Making the jump, from liking to loving
Drivers and effects of brand love
Making the jump, from liking to loving
Drivers and effects of brand love
Stephan Liebregt Student no. 1582976 University of Groningen Faculty of Economics & Business
Msc Business Administration
Specialization Marketing Management & Research Based Marketing
Gelijkheid 8 8014 XC Zwolle Phone: +31 614493504 e-mail: stephanliebregt@hotmail.com
1st Master Thesis Supervisor : dr. J.E. (Jaap) Wieringa
Management Summary
Customer satisfaction has always been one of the most important indicators for a company. A high level of customer satisfaction results in higher profits and more loyal customers. In the last decade consumers are starting to use brands more and more to express their own personalities. In addition, more advanced marketing and production tools are available that make it possible to have personalized marketing actions and products. This creates new ties between brands and their consumers. Hence researchers started to look for other constructs than satisfaction to predict strategic consumer behavior. One highly noted stream of research, that tries to uncover the new ties that bind consumers, is the one on consumers love for possessions and brands.
This report uses interpersonal love as the basis to explain love for possessions and brands. The triangular love theory of Sternberg (1986) is used in pursuit of understanding strategic behavior of the modern consumer. This theory claims that love consists of three first order factors: Intimacy, Passion and Commitment. Based on this theory and recent research on brand love or close related constructs a brand love scale is developed containing the three elements: Intimacy, Passion and Commitment. Hence brand love is defined as the degree of intimacy, passion and commitment a consumer has for a specific brand or trade name.
Three potential drivers of brand love are selected based on their presence in current literature: Satisfaction, Brand-self connection and Hedonic value. In addition, three potential effects that brand love has on consumer behavior are examined: Positive word of mouth, Brand loyalty and Spending resources. The resulting conceptual model has nine latent constructs (three drivers, three effect and the three components of brand love) and six hypothesis (three drivers of brand love and three effects from brand love).
The nine latent constructs are transferred into a questionnaire with thirty-two questions. The questionnaire was distributed via e-mail to different age groups (e.g. students, working adults, seniors). In total two-hundred and two respondents filled in the questionnaire correctly. Using the six stage design from Hair et. al (2006) a research design was developed to test the constructs, the measurement and the structural model. The program LISREL was used to conduct the analysis.
The scales showed moderate validity and some alteration were needed to end up with an acceptable measurement model. Next the model as a whole, including the drivers and effects, was tested. The brand love construct was treated in two ways: As an one order factor containing all the items from the constructs, Intimacy, Passion and Commitment and as a second order factor. The model with brand love as an one order factor showed better fit statistics.
has the most influence on brand love. For marketers this is an indication that focusing on how the products enriches the consumers life instead of the showing how the product offers general fun could increase brand love for their products. Furthermore, brand love showed a significant positive influence on all three drivers. Positive word of mouth was influenced the most by brand love, followed by loyalty and spending resources. The results suggest that marketers should strongly consider adding brand love to their cockpit of measured items to predicts consumer behavior.
Preface
Writing this thesis was a journey with ups and downs and I could not have done it without the ongoing support of my friends and family. Especially my girlfriend Ine-Marije and my parents gave me energy to finish my thesis. On a professional level I would like to thank Jaap Wieringa and Karel-Jan Alsem for their role as supervisors and Aaron Ahuvia and Hean Tat Keh for sending useful articles and (un)published work of their own.
This journey has come to an end.
5
Table of Content
1.
Introduction
1.1 Background 6
2.
Theoretical Framework
2.1 From interpersonal love to brand love 10
2.2 Brand love, satisfaction, brand liking, brand involvement and delight 17
2.3 Drivers of brand love 19
2.4 Effects of brand love 22
2.5 Conceptual model 24
3.
Research Design
3.1 Defining individual constructs 26
3.1.1 Construct validity 28
3.1.2 Discriminant validity 28
3.2 Development and specification of the measurement model 29 3.3 Study design and data collection to product empirical results 29
3.4 Model validity of the measurement model 30
3.5 Development and specification of the structural model 32
3.6 Model validity of the structural model 32
4.
Results
4.1 Data and subjects 33
4.2 Individual construct scores 33
4.3 Discriminant validity 36
4.3.1 Drivers of brand love 36
4.3.2 Components of brand love 36
4.3.3 Effects of brand love 36
4.4 The measurement model 38
4.5 The structural model 40
5.
Conclusions and recommendations
5.1 Brand love construct 43
5.2 Drivers of brand love 43
5.3 Effects of brand love 44
5.4 Limitations of the study 44
6
1.
Introduction
1.1 Background
Try to imagine a world without brands. For today’s consumer this is an impossible assignment. Everything is branded nowadays from clothing to fruit and from furniture to paper. An important reason for brands to exist is the possibility to create a loyal customer base. Loyal customers are a real benefit to a company, e.g. they guarantee profit and create positive word of mouth (WOM+) (Helgesen, 2006). Hence the consumer brand relationship is extensively studied in literature (Fournier, 1998). Until recently the most important link between consumers and loyalty was the level of satisfaction that the consumer achieved when consuming the brand. This link made satisfaction a very dominant subject in scientific literature on loyalty. A lot of models have been developed that try to explain the level of satisfaction that the consumer reaches and to give insight in how satisfaction is developed. Models such as the Expectation-Disconfirmation (ED) model (Cadotte, Woodruff and Jenkins, 1987), the desired model (Spreng, Mackenzie and Olshavsky, 1996), and the experienced-based norms model (Woodruff, Cadotte, and Jenkins, 1983) are widely used in market research. In addition, a lot of research has shown a positive link between an increase in satisfaction scores and firm performance (Anderson, Fornell, and Lehman, 1994), explaining the continuing popularity of the satisfaction score. Today’s market, however, is changing rapidly because products are not evaluated solely on their utility. Brands are also used to signal a way of living thereby creating the need to go beyond satisfying the customer (Carroll and Ahuvia, 2006). Recently researchers started to look for reasons beyond satisfaction, and the CS-paradigm, to explain differences in strategic customer behavior (Chaudhuri and Holbrook, 2001; Fournier and Mick, 1999; Fournier, 1998; Rust and Varki, 1997).
