Affecting consumers through experience
How the Online Brand Experience influences your consumers
Company Guidion Wibautstraat129-2 1091GL Amsterdam
Company supervisor Kees van Lede
0642412928 kvanlede@guidion.net
Affecting consumers through experience
How the Online Brand Experience influences your
consumers
Master Thesis
Master of Science Business Administration
Faculty of Economics and Business
University of Groningen
The Netherlands
By:
Eline Weterings
Ernest Staesstraat 9-1
1061BZ Amsterdam
Studentnr:S2139405
e-mail:elinecw@gmail.com
Supervision University of Groningen, Department of Marketing
Drs. J.H Berger
Supervision Guidion
Preface
This thesis is the last step to complete my Master of Science and Business Administration, Marketing and Management of the University of Groningen. It was a journey writing this report.
My special thanks first of all goes to my supervisor Drs. Hans Berger of the University of Groningen, for his helpful comments and constructive feedback during the writing process. Secondly, my thanks goes out to my second supervisor, Daniela Naydenova who provided me with constructive feedback.
Management summary
Since the introduction of the internet and social media, consumers are using more online media to express themselves and talk directly to organizations. With these changes taking place, another type of brand experience makes its entrance: the Online Brand Experience (OBE). It is interesting to study how the OBE influences consumers in their responses, since no specific research has ever been conducted on the very topic of the effect of the OBE upon the consumer responses; word of mouth (WOM), explicit brand attitude, purchase intention and brand awareness, when also taking a moderating effect of the brand relationship into account. Therefore, this thesis provides a piece of research on the subject. Based on an extensive literature discussion, several hypotheses were formulated, which are all captured by the following research question:
Does the Online Brand Experience influence the cognitive and behavioral consumer response and is this relation moderated by the brand relationship?
The hypotheses were tested using the customer database of the organization Guidion as well as non-customers. This study confirms that the OBE indeed influences various consumer responses and that the Brand Relationship moderates this relation. In addition, an analysis was performed on the direct effect of Brand Relationship on consumer response. Close to all hypotheses were confirmed, however a negative effect was found for the OBE upon Brand Awareness. In addition, research was conducted on the direct effect of Brand Relationship upon consumer response. All these analyses were significant and results indicated that Brand Relationship has a negative effect upon Brand Awareness as well. Other effects were positive.
Table of content
Preface 3 Management summary 4 Table of content 5 Chapter 1: Introduction 9 1.1 Research question 14Chapter 2: Theoretical framework 15
2.1 The independent and dependent variable 15
2.2 The moderator 17
Chapter 3: Hypotheses 19
3.1 Hypothesis 1: OBE and Brand Awareness 19
3.2 Hypothesis 2: OBE and WOM 19
3.3 Hypothesis 3: OBE and Purchase Intention 20
3.4 Hypothesis 4: OBE and Brand Attitude 20
3.5 Hypothesis 5: OBE and Brand Awareness and Brand Relationship 21 3.6 Hypothesis 6: OBE, Brand Attitude and Brand Relationship 21 3.7 Hypothesis 7: OBE, Purchase Intention and Brand Relationship 22
3.8 Hypothesis 8: OBE, WOM and Brand Relationship 22
3.9 Conceptual model 23
Chapter 4: Research methodology 23
4.1 Guidion 24
4.1.1 Guidion Computer Buddies 24
4.2 Data Collection procedure 25
4.5.1 Definition 29
4.5.2 Measuring scales 29
4.6 Explicit Brand Attitude 30
4.6.1 Definition 30 4.6.2 Measuring scale 30 4.7 Purchase intention 31 4.7.1 Definition 31 4.7.2 Measuring scales 32 4.8 WOM 32 4.8.1 Definition 32 4.8.2 Measuring scales 33
Chapter 5: Sample and Population 35
5.1 Missing values 35
5.2 Population 35
5.3 Size of the sample 36
5.4 Data collection 36
Chapter 6: Constructs 37
6.1 OBE 37
6.2 Brand Relationship 38
6.3 Brand Awareness 38
6.4 Explicit brand attitudes 38
6.5 Purchase intention 38
6.6 WOM 39
Chapter 7: Descriptive Data 40
7.1 Gender 40
7.2 Age 42
7.3 Education 45
Chapter 8: Requirements for Multiple Regression analysis 48
8.1 Condition 1: Independent observations 50
8.1.1 Avoiding multicolinearity 51
8.2 Condition 2: Equally divided residues 51
8.3 Condition 3: Linear Relation within Regression 52
9.1 Hypothesis 1: OBE and Brand Awareness 53
9.2 Hypothesis 2: OBE and WOM 54
9.3 Hypothesis 3: OBE and purchase intention 54
9.4 Hypothesis 4: OBE and Explicit Brand Attitude 54
9.4.1 Multiple Regression Analysis 55
9.5 Hypothesis 5: OBE, Brand Relationship, Brand Awareness 57
9.6 Hypothesis 6: OBE, Brand Relationship, Explicit Brand Attitude 57 9.7 Hypothesis 7: OBE, Brand Relationship, Purchase Intention 58
9.8 Hypothesis 8: OBE, Brand Relationship, WOM 59
9.9 Moderator analysis 60
Chapter 10: Conclusion and Discussion 61
10.1 Research question 61
10.2 General Results 61
10.2.1 Results OBE and consumer response 61
10.2.2 Results OBE, Brand Relationship and Brand Awareness 63
10.2.3 Results OBE, Brand Relationship and Purchase intention 65
10.2.4 Results OBE, Brand Relationship and WOM 66
10.2.5 Additional Regression Analyses 67
10.3 Implications for organizations 68
10.3.1 Making the OBE fun 68
10.3.2 Creating Brand Awareness 69
10.3.3 Creating WOM 69
10.3.4 Nurture the Brand Relationship 70
10.3.5 Creating an intention to purchase 71
10.3.6 Changing the consumers’ brand attitude 71
10.4 Limitations and future research 72
Appendix D 97
Appendix E 98
Appendix F 99
Chapter 1: Introduction
Social media are gaining popularity rapidly, while in 2012, more than 1 billion users of Facebook.com (Vance,2012) are active online and organizations are seeking to tap into relationship development potential. Due to the number of people using online media, information can be easily accessed and a consequence of this development is the rapidly changing marketing environment. Organizations do not longer only need to send, but also receive an extended amount of information, on which they have to anticipate. This means interaction, and because this interaction is getting more and more transparent, consumer responses are also changing. These developments are gaining importance, now that there is a shift taking place within the consumer market, concerning the use of traditional media, towards online media (Waters et al.,2009).
