 6 or lower, which results in a net promoter score of -63,41 ((2/41)x100) – ((28/41)x100). The change of the NPS is 46.33 to the negative side (63.41-17.08).

The BRI for the before and after wrongdoing situations, and thus the potential change in relationship quality, are determined by performing a PCA. In the before wrongdoing condition, it is sufficient to extract only two components for analysis. They represent the themes; communication between customers, and product specifics. When studying the after wrongdoing condition, three components should be extracted for analysis. Here, the

components respectively represent the themes; communication between customer, brand image, and brand knowledge.

The value for relationship quality before wrongdoing is 2.489. This value drops to 2.030 for the after wrongdoing condition.

(46)

UvA - Master Thesis - Joni Kort - 10670920 Page 46 Extraversion:

Being the second largest group, this group consists of 34 participants (n=34). For the NPS, there is a decrease in the score of 44.11 (67.65-23.54). This is because the NPS before wrongdoing is -23,54 ((6/34)x100) – ((14/34)x100). Six of the participants scored a 9 or a 10, 14 a 7 or 8 and, 14 a 6 or lower.

After wrongdoing, only three people scored a 9 or 10, five a 7 or 8 and, 26 a 6 or lower, resulting in a NPS of -67,65 ((3/34)x100) – ((26/34)x100).

A PCA is performed to measure the BRI and the change in relationship quality. From the PCA it can be derived that extraction of three components in both the before and after wrongdoing condition is sufficient, because still a majority of total variance is explained in both cases. Before wrongdoing, the three components represent the following themes; communication between customers, company behavior, and product specific. In the after wrongdoing situation, the components can be labeled as; communication between customers, company behavior, and product knowledge.

The change in relationship quality values is to the negative side. The value before wrongdoing is 2.407, while 2.204 after wrongdoing.

(47)

UvA - Master Thesis - Joni Kort - 10670920 Page 47 Openness to experience:

This third group consists of 31 participants (n=31). The score in NPS decreased by 41.93 (64,51-22.58). Before wrongdoing, five participants scored a 9 or 10, 14 a 7 or 8 and, 12 a 6 or lower, resulting is a NPS of -22.58 ((5/31)x100) – ((12/31)x100).

After wrongdoing, only one person scored a 9 or 10, nine participants a 7 or 8 and, 21 a 6 or lower, resulting is a NPS of -64.51 ((1/31)x100) – ((21/31)x100).

The BRI is measured by performing a PCA. The PCA suggest extraction of three components is sufficient in the before wrongdoing condition. The components here represent respectively the themes; product specifics, company behavior, and product satisfaction. After wrongdoing, the again three components should be extracted. However, the themes of the components changes to; communication between customers, company behavior, and brand knowledge.

The relationship quality value drops from 2.562 in the before wrongdoing situation to 2.255 in the after wrongdoing condition.

(48)

UvA - Master Thesis - Joni Kort - 10670920 Page 48 Conscientiousness/emotional stability:

This group consists of respondents scoring even on the conscientiousness and emotional stability dimension and therefore are part of the combination group. The group consists of only 13 participants (n=13) and has an decrease in NPS of 33.73 (69.23-38.50). The NPS before wrongdoing is -38.50 ((2/13)x100) – ((7/13)x100), with two people scoring a 9 or 10, four participants a 7 or 8 and, seven a 6 or lower.

After wrongdoing, nobody scored a 9 or 10 anymore, again four people a 7 or 8 and, nine a 6 or lower, resulting in a NPS of -69.23 ((0/13)x100) – ((9/13)x100).

The PCA performed to measure the BRI, suggest extraction of three components in the before condition is sufficient. The three components represent respectively the themes; communication between customers, brand interest and brand knowledge. After wrongdoing, again the extraction of three components in sufficient. These components can be labeled as; communication between customers, brand knowledge, and brand interest.

The relationship quality value drops as much as 3.129 in the before wrongdoing situation to 2.410 in the after wrongdoing situation.

(49)

UvA - Master Thesis - Joni Kort - 10670920 Page 49 Emotional stability:

This group consists of 19 participants (n=19) and has a decrease in NPS of 26,36 (47.36-21.00). Before wrongdoing, four participants scored a 9 or 10, seven a 7 or 8 and, eight a 6 or lower, resulting in a NPS of -21.00 ((4/19)x100) – ((8/19)x100).

After wrongdoing only two people still scored a 9 or 10, six a 7 or 8 and, 11 a 6 or lower, resulting is an NPS of -47.36 ((2/19)x100) – ((11/19)x100).

The BRI, measured through performing a PCA, indicates extraction of three components in the before wrongdoing situation is sufficient because together they still explain majority of total variance. The three components can respectively be labeled as; communication between customers, product satisfaction, and brand knowledge. After wrongdoing, again three components should be extracted. These components represent the following themes; communication between customers, product satisfaction, and brand interest.

The relationship quality value only slightly dropped from 2.258 in the before wrongdoing condition, to 2.228 in the after wrongdoing condition.