7 consumer brand relationship. Although results are significant the models suffer from multicolinearity and are hard to interpret (Breivik and Thorbjørnsen, 2008).
Finding its origin in interpersonal love, the construct brand love is another candidate for a better understanding of the new ties that bind consumers to their brands (Albert et al., 2010; Batra et al., 2008; Albert et al., 2007; Keh et al., 2007; Carroll and Ahuvia, 2006; Thomson et al., 2005; Whang et al., 2004; Fournier and Mick, 1999; Shimp and Madden, 1988). This construct appears to be accustomed to account for the (long term) emotional bonding between the consumer and the brand which is becoming increasingly important (Fournier, 1998). Research on the effects of brand love shows a number of important positive relations: a positive and significant relation between brand love and loyalty (Albert et al., 2010; Keh et al., 2007; Carroll and Ahuvia, 2006), a positive and significant relation between brand love and a willingness to spend resources (e.g. time and money) (Thomson et al., 2005) and a significant relationship with positive word of mouth (Albert et al., 2010; Carroll and Ahuvia, 2006). Not surprisingly, there is a growing interest on love in a consumption context among consumer researchers and practitioners (Roberts, 2006). Research on the topic of brand love is however, still in its infancy and therefore mostly focused on the discovery of what love means in a consumer brand context (Albert et al., 2010; Keh et al., 2007; Whang et al., 2004; Shimp and Madden, 1988). One prominent source in research on brand love are interpersonal love theories. The triangular theory on love of Sternberg (1986) and the love styles of Lee (1988) are adapted in research trying to understand brand love and for the development of a scale to measure brand love. Although some researchers claim that using only interpersonal love as the basis fails to capture the nuances that direct consumer-object love research can obtain, (Albert et al., 2007; Ahuvia, 2005) the results are clear and indicate a significant amount of overlap between interpersonal love and brand love (Albert et al., 2010; Kamat and Parulekar, 2007; Keh et al., 2007; Whang et al., 2004).
Drivers of brand love are less studied, although some research has been done (Yim et al., 2008; Carroll and Ahuvia, 2006). Research on the drivers of brand love is a logical step in the process of understanding brand love, and has been advised by a number of authors (Albert et al., 2010; Yim et al., 2008; Albert et al., 2007; Carroll and Ahuvia, 2006; Thomson et al., 2005; Oliver et al., 1997). Understanding the drivers of brand love might contribute to:
1) an increased understanding of the causes of brand love;
8 This report will be the first to do research on brand love among Dutch consumers. This is relevant because love is processed prototypically (Aron and Westbay, 1996; Fehr and Russel, 1991; Sternberg, 1986) e.g. there is no clear set of features which determine membership to a class, and prototypes are linked to cultural mental modes (Batra et al., 2008). Results from other studies may be only transferable when the cultures are similar, increasing the utility of this report for the Dutch marketers.
In conclusion, this report will focus on the discovery of drivers of brand love among (Dutch) consumers in the pursuit of understanding strategic consumer behavior (loyalty). The main research question will be “What are the drivers of brand love?” In addition, the following questions are framed: “What are the effects of brand love” and “What are the components of love in a consumer context?”
The following chapter describes the triangular love theory of Sternberg (1986) and the Consumer Object Relationship theory from Shimp and Madden (1988). Based on these two theories and by addressing a number of developed scales to measure brand love or close related constructs this report shows that brand love may have the same foundation as interpersonal love. Next, the differences between brand love and closely related constructs like satisfaction and delight are discussed. Brand love appears to be distinct in three ways: It requires a deep emotional bond between the consumer and the brand, it is based on a long term relationship and it taps into the cognitive and conative part of the consumers mind. Three drivers of brand love: satisfaction, brand-self connection and hedonic value are discussed followed by the discussion of the effect of brand love on positive word of mouth, brand loyalty and spending resources. The drivers and effects of brand love lead to a conceptual model with six hypotheses. Using the six stage design from Hair et al. (2006) the research design is developed and described in chapter three. A questionnaire consisting of 32 questions is used to create input for the models. The constructs are analyzed using the coefficient alpha, the variance extracted and the construct reliability score. Next the measurement model is estimated and judged using different fit criteria (e.g. the comparative fit index). Finally, the hypotheses are tested by estimating the structural model. The last chapter discusses managerial implications of the results, limitations of the study and suggestions for future research.
9 tested as a second order factor and a first order factor. Third, based on known research on interpersonal love, consumer-object love and brand attachment three possible drivers and three possible effects are discussed and selected.