Since consumers have the opportunity to use online media and respond directly towards the company, another type of brand experience makes its entrance: the Online Brand Experience, in this thesis also referred to as the OBE. The Online Brand Experience can be distinguished from the regular brand experience by how it is experienced, which is different for online brands. Indeed, the consumer experience is the key distinguishing feature of an online brand (Christodoulides, 2009;Hoffman & Novak, 1996).However, there are more distinguishing factors, which are among others, illustrated by the research of Kollmann & Suckow (2008). They state that for online brands, the virtual nature of the computer-mediated environment plus the lack of physical clues, heighten the challenges of intangibility and uncertainty. Another difference is that the OBE mainly needs to nurture the continuous active relationship with the user to ensure the brands' survival, while the regular brand experience is not a continuous active relationship. (Christodoulides, 2009; Kollmann & Suckow, 2008; Song et al., 2010).
organizations desperate in terms of developing a five-year marketing plan.
Consumers are changing the rules of the game and with it, the whole marketing landscape, while managers are busy with trying to influence these conversations, for example by strengthening the brand relationship (Lipsman et al.,2012). It is because of the change in the game that organizations need to know if the OBE is affecting the consumer response and how. Is there a change within the cognitive mindset of consumers and their actual behavior? Which actions do organizations need to undertake to survive this consumer rebellion, this earthquake of changes? By researching if the OBE influences various cognitive and behavioral responses of consumers and what role the brand relationship has in this framework, this thesis will provide you with a piece of research about the questions which both the academic world as top managers ask themselves, when coping with these problems on a daily basis. An example of an organization that copes with these questions is Guidion. Guidion operates within the ICT industry, both business-to-business as business-to-consumer. It offers computer aid to small businesses within the Netherlands (MKB Nederland) and consumers. More information about this organization can be found within Chapter 4: Research methodology, Paragraph 4.1.
To be able to give an answer to these questions, it is useful to explore various sources of literature and researches on the topics of the: OBE, brand experience, consumer experience, brand relationship and several different factors within both cognitive as behavioral consumer responses. Various researches were conducted concerning the topic, however never did these studies include specific research about the OBE influencing the consumer response.
Concerning behavioral and cognitive consumer responses, the academic necessity and value is illustrated by the following examples.
The research conducted by Morgan-Thomas and Veloutsou (2013), mentions that although the results of their study support the importance of branding in online environments, that the study only includes a small subset of branding construct, which causes an extended research regarding the influence of OBE upon various consumer responses, to be necessary.
Factors that determine the customer equity are among others the brand awareness and the explicit brand attitude, both of which are also cognitive consumer responses.
The gaps in literature described above make research on the topic of the relation between the OBE and various cognitive and behavioral consumer responses, necessary and valuable. Other studies that propose further research to be required on the topic of the OBE and specific cognitive consumer responses are mentioned within the following examples.
Within the research of Bruhn et al. (2012) it is concluded that the effect of user generated social media communication on brand awareness is rather complex. Therefore, further research should be done on the topic of the OBE upon brand awareness. Concerning the explicit brand attitude, the research of Chang (2009) states that more research efforts are needed to better understand the consumer’s explicit brand attitude as a consequence of the Online Brand Experience. This is why in this thesis, both the brand awareness and the explicit brand attitude will be researched. On top of this, according to Chang (2009) there is a substantial amount of studies that require further research on the topic of the cognitive responses, for example on brand awareness and explicit brand attitude.
Concerning the relation between the OBE and factors within the behavioral consumer response, further research is required as well. The study of Harrison-Walker (2001) concludes with the issue that the role of the internet in word of mouth (or in this thesis referred to as WOM) communication is not investigated in their research. This gab opens up possibilities for further research regarding the relation between the OBE and WOM. The study of Aggarwal (2004) claims that future research should examine relationship norms more extendedly by using measures such as purchase intention. The relation between the OBE and the behavioral factors of purchase intention and WOM will therefore be further researched within this thesis.
business-to-consumer environment.
Furthermore, the research of Aggarwal (2004) mentions that research concerning the role of the brand relationship is necessary, when looking at the relation between the OBE and consumer response. The brand relationship and its role are important to take into account in this research, while it is expected that the brand relationship will moderate between the OBE and consumer response.
An important aspect worth mentioning, concerning brand relationship is that it is a concept that differs completely from the OBE by definition. As defined by Thomson, MacInnis, and Park (2005), the concept of brand relationship consists of an attachment to brands. In addition, it consists of involvement, as defined by Zaichkowsky (1985). Basically, a brand relationship can be defined as an internal consumer response, which involves an emotional connection or relationship with the brand, while the OBE is not only or necessarily only an emotional relational concept. The OBE can actually evoke emotion, thus an affective dimension is included within the concept of OBE. However, an OBE can also take place when consumers do not yet have strong emotional feelings for a brand, which is why it consists of more than only an affective dimension. In addition, earlier research by Park and MacInnis (2006) and Thomson, MacInnis, and Park (2005) found that consumers engaged in a deep relationship with the brand, do not have the most intense experience per se. These findings enforce the finding that the OBE consists of more than only the emotional aspect and therefore differs from the Brand Relationship, of which an emotional connection to the brand is the basic condition to form a Brand Relationship. In fact, all these studies conclude that the OBE and the brand relationship are not the same construct.
brand relationship could however change this relation (Morgan-Thomas and Veloutsou, 2013. Within the research of Edwards & Lambert (2007) the difference between mediation and moderation is clearly demonstrated. The general framework for mediation is the basic mediated model as shown in Figure A. Within this model there is a direct effect of X on Y and an indirect effect of X on Y mediated by M.
Within this thesis, the initial moderated model is applied for the variable ‘brand relationship’ (figure B). This model incorporates Z as a moderator (brand relationship) of the path from X (OBE) to Y (consumer response). We call this first stage moderation because, the moderating effect applies to the first stage of the indirect effect of X on Y. (Edwards & Lambert, 2007) To research whether including the brand relationship as a moderator is correct, the construct was also included in the conceptual model as a mediator.
However, the brand relationship cannot be considered a mediating variable between the OBE and consumer response, since a mediator is considered property of a construct that transforms the predictor/input variable in some way. The brand relationship does not transform the OBE in the effect it has upon consumer response. Instead, it has an effect on the relationship between the OBE and consumer response. In addition, a mediator specifies how or why a relationship occurs (Baron & Kenny, 1986), while the brand relationship does not specify how or why there is a relationship with the cognitive and behavioral response of consumers.
Moderators specify when a certain effect will occur and when the variable brand relationship is included in the conceptual model as a moderating variable, it is considered to change the relation between the OBE and consumer response as a function of brand relationship. More specifically, it accounts for the relationship predictor variable and an outcome variable. This is an expectation that matches those described in literature, as previously discussed within this chapter. Therefore, the brand relationship is considered a moderator in this thesis.