10
2. Theoretical Framework
2.1 From interpersonal love to brand love
One of the most adopted theories from interpersonal love is the triangular love theory of Sternberg (1986). According to this theory love has three components based on different psychological processes: a) intimacy, which is based on the psychological process of emotion, encompasses the feelings of closeness, connectedness, and bondedness one experiences in loving relationships; b)
passion, which is based on the psychological process of motivation, encompasses the drives that lead
to romance, physical attraction, and sexual consummation; and c) decision/commitment, which is based on the psychological process of cognition, encompasses in the short term the decision that one loves another and in the long term the commitment to maintain that love. Following this, the amount of love one experiences depends on the absolute strength of these three components. In addition, the three components interact with each other and with the actions that they produce and that produce them as to form a number of different kinds of loving experiences. For example a mother – child love experience will contain high levels of intimacy and decision/commitment but a low level of passion. Conversely, a fling between adults will have a high level of passion, a low level of intimacy and a low level of decision/commitment. Depending on the absence or presence of the components eight different kinds of love are possible. Ranging from non-love, where none of the components is experienced to consummate love, where all of the components are experienced.
11
Components of PPRs and CORs
Sternberg’s Component Consumer object process analog Psychological process______
Intimacy Liking Emotion (warm component)
Passion Yearning Motivation (hot component)
Decision/Commitment Decision/Commitment Cognition (cold component)
____________________________________________________________________________
Table 1. Shimp and Madden’s components of Consumer Object Relations
Table one shows the transformation of love from people-people relations to consumer-object relations. The proposed model of Shimp and Madden has not been tested empirically. Nevertheless, its contribution can be seen in later work on this topic. Following the conceptual work of Shimp and Madden (1988) researchers looked in detail at the role of love in the consumer brand relationship and developed scales that could statistically confirm similarities between interpersonal love and brand love (Albert et al., 2010; Kamat and Parulekar, 2007; Keh et al., 2007; Ahuvia, 2005, 1993; Thomson et al., 2005).
12
Figure 1. The emotional attachment model of Thomson et al. (2005)
Related results show up in the work of Keh, Pang and Peng (2007) who adopt Sternberg’s triangular theory on love to develop a scale to measure brand love. They define brand love as: “the intimate,
passionate, and committed relationship between a customer and a brand, characterized by its
13
Brand Love Scale
Intimacy
1. I receive considerable support from this brand 2. My relationship with the brand is intimate
3. There is something special about my relationship with this brand
Passion
1. I would rather own this brand than any other brand
2. I cannot imagine another brand of the same product making me as happy as this brand 3. I find myself thinking about this brand frequently during the day
4. I have the impulse to buy it at the sight of the brand
Commitment
1. I will continue to buy this brand when I need the same product 2. I am willing to recommend this brand to my friends
3. I feel that I really can trust this brand
4. This brand is my first choice when I buy the same product
Table 2. The brand love scale developed by Keh et al., 2007
Quite similar results are found by Kamat and Parulekar (2007) in their study on brand love. They end up with five somewhat different factors named: Friendship, Contentment, Admiration, Commitment and Yearning. One remarkable finding is the component Yearning which consists of, Desire Physical Contact (0.72 factor loading), Physical contentment (0.69 factor loading) , Desire Frequent Physical Contact (0.66 factor loading) and Acceptance (0.64 factor loading). This transformation of the passion component into a yearning component was also advised by other researchers (Bengtsson 2003; Breivik and Thorbjørnsen 2008; Shimp and Madden 1988). The discussion about the existence of passion in consumer-object love similar to passion found in interpersonal relations is a vivid one. This report argues that the different results on the passion component may be attributed to three causes.
First, different questions are used to identify the level of passion present. Questions such as “I feel
this brand and I were really meant for each other” may be difficult for consumer to answer. Where questions such as “I would rather own this brand than any other brand” and “I Have an impulse to buy this brand at the sight of it” may be better accustomed to operate in a consumer-object context.
Second, the type of respondents that were used in the studies were very different. High levels of
14 feelings than surveys. It seems plausible to say that a depth interview of three hours can reach a deeper emotional level than a questionnaire consisting of thirty questions.
A comprehensive study of Albert et al. (2010) emphasizes the similarity between brand love and interpersonal love. Dimensions of product love are gathered through a set of structured interviews and then confronted to interpersonal literature (Sternberg, 1997; Hatfield and Sprecher, 1986; Rubin, 1970) to retain true dimensions. Based on the analysis of more than 800 questionnaires they end up with five first order dimensions (Uniqueness, Pleasure, Intimacy, Idealization, Duration, Memories and Dream). These five dimension are linked to two second order factors, Affection and Passion.
The above research clearly indicates the usefulness of transferring known knowledge on interpersonal love to the consumer-object context (Yim et al., 2008). The three components of interpersonal love, Intimacy, Passion and Commitment are found in different studies albeit different naming is sometimes used, e.g. Eros instead of Passion (Whang et al., 2004). Nevertheless there are a number of reasons to be cautious when transferring the knowledge to this area. First, Yoon and Gutchess (2006) have found that consumer-product relationships are processed in a different part of the brain from interpersonal relationships. Second, the methods and items used to develop and validate the brand love / emotional attachment scales are not uniform. Third, the results on the component Passion must be transferred with extra care because of the mixed results on this component. This means that the results, although similar between a number of studies, must be interpreted with care.
15 consumer brand relationship. It reflects the consumers choice to buy the brand and to keep buying the brand in the future. In addition, it is the conscious choice of the consumer to use the brand in his or her daily life. It is the cold component (Shimp and Madden, 1988) of the consumer brand relationship. Together these three components represent the construct brand love. Depending on the absence of presence of each of the components, the feelings for a brand can have several manifestations (Shimp and Madden, 1988) but real brand love is only present when all three components are activated. This multidimensional approach is not only consistent with research on interpersonal love but also advised by a number of researchers e.g. Batra et al. (2010) and Thomson et al. (2005). The construct is shown in table 3 below. This report recognizes that consumers may speak loosely when using the word love in reference to commercial products. As such, many instances of brand love will not be fully analogous to the stronger forms of interpersonal love (Carroll and Ahuvia, 2006; Thomson et al., 2005; Oliver, 1999; Shimp and Madden, 1988).
Intimacy- four items, five points each, Likert-type scale
I have a special relationship with this brand This brand is close to me
I am attached to this brand This brand gives me a good feeling
Passion- three items, five points each, Likert-type scale
Whenever I see this brand I immediately want to buy it I am enthusiastic about this brand
I have a passion for this brand I am really fund of this brand
Commitment- three items, five points each, Likert-type scale
I am loyal to this brand
I will remain loyal to this brand in the future The purchase of this brand is a conscious choice
16
Model Dimensions Author
__________________________________________________________________________________
Consumer-object
relationship Liking, Yearning, Commitment Shimp and Madden, 1988
Brand Relationship
Quality Love/Passion, Self-connection Commitment, interdependence
Intimacy, Brand Partner Quality Fournier, 1998
Object Love model Mania/Agape, Pragma, Eros Whang et al. 2004
Emotional Attachment Affection, Love, Peaceful, Friendly Attachment, Bondedness, Passion
Connection, Delight, Captivated Thomson et al. 2005
Brand Love Scale Passion, Attachment, positive
evaluation, positive emotions, love Carroll and Ahuvia, 2006
Love dimensions Passion, Duration, Self-congruity Dreams, Memories, Pleasure Attraction, Uniqueness, Beauty
Trust, Declaration of affect Albert et al. 2007
Brand Love Scale Friendship, Contentment, Admiration
Commitment, Yearning Kamar and Parulekar , 2007
Brand Love Scale Intimacy, Passion, Commitment Keh et al. 2007
CAAM model Brand-self connection, Automaticity of though Park et al. 2007 The love system Satisfaction with quality, positive emotions
Attachment, Anxiety, Natural fit, Expend resources, Long history, Desire
Intrinsic reasons, Desired self-image Batra et al. 2008 Brand Love Scale Uniqueness, Pleasure, Intimacy
Idealization, Duration, Memories
Dreams Batra et al. 2010
17
2.2 Brand love, satisfaction, brand liking, brand involvement and delight
This part will discuss the difference between the brand love construct and the constructs, Satisfaction, Brand affect/liking and Delight. The choice for these constructs is based on their common use in scientific literature (Thomson et al., 2005; Fournier, 1999; Rust and Varki, 1979).
Satisfaction.
Consumers may be satisfied with a number of different brands. Feelings of brand love only exist for a few brands (Carroll and Ahuvia, 2006). Moreover, satisfactionis
often basedon an expectation-disconfirmation model like the experienced-based norms model. Expectation-disconfirmation is a result from a type of comparison process. Prior to the purchase the consumer has a level of expectation of the brand’s performance. The experience-based norms model states that the expectation is based upon performance norms, brand attitudes and brand expectations (Woodruff, Cadotte, and Jenkins, 1983). Brand love on the other hand does not require any expectation nor does it rely on any disconfirmation. Linked to this is the absence of a lower bound on brand love. In comparison to satisfaction which has dissatisfaction as its lower bound, the absence of brand love is the lower bound of brand love. Furthermore, satisfaction may be reached by deploying a zero-mistake strategy (Oliver et al., 1997). While creating brand love requires an emotional bond with the consumer. In other words, satisfaction is mostly based on a cognitive judgment where brand love also involves a strong conative (emotional) component (Carroll and Ahuvia, 2006).18
Brand involvement
. Involvement is a state of mental readiness that typically influences the allocation of cognitive resources to a consumption object, action or decision (Park and Mittal, 1985). The love emotion goes beyond this mental readiness and is sometimes beyond one’s conscious control. Sternberg (1986) states that passion is a process that cannot be controlled by the consumer. So, love taps more into the emotional realm whereas involvement is more active in the cognitive realm (Thomson et al.,2005).Delight
. Delight is based on the expectations-disconfirmation model (Oliver, 1989) and the primaryeffect of delight occurs when the product performs outside the range of expectations. This implies that delight can occur within a single transaction. So delight is more a momentum emotion whereas brand love is based upon a more continuous feeling towards the brand. Delight can occur quickly and disappear just as fast. Brand love is build up over time and will normally not disappear suddenly.
19
2.3 Drivers of brand Love
This part will discuss the drivers of brand love. The drivers are adopted from research on interpersonal love, consumer-object love and brand attachment (Batra et. al, 2008; Carroll and Ahuvia, 2006). Three drivers are chosen in this report: Satisfaction, brand-self connection and hedonic value. The drivers will be defined and the usefulness of the drivers will be explained in this section.