Figure A. A Basic Mediated model
M
X
Y
Figure B. First stage moderation model
Z
1.1 Research question
When looking at background studies and various sources of literature, it can be concluded that these studies discuss the topics of the brand relationship, the OBE and various cognitive and behavioral consumer responses separately. However, no specific research was found and thus ever conducted on the very topic of the relationship between these variables separately and specifically on the following consumer responses; WOM, explicit brand attitude, purchase intention and brand awareness, when also taking the moderating effect of the brand relationship into account. Therefore, this lack of research is considered a gap in literature that requires further research. All the factors mentioned are included within the following research question:
Chapter 2: Theoretical framework
There is gap in literature on providing answers to the question of how the OBE influences various consumer responses and especially regarding certain cognitive and behavioral responses, research is limited. This research will provide direct answers on the question whether the OBE influences consumer response. Furthermore, an answer is given to the question of whether the brand relationship plays a moderating role in this framework.
To be able to provide answers, in this chapter, literature is discussed concerning these concepts and their relations.
2.1 The independent and dependent variable
It is important to first look at the definition of the OBE in order to finally determine how consumers are changing the marketing landscape. As discussed earlier in this report, the OBE is defined as a concept derived from the brand experience, which is now an Online Brand Experience. Therefore, it is useful to explore theory on these subjects. On the topic of the OBE, there is a variety of definitions, such as the internet experience as defined by Herbørn & Helge, (2004), the online customer experience as described by Novak, Hoffman, & Yiu-Fai, (2000) or the total customer experience as stated by Petre et al. (2006).
However, the view of Brakus et al. (2009) on the relation between the OBE and behavioral responses, is partly in contradiction with the view as portrayed in this thesis.
2.2 The moderator
Defining the meaning of a brand relationship is an important facet when examining how the behavior and cognitions of consumers are influenced by the OBE. It is therefore useful to explore and refine existing theory on the topic. Regarding online (social) media and brand relationships, the research of Hwang & Kandampully (2012) offers an insight. Their research states that the brand relationship of consumers is similar to interpersonal relationships, because they involve exchanges between the two through repeated actions.
The quality and strength of a brand relationship can be defined by the Brand Relationship Quality model (BRQ model) designed by Fournier (1998). This model is often cited in literature and has been adapted several times. The original model includes seven aspects: intimacy, passionate attachment, love, self-concept connection (perception of a brand as part of the self), personal commitment, nostalgic connection and partner quality (taking good care of consumers). To indicate if a relationship exists between the consumer and the brand, Veloutsou (2007) conducted research on the topic and found that there are two aspects that indicate this existence: the emotional connection and communication. These concepts are linked as well, because consumers with a deep emotional bond show different responses to the brand after having an OBE. These findings support the definition of the brand relationship as defined in this thesis: “The extent to which consumers feel an emotional connection to the brand.” (Aaker, 1996; Smith et al., 2007) This definition captures the meaning of the concept of brand relationship as a whole perfectly, also with regard to the other variables researched within this thesis.
responses.
The scientists Bagozzi, Gopinath, & Nyer (1999) and Kollmann & Suckow (2008) discuss the functional benefits of brands, which can be easily duplicated sometimes, while the brand relationship provides a better basis for long-term success. Meaning, this variable is highly important when gaining success concerning consumer response. Another study finds that in online marketing, the absence of direct sales causes the brand experience to be more intimate and direct (Petre, Minocha, & Roberts, 2006). This research is stressing the importance of the difference between the brand experience and the OBE. It emphasizes on the intimate part and points out the importance of the brand relationship in having an effect on the relation between the concepts of OBE and consumer response. Finally, the research of Aaker (1996) and Smith et al. (2007) demonstrates that when consumers find that there are similarities between them and the brand, the emotional bond will grow and this will influence their responses.
Chapter 3: Hypotheses
3.1 Hypothesis 1: OBE and Brand Awareness
Research on the topic of the relation between the OBE and the cognitive consumer response of brand awareness, is very limited. A study conducted by the scientists Rice & Bennett (1998) and Okechuku (1992) discusses the brand experience and found that consumers who experience a product by usage, are more likely to remember the advertisement. This statement suggests a positive relation between the brand experience and the brand awareness. However, this research includes only the regular brand experience, meaning it is scientifically relevant to research the effect of the OBE upon brand awareness. Taking into account this research and earlier discussed research on the relation between the OBE and consumer response in general, this thesis needs to find whether this relation exists. The following hypothesis is constructed. H1: There is a positive relationship of the OBE upon brand awareness as a cognitive consumer response.
3.2 Hypothesis 2: OBE and WOM
Specific research about the relation between the OBE and WOM is limited as well. A study on the topic, includes the research of Eroglu et al. (2003).This research states that the OBE will lead to behaviors and this finding is supported by the research of Flavian et al.(2006), which concludes that the reputation of the OBE has an effect on the perceived ease of use and indirectly effects the behavioral intentions. Then there is the research of Lipsman (2012), which states that the value of a fan on social media is determined, among others by its power to influence friends by spreading WOM. Finally, Jones et al. (2009) claims that past research showed that decisions involving experiences are sensitive to WOM communication. Altogether, when looking at previous studies about the general relation between the OBE and behavioral response and decisions involving WOM, it is suggested that a positive relation between the OBE and WOM could exist, but that it was not researched before. Therefore, the
following hypothesis is formulated.
3.3 Hypothesis 3: OBE and Purchase Intention
Regarding the relation between the OBE and purchase intention, as mentioned earlier, the research of Flavian et al. (2006) concludes that the reputation of the OBE indirectly affects the behavioral intentions. Purchase intention, as WOM, is a behavioral consumer response. Other research on this relation is very limited. However, the study conducted by Koufaris (2002), claims that the representation of the product online leads to, among others, an intention to purchase. Meaning, there is a gab in literature, when it comes to research on a positive relation between the actual OBE and purchase intentions of consumers. These studies lead to the following hypothesis.
H3: There is a positive relationship of the OBE upon purchase intention as a behavioral consumer response.
3.4 Hypothesis 4: OBE and Brand Attitude
When looking at the relation between the OBE and the brand attitude, the research of Howard and Kerin (2011) concludes that results indicate that simple changes in a consumers’ subjective probability of previously seeing a brand independent of actual exposure, can have significant attitudinal consequences for that brand. This exposure however, was not researched online when the consumer is engaged in an OBE. Research conducted by a scientist who was ahead of his time, by Atwood (1969) claimed that there are two attitudes at work in every situation. These two attitudes are that towards an object and that towards the situation. A situation as such still needs to be researched for the OBE.