Satisfaction –
The construct is definedas: A
n overall evaluation based on the total purchase and consumption experience with a good or service over time (Fornell, 1992). Because the focus here is on the consumer brand relationship customer satisfaction is treated as a cumulative measure. It is a judgement that is based on a time period where the product or service was consumed. Consumers who love a brand are presumed to have a strong relation with this brand. Research shows that strong relationships are based on a rooted belief about superior product performance (Fournier, 1998). This implies that consumers who have a strong relation with a brand believe the brand has a superior performance and therefore will experience a high level of satisfaction (Oliver, 1999). A strong connection makes consumers also less vulnerable to negative information about the brand, consequently satisfaction levels will remain high (Swaminathan et al., 2007). Carroll and Ahuvia (2006) also noted that satisfaction is the dominant driver in creating brand love. Results show that satisfaction builds trust and trust is the basis for the consumer brand relationship. Finally, results from Keh et al. (2007) show a significant and positive relation between customer satisfaction and brand love. In summary I formulate the following hypothesis:20
Brand-Self Connection
– The construct is defined as: The extent to which one feels connected to a brand (Park et al., 2007). It reflects the level of connectedness a consumer feels with a brand. The brand is recognized as being a part of the consumer him/herself. Strong attachment is mostly based on a rich set of affectively laden memories that link the object to the self (Thomson et al., 2005). It follows that brands that are an integrated part of the self-identity will be more loved (Ahuvia, 2005). The same results are found in a study by Fournier and Mick (1999) which shows that consumers actually claim certain object to be just as themselves. Especially consumers that claimed to have overlapping selves felt love for their object/brand. This strong bond between the self and the brand is often due to a reference group (Schouten and McAlexander, 1995), a community (McAlexander, Schouten, and Koenig, 2002), or the symbolic meanings of the brand (Wallendorf and Arnould, 1988). Harley Davidson for example, stands for freedom and symbolizes the values of America. People who own and ride a Harley use it to show their own personality (Batra et al., 2009). This symbolic meaning helps consumers to create an identity or resolve life themes (Ahuvia, 2005; Fournier, 1998; Fournier and Yao, 1997). This symbolic meaning seems to enhance peoples love for the object because it takes a deeper meaning (Wallendorf and Arnould, 1988). Also noted in research was the link between “falling in love” with a brand due to the power to generate self-esteem and self-efficacy (Fournier and Yao, 1997; Aron, Paris and Aron, 1995). This connection between the brand and the identity of the consumers makes it likely that the brand plays a central role in the life of the consumer. Finally, the dimension self-congruity, which means having a good fit with the (ideal) self, was found to have a strong association with feelings of love (Albert et al. 2007). This finding is in line with the conceptual work of Sternberg (1986) who states that the arise of passion is a process that cannot be controlled by the consumer. Hence experiencing passion will often be parallel to the feeling of a natural fit with the brand. Feelings of a natural fit are therefore likely to be positive related to feelings of brand love. In summary I formulate the following hypothesis:21
Hedonic value
– Hedonic value is defined as: The consumer’s perception of the relative role of hedonic (as compared to utilitarian) benefits offered by the brand (Carroll and Ahuvia, 2006). Hedonic products are products that deliver pleasure, fun or excitement as one of the primary benefits. Game consoles is a product category with high hedonic value because it gives the user a lot of joy when playing on it. Research suggests that the hedonic value of a brand is positive and related to brand affect (Carroll and Ahuvia, 2006; Chaudhuri and Holbrook, 2001). Some results even show that the utilitarian value of the product category is significantly but negative related to brand affect. Brands that generate more hedonic value (e.g. pleasure, fun and excitement) will probably create a stronger emotional reaction. This implication is supported by research of Albert al. (2007) that shows different results on feelings of love between product categories. Shoes, lingerie and perfume are product categories that were associated with stronger feelings of love. These product categories are able to deliver benefits like pleasure and fun. In addition, pleasure/hedonic value was one of the two dimensions that was explicitly shared by both French and US consumers in a study on object love. Products or objects that are loved often fulfill needs that are not linked to, or go beyond the performance attributes (Wallendorf and Arnould, 1988). One may think of a pair of designer jeans that will create a different emotional response compared to an unbranded pair of jeans. The designer jeans will give the consumer self-esteem, fun and pleasure because it is able to deliver benefits at the emotional level. Unbranded jeans on the other hand can only deliver satisfaction on an attribute level, it is simply a piece of clothing. Batra et al. (2008) also notes that respondents mentioned love objects ability to provide a wide variety of benefits, such as, comfort, pleasure and excitement. Consequently, the hedonic characteristics of the brand may be positive related to brand love. In summary I formulate the following hypothesis:22
2.4 Effects of brand love
Next the effects that brand love may have on strategic consumer behavior will be described. This report covers the following strategic consumer behaviors: Positive word-of-mouth, brand loyalty and spending resources. All constructs will be defined and the proposed relation with brand love is explained.