Finally, the research of Koufaris (2002) is useful as well, while he states that the representation of the product online leads to among others, a different attitude. This research is about the representation of a product online. Therefore, a more in-depth research on this the relation when including the OBE would be valuable as well. Taking into account all these separate studies, the following hypothesis was constructed.
3.5 Hypothesis 5: OBE and Brand Awareness and Brand Relationship
Regarding the moderating effect of the brand relationship between the variables OBE and brand awareness, the literature researched on these concepts and its relations is limited. However, there is a research conducted by Hwang & Kandampully (2012), which states that by the brand experience, consumers create brand associations, which are similar to characteristics in human relationships. These associations within the brand relationship influence among others, the consumer’s memory. This study however does not research the OBE. A study by Petty and Cacioppo (1979) and Chaiken (1980) and Krugman (1977) agree on the brand relationship having an effect on the brain, while they state on this topic, that a brand relationship increases the amount of processing. Finally, according to Zinkhan and Muderrisoglu (1985), the brand awareness is influenced indirectly by the brand relationship. None of these studies includes specific research on the OBE, nor the moderation of the brand relationship upon the OBE and brand awareness. All these findings lead to the following
hypothesis.
H5: The positive relationship of the OBE upon brand awareness is moderated by the brand relationship.
3.6 Hypothesis 6: OBE, Brand Attitude and Brand Relationship
Now, literature on the moderating effect of the brand relationship on the relation between the OBE and brand attitude is reviewed. Research on these specific relations is scarce. Lipsman (2012) states that for social media, the value of a ‘fan’ can be assessed by an increasing depth of engagement and loyalty. In addition, the research of Hwang & Kandampully (2012) concludes that by the brand experience, consumers create brand associations and these associations within the brand relationship influences among others, the consumer’s attitude. These studies imply that moderation of the brand relation between the OBE and brand attitude is possible, however it was not researched in these studies. Taking into account all previous
research, the following hypothesis is constructed.
3.7 Hypothesis 7: OBE, Purchase Intention and Brand Relationship
In this paragraph, the relation between OBE and purchase intention, and the brand relationship as a moderator is discussed. In previous findings, scientists suggest that emotional attachment to a brand has among others, strong behavioral implications (Bowlby, 1979; Park and Macinnis, 2006). The research of Morgan-Thomas & Veloutsou (2013) is a little more specific concerning the effect of the brand relationship upon the purchase intention. They found that trust has a positive effect on online purchasing and that trust is a factor embedded within a good relationship. Finally, the research of Nobre (2011), found that affective relationships were strong drivers of repurchase. Meaning, specific research on the OBE is missing, as is the effect of moderation of the brand relationship upon purchase
intention. It is therefore valuable to research if the following hypothesis can be confirmed. H7: The positive relationship of the OBE upon purchase intention is moderated by the brand
relationship.
3.8 Hypothesis 8: OBE, WOM and Brand Relationship
Now the relation between OBE and WOM, with the moderating effect of the brand relationship is discussed. As discussed earlier by Bowlby (1979) and Park and Macinnis (2006), emotional attachment to a brand has strong behavioral implications. Since WOM is a behavioral factor, this is an important piece of research. The previously discussed research of Nobre (2011) found that affective relationships were strong drivers of WOM as well. However, these studies do not include research on the topic of the OBE or the moderation of brand relationship upon WOM, leaving a gab in literature. Research on this topic, leads to the
following hypothesis.
3.9 Conceptual model
To illustrate the hypotheses mentioned, in this paragraph a visual model of the relations among the various constructs, is displayed within Figure 1. The independent variable is the OBE and the dependent variable is consumer response, which is divided into behavioral and cognitive consumer response. The moderator is the brand relationship.
Chapter 4: Research methodology
4.1 Guidion
This chapter will start with some background information on the company Guidion. Guidion is a company that operates b-to-b and b-to-c. The organization offers computer aid to small businesses within the Netherlands (MKB Nederland) and consumers. Guidion works together with partners such as XS4ALL, Telfort, UPC, ZIGGO, BCC, Apple and MediaMarkt to offer computer aid to consumers. Guidion operates with more than 500 local computer experts and is expanding rapidly. Currently it is the fastest growing business within the Netherlands. The company has 350.000 private and business customers.
Guidion made name in the business-to-business market by working together with big companies within the Netherlands, however the company is still gaining brand awareness within the consumer market, by using online tools and among others social media. This part of Guidion is called Guidion Computer Buddies.
4.1.1 Guidion Computer Buddies
Guidion Computer Buddies is a separate division within Guidion, which has 40 franchisers who offer computer aid to direct customers of the company. Other divisions within Guidion take care of the orders of XS4ALL, Telfort, UPC, ZIGGO and BCC, with local experts who are both franchisers as ZZP’ers (entrepreneurs). This research conducted for the department of Guidion Computer Buddies, which is operating as an MKB company.
Because Guidion Computer Buddies is now ready to take the step of expanding its customer base within the consumer market, it is looking for possibilities to explore the Online Brand Experience as a strategy to reach private consumers and to build a trustworthy and solid relationship with them, while influencing them in both a cognitive as well as in a behavioral way.
4.2 Data Collection procedure
Within the following paragraphs, data collection as well as the operational definitions and the indicators of the concepts within the conceptual model are explained, to guarantee the construct validity and the content validity.
To measure relations between the constructs, quantitative research was used. This form of research was used, because it is the aim of this thesis to determine the relation between the independent and the dependent variable and to reach conclusions about the population.
A sample was randomly selected from the population, in order to avoid a biased sample. Quantitative research took place in the form of a survey that was send to customers of Guidion, by including the survey within different newsletters, advertising of the survey on the Facebook page and webpage of the company and to non-customers. Non-customers are classified in this thesis as Dutch consumers who are in the possession of at least one desktop computer, or laptop with internet. Furthermore, within the survey, all statements and questions are translated to Dutch, because the sample was taken within the Netherlands. The survey was tested as a pilot among eight Dutch consumers for clarity and correct translation by the technique of ‘back translation’ studied by Douglas and Craig (2007), before publication on the internet and including it in newsletters. The original survey can be found in Appendix A.
4.3 OBE
4.3.1 Definition
The operational definition of the OBE is: “The extent to which consumers experience positive feelings, sensations and cognitions as an internal and subjective response to contact with the company” (Brakus et al, 2009). The indicator is: “In how far consumer experience positive cognitive thoughts”.