Positive Word of Mouth
– The construct is defined as: The degree to which the consumer praises the brand to others (Westbrook, 1987). People who have passionate feelings about something are often talking about it. Hence the expectation is a positive relation between brand love and positive word of mouth. Fournier and Yao (1997) show that feelings for a brand can become so strong that respondents become recruiters for the brand. In one of their interviews a woman states that she is so strongly attached to ‘her’ coffee brand that she tries to recruit other people for the brand by spreading the word about its quality. Consequently, consumers who experience love for the brand and use the brand to express themselves may also be more likely to speak positive about the brand to other people. Research in the service domain by Verhoef et al. (2002) shows that affective commitment has a positive relation to customer referrals. The same relation may be valid for a consumer and their brand, if the consumer has an affective commitment to the brand it is likely that he or she actively talks about it. The results of Carroll and Ahuvia (2006) also suggest that there is a positive link between brand love and positive word of mouth. In summary I formulate the following hypothesis:23
Brand Loyalty
- The construct is defined as: The degree to which the consumer is committed to therepurchase of the brand or conative loyalty (Oliver, 1999). Fournier and Yao (1997) report that brand loyalty can be misleading because their results show that exclusive loyalty was only present in one out of six cases. This implies that it is more useful to define loyalty as a (conative) commitment instead of a repurchase rate or a percentage of total spending. Results indicate that consumers who experience higher levels of brand love are more loyal to their brand (Carroll and Ahuvia, 2006; Thomson et al., 2005). If consumers use brands to express themselves it seems logical to think that there will be loyalty towards that brand. Consumers that love their brands will also experience satisfaction with the brand, and satisfaction is the beginning of brand loyalty (Oliver, 1999). This implies a link between satisfaction – brand love – brand loyalty. Research in the service market also reveals that cultivating affectionate ties that comprise both intimacy and passion lead to more loyalty towards the company (Yim et al., 2008). In conclusion I formulate the following hypothesis:
H5: Brand love will have a positive effect on brand loyalty
Spending resources
– This construct is defined as: The willingness to spend more resources (timeand money) on a particular brand compared to similar brands of the same product (Thomson et al., 2005). Consumers will always search for a balance between spending resources and getting the advantages they desire. Brands that are loved are able to satisfy needs and deliver advantages on the emotional level. Therefore it is expected that consumers are willing to spend more resources, like time and money, on brands that they love. For example, playing the guitar can be a hobby but it will take a lot of time and a significant amount of money (buying a guitar and paying for the lessons). Nevertheless a lot of people play the guitar or another instrument because they just love doing it. The same may be true for commercial brands if they can deliver benefits beyond attribute performance. Think of fashion brands. Why are consumers willing to pay more for of a designer dress? Not only the shape of the dress is important but also the additional benefits that come along with it. It may show that the person wearing it is successful or unique. This hypothesis was confirmed in a study by Thomson et al. (2005), their results show that consumers were willing to spend half or even double the amount of money of brands that they loved. In addition, as mentioned before consumers also use brands and objects to show their personality or to reach an ideal self (Ahuvia, 2005). If a brand is able to do this, hence it must be a love brand, then is it logical to accept that consumers are willing to spend more resources on these kind of products. In summary I formulate the following hypothesis:
24
2.5 Conceptual model
The conceptual model shows the expected relations between the drivers of brand love, brand love, and the effects of brand love. Although indirect effects are likely to exist, this report only takes direct effects into account. The following hypotheses are proposed: Satisfaction has a positive effect on brand love, brand-self connection has a positive effect on brand love and hedonic value has a positive effect on brand love. It is expected that brand love will be represented by three components: The component Intimacy, the component Passion and the component Commitment. Finally, brand love has a positive effect on brand loyalty, positive word of mouth and spending resources. Figure 2 represents the proposed hypotheses in a conceptual model.
Complete Path diagram showing Specific Hypothesized Structural
Relationships and Complete Measurement Specifications*
Figure 2. Complete conceptual model including hypothesized relationships
*error terms are not shown for simplicity
Drivers
H1 Satisfaction has a positive effect on brand love
H2 Brand-self connection has a positive effect on brand love H3 Hedonic value has a positive effect on brand love
Effects
H4 Brand love has a positive effect on positive word of mouth H5 Brand love has a positive effect on brand loyalty
H6 Brand love has a positive effect on spending resources
25
Defining the Individual Constructs What items are to be used as measured variables?
Develop and Specify the Measurement Model Make measured variables with constructs Draw a path diagram for the measurement model
Design a Study to Produce Empirical Results Asses the adequacy of the sample size
Select the estimation method and missing data approach
Assesing Measurement Model Vailidity
Asses line GOF and construct validity of measurement model
Specify Structural Model
Convert measurement model to structural model
Asses Structural Model Validity
Asses the GOF and significance, direction, and size of structural parameter estimates Measurement model Valid? Structural model Valid? Refine measures
and design a new study
Proceed to test structural model with stage 5 and 6
Refine model and test with new
data
Draw substantive conclusions and recommendations
3.
Research Design
The six stage design from Hair et al. (2006) is used to develop and test the measurement and structural model (i.e. conceptual model). This chapter will discuss the design of the six stages. The results are discussed in chapter four. The six stage design is shown below in figure 3.
Stage 1 Stage 2 Stage 3 Stage 4 No Yes No Yes
Figure 3. Six-stage process for structural equation modelling by Hair et. Al (2010)
Stage 5
26
3.1 Defining individual constructs
A new scale was developed for the first order factors of Brand love, Intimacy, Passion and Commitment to fit the proposed meaning of these three components and to better fit the meaning of love of the Dutch consumers. The scales are based on work of Albert et al. (2007), Carroll and Ahuvia (2006), Keh et al. (2007), Thomson et al. (2005) and Yim et al. (2006). In addition, a new scale was developed for the construct Spending resources because it did not exist in current literature.