4.3.2 Measuring scales
Throughout the years, various scales were constructed to measure the OBE, such as the visual product aesthetics scale by Bloch, Brunel, and Arnold (2003). Other examples are the emotions scale by Izard (1978), or the need for- cognition scale by Cacioppo and Petty (1982). These scales focus on individual reactions to specific stimuli (well-designed products) or tasks (abstract thinking). They use multifaceted items as used by Bloch, Brunel, and Arnold (2003) to measure specific sensations (“I enjoy seeing displays of products that have superior design”). Also, specific emotions such as affectionate, attached and passionate as used by Thomson, MacInnis, and Park (2005) or specific intellectual processes (“The notion of thinking abstractly appeals to me”) designed by Cacioppo and Petty (1982) were used. Although reviewing such scales is important, the specific items on these scales are of limited value to a more general measurement of the degree of a sensory, intellectual and affective experience. These scales are of limited value, because the scales mentioned measure other constructs than sensory, intellectual and affective experience, while the definition of the OBE as in this thesis, is not in accordance with the definitions in these studies. This would cause using their measuring scale, to be invalid when applying it within this research. Therefore, the research of Brakus et al. (2009) and their constructed measuring scale is used to measure the construct of Online Brand Experience.
scale is short and easy to administer, and it is internally consistent and reliable.
A scale, which does not include the behavioral construct was not researched and tested.
However, in this thesis the behavioral items of this scale are not included to research the brand experience. Only the internal outcomes of the brand experience are included. Since Brakus et al (2009) found internal reliability between the affective and sensory constructs solely, it is expected that not including behavioral responses within the measuring scale, will not influence the CA, which would keep the scale reliable. However, not taking into account the behavioral responses will influence the result of this thesis’ research in the sense that it will not measure what the model of Brakus et al. (2009) measured. Meaning, a different measuring scale is applied, because of the definition of the OBE within this thesis.
As explained earlier, the definition in this research includes only internal affective, sensory and intellectual responses as being part of the OBE, while different researches support this definition such as the research ofSautter et al. (2004) by stating that brand experience only includes affective, emotional and intellectual aspects, which will lead to behavioral implications. The research of Dewey, Dubé and LeBel (2003) also support behavioral aspects not being part of the brand experience and describe four “pleasure dimensions” in their research: intellectual, emotional, social, and physical pleasures. These dimensions are defined as affective or implicit emotional responses. Finally, the research of Pinker (1997) identifies four mental modules that correspond closely to the experiences that Dewey (1922, 1925) and the research of Brakus et al. (2009) describes, which is a sensory perception, feelings and emotions, creativity and reasoning and social relationships. However, Pinker does not list a separate “doing” module; he considers bodily experiences, motor actions and behaviors part of the “sensory-motor module,” similar to Lakoff and Johnson’s (1999) concept of embodied cognition.
4.4 Brand relationship 4.4.1Definition
The operational definition of brand relationships is: “The extent to which consumers feel an emotional connection to the brand”. This emotional connection develops into a relationship that involves reciprocal exchanges between partners through a series of repeated actions. Also, a relationship provides benefits to the participants, such as perceived commitment (Aaker, 1996; Smith et al., 2007). According to literature, the indicator is: ´The consumers´ specific feelings toward a brand, concerning their affection, passion and connection`
(Thomson et al.,2005). 4.4.2. Measuring scales
People can form emotional attachments to a variety of objects, including brands (McAlexander and Schouten &,1995).The notion that such attachments reflect an emotional bond is also suggested by research in consumer behavior (Shimp & Madden, 1988). For example, Slater (2000) found that a variety of emotions (love, warm feelings) describe the collectors’ emotional attachments to Coke and Hallmark. Based on the literature it was found that emotional attachment consists of: person attachments (Ball & Tasaki, 1992; Bowlby,1979; Brennan et al., 1998, Collins & Read,1994;, object attachments (Kleine, Kleine, & Allen, 1995; Richins, 1994) and love (Aron & Westbay, 1996, Hazan & Shaver, 1994; Hendrick & Hendrick, 1989 and Sternberg, 1987).
The study by Brakus et al. (2009) however, suits the definition of the brand relationship consisting of an emotional connection perfectly. These scientists found that emotional attachment can be distinguished from other constructs such as brand attitude favorability, satisfaction, and involvement. Also, this research displays a clear difference between the OBE and the brand relationship.
4.5 Brand Awareness 4.5.1 Definition
Brand awareness consists of two aspects, the recognition and the recall (top of mind). The top of mind awareness shows whether the consumer will remember the brand at the point of purchase and choose it over other brands (Miller & Berry,1998). Brand recognition measures whether consumers (customers and non-customers) recognize the company’s logo, name and the product.
Extended research is conducted on brand awareness, including the study of Romaniuk and Wight (2009) that shows that for all advertising awareness measures, users who remember the brand experience have a higher propensity to remember the advertisements than non-users who never experienced the product. However, apart from these studies no research was conducted on the brand awareness as an outcome of the OBE.
The brand awareness can be defined as “The conscious effect of advertising exposure on the consumer.”(du Plessis,1994). This definition is consistent among different researches on the topic and therefore fits as a good description of the construct. One indicator of brand awareness is the `brand recognition` and another is `brand recall`. The brand recognition is defined as “Recognition of the logo/name/product by the consumers” when they see those of multiple brands. The brand recall is defined as: “In how far the brand is at the top-of-mind of consumers. “ (Romaniuk and Wight, 2009).
4.5.2 Measuring scales
The research of Romaniuk and Wight (2009) describes three measures which are used
to assess advertising effectiveness across the whole market and between brands. These measures, which are designed to measure brand recognition and brand recall consists of: (1) Top-of-mind advertising awareness, or the first brand retrieved/ brand recall;
(2) Total unprompted advertising awareness, that part of the market that recalled the brand as advertising without prompting;
How this research was conducted can be found within the study of Romaniuk and Wight (2009). Because the survey for this thesis was included within the monthly newsletter to customers of the organization, the top-of mind awareness cannot be measured within this thesis’ research. In addition, the category of unprompted advertising cannot be measured, because of this reason. The consumers will recognize the logo and/or name of the organization within the newsletter and might be influenced by its display. However, the variable ‘prompted’ can still be measured, because the respondents will actually be prompted. The variable brand awareness is measured by asking the respondent if they remember exposure to the brand when given both the name of Guidion Computer Buddies and the industry it is operating in: the ICT. The consequence of not using the variables unprompted and top-of-mind awareness is that for the brand awareness, only one indicator will be measured, which is brand recognition, this will influence the results of this thesis research as well. The measuring model used for the question within the survey, can be found in Table 3, in Appendix C.
4.6 Explicit Brand Attitude 4.6.1 Definition
The explicit brand attitude is defined as the feelings consumers have when purchasing a product or service and their attitude toward the brand. The explicit attitude is in contrast with the implicit attitude, because an explicit attitude is the conscious process within the mind of consumers, while the implicit attitude is an unconscious process. Earlier research on the topic touches another difference, which is between the consumer attitude and the brand experience. In this thesis, the explicit brand attitude is considered part of the cognitive consumer response. The explicit brand attitude is defined as: “The conscious process in the mind of the consumer, which concerns the feelings related to an organization and in how far consumers like the organization and their marketing” (Summers, 1970 and Fennis & Stroebe, 2010). This is a definition, which covers the concept of brand attitude completely. The indicator of explicit brand attitude is: “In how far the organization and its marketing efforts are consciously liked by consumers”. (Summers, 1970)
The brand attitude was measured among consumers by asking about their feelings related to the brand in the research of Fennis & Stroebe (2010). When a service is purchased and when visiting the websites, certain feelings are evoked. Statements on a scale of 1 to 10 measured these feelings. These statements measure in how far consumers experience positive feelings towards the brand. In the research of Redker & Gibson (2009), a group of 30 participants in an introductory psychology class evaluated 30 pieces of country music to assess the explicit brand attitude. After each advertisement was presented, the participants rated their preference for the product on four Likert-type scales. The participants first responded on 7-point Likert-type scales to three general statements regarding country music.
The research of Howard and Kerin (2011) supports this measurement scale, used by Redker & Gibson (2009) and by Fennis & Stroebe (2010). In their study, the explicit attitude was also measured at a 7-point Likert scale, by the statement: ‘I personally feel that the products sold at [retailer] are: good–bad, undesirable–desirable, beneficial–not beneficial’. The statements used to capture the measurement for this thesis’ research on the construct of explicit brand attitude are based on the research and proven measuring scale of Redker & Gibson (2009). This measuring scale is used within this thesis, because these scientists provided the measuring scale within their research, where as other researches solely describe their methods, without including the actual measuring scale. In addition, this measuring scale suits the definition of the explicit brand attitude perfectly, as discussed earlier within this thesis. Meaning, the measuring scale is considered a valid measuring instrument. The survey questions and their translation, to measure the construct of explicit brand attitude can be found in Table 4, in Appendix C.
4.7 Purchase intention 4.7.1 Definition
E
xisting literature on the topic of purchase intention found that consumers are using online media more and more and are reducing their reliance on sources such as traditional advertising to make their purchase decisions (Mangold and Faulds, 2009). The purchase intention is operationally defined as: “In how far it is likely that consumers will purchase the services of an organization”(Andrews et al. 1992; Beerli and Santana 1999).4.7.2 Measuring scales
The purchase definition can be measured by asking the respondents how likely it is that they will purchase the services. Participants will evaluate an advertisement or product and then their intention to buy is tested. This method is often used within literature (Andrews et al. 1992; Beerli and Santana 1999).The research of Andrews et al. (1992) provides a fine measuring scale, designed to measure exactly the definition of the purchase intention as displayed in this thesis, described above. By using the measuring scale of these scientists, this research will be valid for the concept of purchase intention, because it fits the definition perfectly. The concept is measured on a scale anchored by seven-point, semantic differential (7)"Extremely likely" to "Extremely unlikely." (1) How this scale is translated to this research and in Dutch is showed in Table 5, in Appendix C.
4.8 WOM
4.8.1 Definition
The way WOM is referred to is: “user generated content”, which is a part of consumer- to- consumer communication. Within the research of Tirunnilai and Tellis (2011) it was found that volume of chatter shows the strongest relationship with returns and trading volume.
WOM is an important outcome of the OBE and brand relationship, because consumers are turning away from traditional sources such as printed media, to collect information.
4.8.2 Measuring scales
WOM refers to interpersonal communication or simply an exchange of information between individuals (King and Summers 1970). The research of Harrison-Walker (2001) researches the various ways to measure WOM. For example, Swan and Oliver (1989) used a single item Likert-type scale to assess the favorableness of WOM communications, following the purchase of a new car. New automobile buyers were asked, “Did you tell (the message receiver) mostly positive or mostly negative things about the car?”
Singh (1990) measured unfavorable
WOM using a single-item dichotomous scale (“told my friends and relatives about my bad experience”). Richins (1983), who focused on unfavorable WOM as a response to dissatisfaction, measured WOM communication as “the act of telling at least one friend or acquaintance about the dissatisfaction.” File, Cermak, and Prince (1994) focused their conceptualization on intentions.
Chapter 5: Sample and Population
5.1 Missing values
Since not all 233 respondents answered every question, it is firstly important to determine which procedure for missing values needs to be used. According to Marsh (1998), the most widely used approaches are listwise deletion (eliminating all cases with any missing data), pairwise deletion (computing each covariance on the basis of all cases with non-missing values for that particular pair of variables), and various imputation procedures whereby missing values are replaced by some estimate of the missing value. In this thesis, the method of pairwise deletion will be used to save more data, since listwise deletion results in a substantially reduced sample size and less efficient estimation. Multiple imputation will not be used, because it is most suited for only some limited situations where there is a small number of discrete patterns of missing data (Kaplan, 1995; Muthen et al., 1987) and this is not the case for the data collected for this thesis’ research.
5.2 Population
5.3 Size of the sample
In order to collect a representative sample, it is important that an appropriate sample size is calculated based on the population. By distributing the survey through various media, the total response was N=233. When looking at other comparable research on the topic it is found that this is a reliable number of respondents. An example of comparable is research is that of Veloutsou (2007). In his research, it is concluded that by having a total of 277 usable responses, the sample size is sufficient to perform this analysis, since it is higher than the adequate 15:1 ratio of respondents to items (Hulin, Cudeck, Netemeyer, Dillon and McDonald 2001, p. 58). Another example of comparable research is that by Batra, Ahuvia, & Bagozzi (2012) in which N=187, for their research on brand love. Within the research of Wang and Head (2007), N=167, which was also found to be sufficient.
5.4 Data collection
Chapter 6: Constructs
To test the different constructs, at first internal reliability was analyzed among questions measuring the same construct. Which questions were used to form per construct, can be found in Appendix C. Internal reliability is assessed by the Cronbach’s alpha (CA). When the CA is sufficient, questions measuring the same construct can be transformed into one construct. A widely advocated level of adequacy for CA stated by Nunally (1978) is .70. It is because of Nunally’s 1978, text on psychometric theory, which is quoted extensively over the years in various studies, that this CA is widely adopted. In his research, he points out that for early stages of basic research, reliabilities of .50 to .60 suffice and that increasing reliabilities beyond .80 is probably wasteful. Because the constructs researched within this thesis have been included into various studies before, the cut off for the CA is set at 0.7, to ensure a high enough standard for internal reliability. An overview of the CA’s measured among the different constructs can be found below in Table 1 below.