27
Intimacy- four items, five points each, Likert-type scale
I have a special relationship with this brand This brand is close to me
I am attached to this brand This brand gives me a good feeling
Passion- three items, five points each, Likert-type scale
Whenever I see this brand I immediately want to buy it I am enthusiastic about this brand
I have a passion for this brand I am really fund of this brand
Commitment- three items, five points each, Likert-type scale
I am loyal to this brand
I will remain loyal to this brand in the future The purchase of this brand is a conscious choice
---
Satisfaction- four items, five points each, Likert-type scale
I am satisfied with the quality of this brand If I had to choose again, I would choose this brand Everything accounted I am very satisfied with this brand My choice to buy this brand was a wise one
Self-connectedness - three items, five points each, Likert-type scale
This brand is part of my identity This brand symbolizes my personality I feel personally connected to this brand
Hedonic value- three items, five points each, Semantic-type scale
This brand is functional / pleasurable
This brand is a sensory experience / does a job (-) This brand is useful / pleasurable
---
Positive Word of Mouth- four items, five points each, Likert-type scale
I have recommended this brand to a lot of people I recommend this brand to friends
I actively try to spread the good news about this brand
If asked, I give this brand tons of free word-of-mouth advertising
Brand Loyalty- four items, five points each, Likert-type scale
This is the only brand of this product that I will buy
When I buy this brand, I don’t even notice competing brands
If my store is out of this brand, I’ll postpone buying or go to another store I’ll ‘do without’ rather than buy another brand
Spending resources- three items, five points each, Likert-type scale
I am willing to pay more for this brand than for products of a competing brand
I am willing to spend more time on the purchase of this brand than on products of a competing brand
I am willing to put more energy in the purchase of this brand than in products of a competing brand
28
3.1.1 Constructs validity
In order to asses construct validity three different measures are used:
Coefficient alpha =
∑
Where represents the variance off the total score and the item score . N represents the number of items in the factor. The coefficient alpha indicates how strongly the variables are related as a group. Rule of thumb suggest values of .7 or higher to indicate sufficient correlation (Hair et. al, 2006)
Variance extracted (VE) = ∑
Where represents the standardised factor loading and the number of items. So for items, VE is computed as the total of all squared standardized (estimates) factor loadings divided by the number of items. Values must be .5 or higher to indicate good convergence (Hair et. al, 2006).
Construct reliability(CR) = (∑ )
(∑ ) (∑ )
Construct reliability is computed by dividing the squared sum of factor loadings ( ) for a construct by the squared sum of factor loadings ( ) plus the sum of error variance terms for a construct ( .
Values must be .7 or higher to suggest good reliability (Hair et. al, 2006)
3.1.2 Discriminant validity
29
3.2 Development and specification of the measurement model
The LISREL measurement model is defined as:
=
=
The measurement model specifies how the observed variables, and , are determined through
and
by the latent variables, and . The andterms represent the residuals in and
unexplained by and (Jöreskog and Sörbom, 1983). The measurement model estimation includes 11 x-variables (χ), 21 y-variables (γ), 32 error terms (δ,ε) and 36 covariance terms(Ф). Figure three shows how the observed variables ( ) are determined through the latent variables (
,
error and covariance terms are not shown. The measurement model represents the input for the syntax of LISREL.
Figure 3. Measurement model LISREL without error and covariance terms
3.3 Study design and data collection to produce empirical results
Respondents were recruited using snowball sampling (Goodman, 1961). Close contacts were asked to fill in a questionnaire and to provide a new contact who would be willing to participate. Initial contacts were created among student, working adults and seniors to secure a good representation of the Dutch consumers. Participants were mailed with a link to the website www.thesisstephan.tk. At the first page of the questionnaire responds where faced with the following introduction. The word love was not used to secure that bias in the data collection was minimal.
satis faction
brand-self connect
hedonic
value intimacy passion
30 Next respondents were introduced to thirty-two questions related to the nine latent variables of the conceptual model.
All constructs except hedonic value are measured using a 5-point Likert scale ranging from strongly disagree to strongly agree. Hedonic value is measured using a Semantic-type scale with one questions coded reversly. This report follows other researchers on consumer brand relationship and will treat the 5-point Likert scale used in this study as a continuous variable using the product-moment correlation matrix to estimate the models. (Ahuvia et al., 2010; Batra et al., 2008; Albert et al., 2007; Keh et al., 2007; Carroll and Ahuvia, 2006). In addition, all proposed relations are derived from literature. Hence a Confirmatory Factor Analysis is chosen to test the hypotheses. The programme LISREL 8.3 is used to conduct the different analysis described below.
Based on guidelines from Hair et al. (2006) and Tanaka (1993) the ideal sample size should be around 200 respondents. The estimation technique used will be maximum likelihood or robust maximum likelihood (RML) / maximum likelihood mean-adjusted estimation depending on the distribution on the data (Schumacker and Lomax, 2004; Satorra and Bentler, 1994).
3.4 Model validity of the measurement model
A simple rule for index values that can distinguish poor models from good models is currently not available. Research shows that the use of fit indexes is a long way from being standardized (Jackson et al., 2009). However, several guidelines exist and can be used to assess the acceptability of fit for the model. Hereby using different types of indices and adjusting the cut-off values based on the model characteristics and number of respondents. This report will use fit indexes to assess the model fit based on guidelines and work from a number of researchers: Jackson and Gilapsy (2009), Hair et al,. (2006), Schumacker and Lomax (2004), Kenny and McCoach (2003), Nevitt and Hancock (2000), Schumacker (1992) and Tucker and Lewis (1973).
Read the part below with care
31
Chi-square
( 2), absolute fit index:The chi-square index is the only index that has a statistical significant level which is related to the degrees of freedom of the model. The fit is determined by comparing the observed covariance matrix ( ) with the estimated covariance matrix ( (Schumacker and Lomax, 2004). The observed covariance is computed from sample observation. In this case the covariance matrix is a 32 * 32 matrix with 496 unique values. The estimated covariance matrix is computed by summing the direct and indirect path estimates.