Table 1: Reliability of constructs
Reliability test CA>.06 CA<.06 CA if item is
deleted
OBE .631 -q3 = .726
Brand Relationship .956
Brand Awareness .04
Explicit Brand Attitude .776
WOM .938
Purchase intention Not applicable Not
applicable 6.1 OBE
measuring the construct will be used to form a construct, which is called OBE. 6.2 Brand Relationship
To measure the brand relationship, 3 variables were measured: connection, affection and passion. These variables included statements on which respondents had to answer in how far the description of the word was applicable to Guidion Computer Buddies. Within earlier research, a reflective scale was designed and used to measure the brand relationship, which is why the observed variables should correlate, because each variable is related to the brand relationship. To find if all statements on the construct brand relationship could be transformed into one construct, the CA needs to measure a minimum of 0.7. For all items, the CA=.956. Therefore, the statements were transformed into one construct, which is called the Brand relationship.
6.3 Brand Awareness
To measure brand awareness within this thesis, 1 question was used, which is referred to as Brand Awareness PROMPTED. In earlier research, it was measured on a reflective scale. However, the measuring scale used in this thesis, is not exactly the same measuring scale as used by Romaniuk and Wight (2009). Using their measuring scale in the same way was not possible, since consumers could not be tested for brand recognition, because the survey was send by a standard e-mail template from Guidion Computer Buddies. This template displayed the logo. In addition, this e-mail was send to current customers. Since only one question is included within the survey to measure brand awareness, the CA does not have to be analyzed. 6.4 Explicit brand attitudes
To measure the construct of explicit brand attitude, four questions were asked. This construct was also measured as a reflective scale. To find out whether there was internal reliability between these 4 questions, the CA was analyzed and measured α=.776, which is more than sufficient in order to form a construct. This construct variable is called Explicit Brand Attitude.
6.5 Purchase intention
6.6 WOM
Chapter 7: Descriptive Data
To find out who the respondents are and if they meet the requirements of the population, the variables: gender, age and level of education were included within the survey. These are the parameters of the population. To find whether the sample is representative, these parameters were standardized. Afterwards, a Chi2 test was performed with categorical variables, to find whether there are differences between the expected and observed cells. The equation for this test is
7.1 Gender
This test was first performed on the gender of the respondents. Concerning gender, this thesis’ research sample consists of 88 males, 85 females and 60 respondents of an unknown sex, as displayed by Table 2 below .
The percentages in census refer to Dutch consumers in general, since more than 99% of all Dutch consumers are in the possession of at least one computer, laptop or desktop with internet, according to the CBS (2012).
Table 2: Sample and census gender of Dutch consumers
Gender Number in Sample Percentages in census
Male 88 49,5%
Female 85 50,5%
Total 173 100%
Using the data within table 2, a Chi Square Goodness Of T-test was conducted, to determine if the sample provides a good match to the known division of gender of Dutch consumers. The frequencies of occurrence of each gender are given in the middle column and are the observed values Oi, and there are k = 2 categories into which the data has been grouped so that i = 1,2. If the representativeness of gender is presented as hypotheses, it is presented like this:
H0: The gender of respondents in the sample is the same as the gender distribution of Dutch consumers, based on the census.
To proof H0, the equation : will be filled in with the values from Table 2,where the summation proceeds across all k categories and there are 2-1=1 degree of freedom.
As can be found in Table 2, the sum of this sample is 173. Males are category i=1 and females category i=2. The group of males contains 49, 5% of the consumers and if the null hypothesis were to be exactly correct, there would be 49, 5% of the 173 cases in category 1. This is: . Both the observed and the expected values, as well as calculations for the statistic are given as in Table 3 for both males and females. The difference between the observed and expected numbers of cases is given in the column labeled Oi- Ei, with the squares of these in the column to the right. Finally, these squares of the observed minus expected numbers of cases are divided by the expected number of cases, which is given in the last column. The Chi2 value is also displayed wit in the last column.
Table 3: Goodness of Fit Calculations for Sample of Dutch consumer’s gender Category
1:Male 88 85.63 2.37 5.62 .065
2:Female 85 87.36 -2.36 5.57 .063
Total 173 172.99 0.0
7.2 Age
Concerning age, the mean age of the population was 53,4 years, with a standard deviation of 18,3 years. The mode age of all respondents was 23 years old. Within the survey, 77 respondents did not include their age.
Concluded from Figure 2 below is that residues for age are normally distributed. Meaning, there is no sampling error. Even though the histogram shows a bell shaped line, there are some outliers. For this reason, conducting a normal probability plot(also called Lilliefors test (Lilliefors,1967)) is recommended, as described by Gujarati (1999). This plot is displayed in Figure 3. For the probability plot in Figure 3, the dots are on a straight line, meaning the normality assumption of the residues being normally divided is accepted for the variable age.
Figure 2: Age histogram Figure 3: Normal Probability Plot
Table 4: Sample and census age of Dutch consumers
Age Number in Sample Percentages in census
20-25 20 8,38% 26-30 7 8,07% 31-35 7 8,06% 36-40 1 8,56% 41-45 9 10,29% 46-50 18 10,36% 51-55 15 9,72% 56-60 16 8,82% 61-65 20 8,55% 66-70 18 6,98% 71-75 10 5,19% 76-80 6 4,04% 81-85 9 2,94% Total 156 100%
Using the data within Table 4, a Chi square Goodness of T-test is conducted, to determine if the sample provides a good match to the known division of ages of Dutch consumers. A significance level of .05 is used (Hair et al, 2010).The frequencies of occurrence of each age are given in the middle column and are the observed values Oi, and there are k = 13 categories into which the data has been grouped so that i = 1until 13.
The hypotheses to find whether the sample is representative concerning age are the following: H0: The age of respondents in the sample is the same as the age distribution of Dutch consumers, based on the census.
H1: The age distribution of respondents in the sample differs from the age distribution of Dutch consumers in the census.
Table 5: Goodness of Fit Calculations for Sample of Dutch consumers on age Category 1:20-25 20 13,07 7 48 3,67 2:26-30 7 12,59 -6 31 2,48 3:31-35 7 12,57 -6 31 2,47 4:36-40 1 13,35 -12 153 11,42 5:41-45 9 16,05 -7 50 3,10 6:46-50 18 16,16 2 3 0,21 7:51-55 15 15,16 0 0 0,00 8:56-60 16 13,33 2 5 0,36 9:61-65 20 13,34 7 44 3,33 10:66-70 18 10,89 7 51 4,64 11:71-75 10 8,10 2 4 0,45 12:76-80 6 6,30 0 0 0,01 13:81-85 9 3 4 19 4,25 Total 156 155,94 0.0 36, 41
The level of significance used is .05(Hair et al, 2010). The number of degrees of freedom is the number of categories minus one, so d=k-1, which means 13-1 is 12 degrees of freedom. From the .05 column and the 12th row of the table in Appendix E, the critical value is
7.3 Education
Concerning education, 171 respondents included their level of schooling within the survey. The sample includes 62 missing values. With 6%, most respondents finished a higher level of education (HBO within the Netherlands), while 15 % has an MBO background and 7,7% finished high school. 18% had a university degree or higher. A normal probability plot was conducted and as shown by Figure 4, the residues are normally distributed.
Figure 4: Normal probability plot of Education
The percentages in census as described in Table 6 below, refers to Dutch consumers in general, since more than 99% of all Dutch consumers are in the possession of at least one computer, laptop or desktop with internet, according to the CBS (2012). Unfortunately, the CBS formulation of education is somewhat different from this thesis’ division of categories, while the CBS includes both the level HAVO, which is given in High school into the same category as MBO, which is further education. In addition, HBO and WO are included into one category. Therefore, the division of education is categorized in the same manner, by forming two new constructs consisting of Highschool+MBO and HBO+WO level.
Table 6: Sample and census education of Dutch consumers
Using the data within Table 6, a Chi square Goodness of T-test is conducted, to determine if the sample provides a good match to the known division of education among Dutch consumers, using a significance level of .05 as set by (Hair et al .,2010) and k = 2 categories. The following hypotheses are formulated to test if the sample is representative on education. H0: The education of respondents in the sample is the same as the education distribution of Dutch consumers, based on the census.
H1: The education distribution of respondents in the sample differs from the education distribution of Dutch consumers in the census.
As can be found in Table 7, the sum of this sample is 171. Highschool+MBO are category i=1 and HBO+WO category i=2. The first group contains 68,4% the consumers, and if the null hypothesis were to be exactly correct, there would be 68,4% of the 171 cases in category 1. This is: E1= 171 * (68,4/100)= 171 *.684= 116,96
. To proof H0, again the equation: will be filled in with the values from Table 6. Both the observed and the expected values, as well as calculations for the statistic are given in Table 7 below.
Table 7: Goodness of Fit Calculations for Sample of Dutch consumers on education Category 1:Highschool+MBO 53 116,96 -63,96 4090,88 34,98 2:HBO+WO 118 54,02 63,98 4093,44 75,78 Total 171 171 0.0 110,76
The level of significance used is .05. The number of degrees of freedom is the number of categories minus one, which means 2-1 = 1 degree of freedom. From the .05 column and the first row of the table in Appendix E, the critical value is =3,841.
Chapter 8: Requirements for Multiple Regression analysis
Both the independent variable as the moderator ’Brand Relationship’ were measured on a quasi-interval scale, meaning both the predictor as the moderator are continuous. All dependent variables were measured on a quasi-interval or interval level, except for Brand Awareness PROMPTED, which is measured on a nominal scale. Therefore, all these variables can be included within a Regression Analysis. The relation between the variables, is analyzed separately for each construct, while the dependent variables do not correlate significantly (with a CA of.569), to measure consumer response as a whole construct. To perform the correct analyses, it is important to take notice of the fact that the cognitive and behavioral variables within the consumer response are considered continuous. It is presumed that the effect of the independent variable (X) on the dependent variable (Y) varies linearly with respect to the moderator (Z). This will be tested by adding the product of the moderator and the independent variable to the regression equation, as described by Cohen and Cohen (2003) and Kessler & Cleary (1981), this product is called the interaction effect. Because this thesis describes relations including a moderating variable, a multiple Regression Analysis needs to be performed. When Multiple Regression includes two quantitative independent variables (both the moderator as the predictor), in foremost all studies, scientists will divide the quantitative variables into categories (for example: high versus low). An example of such a research is that of Irwin& McClelland (2001).However in general this practice is not recommended according to Cohen et al (2003), which is why it will not be applied within this thesis. Instead, in order to use analysis of variance (ANOVA) and interpret the effect of interaction, the predictor (OBE) and moderator (Brand Relationship) are centered (Voeten & van Bercken,2004). This method is also in accordance with the research of Kenny &
Following the conclusions of Anscombe,(1973); Chatterjee & Hadi(2006) and Dohoo & Martin (2003), to be able to perform a regression analysis, the following requirements apply:
1. The sample needs to consist of independent observations 2. The residues should be equally divided
3. The variables should have a linear relation
8.1 Condition 1: Independent observations
The respondents within the sample should be drawn independently. This expectation is called the independence assumption. If the variables are dependent, the estimated standard errors will be smaller than they are in reality, which causes the regression coefficients estimated, to be a lot less accurate, meaning they will be biased (Siero, Huisman and Kiers, 2009). To test whether the dependent variables are independently drawn with regard to the predictor and moderator, a Chi2 analysis, with the following equation: is performed within SPSS. The results are displayed in Table 8.
Table 8: Chi2 analysis of independence
Variable Pearson Chi2
value
P-value
WOM 1644.806 .039
Purchase Intention 234.944 .470
Brand Awareness PROMPTED 57.620 .028
Explicit Brand Attitude 970.583 .044
8.1.1 Avoiding multicolinearity
Multicolinearity means that two independent variables (Brand Relationship and OBE) have a high correlation score, which implies that almost the same variance is explained on the dependent variables, causing discrimination between the two to be difficult. (Cohen, 2003 ;Marquardt, 1980).To find whether the moderator and predictor constructs are independent with respect to each other, a Pearson correlation test was performed. The results shown in Table 9 give a high correlation score, indicating multicolinearity issues.
Table 9: Pearson Correlation table Pearson Correlation Table OBE
Brand Relationship .000
To find if multicolinearity is a problem, a Regression Analysis is performed. To exclude multicolinearity, the tolerance indicator needs to be >0.1, and the VIF values <10, while 1/ tolerance = VIF (Wei ,2009). The results of the Regression analysis are displayed in Table 10 below. For all dependent variables, this same output was generated.
Table 10: VIF and tolerance scores on multicolinearity Construct/Variable(standardized) BetatSig Brand Relationship Tolerance score Brand Relationship VIF score OBE .556 1.798
When tested on the dependent variables Brand Awareness Prompted, Purchase intention, WOM and Explicit Brand Attitude, all results indicated there is no multicolinearity between the OBE and Brand Relationship.
8.2 Condition 2: Equally divided residues