Degrees of Freedom
: [ ]The degrees of freedom indicates the amount of mathematical information that is available for model estimating. Where p is the number of variables and k is number of (free) parameter estimates (Schumacker and Lomax, 2004).
Comparative Fit Index
(CFI), incremental fit index: [ ]The CFI is a incremental fit index. It compares the measurement model with the null model. The null model specifies that all variable are uncorrelated. The CFI is normed so that values range between 0 and 1, with higher values indicating a better fit. Research indicates values of >.95 are needed for an acceptable fit (Hair et. Al, 2006).
Tucker Lewis Index
(TLI), relative fit index: / )The TLI (Tucker Lewis index) compares the measurement model to the null model. The difference with the CFI is that the TLI is not normed so values can be below 0 or above 1. Research indicates values of >.95 are needed for an acceptable fit (Schumacker and Lomax, 2004).
Standardized Root Mean Square Residual
(SRMR), badness-of-fit index: [ ∑ ( ) ]1/232
3.5 Development and specification of the structural model
The LISREL structural model is defined as:
The structural model specifies the causal relationships among the endogenous (determined within the model) variables , between the exogenous (determined outside the model) and endogenous variables and describes the unexplained residuals of the latent factors ( öreskog and Sörbom, 1983). Two structural models have been developed one with brand love as a second order factor (1) and one with brand love as a one order factor (1A) containing the variables from the construct Intimacy, Passion and Commitment. Figure 4 is a simplified representation of structural model 1 and figure 5 shows a simplified representation of structural model 1A. Each arrow represents one of the six hypotheses from the conceptual model.
Figure 4. Structural model 1 Figure 5. Structural model 1A
3.6 Model validity of the structural model
To test the validity of the structural model the same fit criteria will be used as described above.
satisfacti on hedonic value BS-connect brand love WOM+ spending resource brand loyalty satisfacti on hedonic value BS-connect WOM+ spending resource brand loyalty brand love
Intimacy Passion Commit
33
4.
Results
This chapter will discuss the results of the six stage process described in chapter 3.
4.1 Data and subjects
In total 205 respondents filled in the questionnaire of which 202 questionnaires were filled in correctly. Those who only partially filled in the questionnaire were deleted from the sample. Of the 202 respondents, 102 (51%) were male and 100 (49%) were female. Respondents age was reasonable well balanced between groups, <25years(25%), 26-40years(34%), 41-55years(39%) and >55years(2%). The 202 respondents reported on 103 unique brands of which most were related to fashion(22%), electronics(20%) and consumer goods(19%). The vast majority(75,7%) of respondents reported on a brand with which they had a long-term relationship with that was more than five years.
Table 6. Brands divided between categories
4.2 Individual construct scores
Prior to testing the constructs were validated by an expert panel consisting of three members of the faculty Business and Economics of the university of Groningen to confirm face validity. Based on the results of the discussion some questions were replaced or slightly changed. A pre-test was conducted to trace and eliminate vagueness and errors from the questionnaire.
Six respondents participated in the pre-test of the questionnaire. The pre-test results did not indicate any vagueness or errors. Only the statement “does a job” seemed hard to interpret but because no alternatives were found the question remained in the questionnaire. In combination with the theoretical grounding this secured face validity on all scales used in the model.
Intimacy
Scale score: Minimum 4, maximum 15, ̅ = 12,0, s = 2,31 maximum likelihood factor analysis produced one factor with all loading in excess of .78 and explaining 77,59% of the variance in the items, coefficient alpha = .86, CR (construct reliability) = .86, VE (Variance extracted) = .66
Brand categories Percentage (%) Number (#)
34
Passion
Scale score: minimum 5, maximum 20, ̅ = 14,17, s = 3,15 maximum likelihood factor analysis produced one factor with all loading in excess of .56 and explaining 56,9% of the variance in the items, coefficient alpha = .73, CR = .75, VE = .43
Commitment
Scale score: minimum 3, maximum 15, ̅ = 10,59 , s = 2,77 maximum likelihood factor analysis produced one factor with all loading in excess of .77 and explaining 74,46% of the variance in the items, coefficient alpha = .83, CR = .83, VE = .61
Satisfaction
Scale score: minimum 4, maximum 20, ̅ = 15,6, s = 3,1 maximum likelihood factor analysis produced one factor with all loading in excess of .72 and explaining 70,2% of the variance in the items,
coefficient alpha = .86, CR = .87, VE = .63
Brand-Self connection
Scale score: minimum 3, maximum 15, ̅ = 8,7, s = 2,7 maximum likelihood factor analysis produced one factor with all loading in excess of .53 and explaining 61,9% of the variance in the items,
coefficient alpha = .69, CR = .75, VE = .43
Hedonic value
Scale score: minimum 3, maximum 15, ̅ = 11,23 , s = 2,85 maximum likelihood factor analysis produced one factor with all loading in excess of .68 and explaining 79,9% of the variance in the items, coefficient alpha = .87, CR = .75, VE = .71
Positive Word of Mouth
Scale score: minimum 5, maximum 20, ̅ = 13,14, s = 3,5 maximum likelihood factor analysis produced one factor with all loading in excess of .57 and explaining 59,6% of the variance in the items, coefficient alpha = .77, CR = .80 ,VE = .45
Brand Loyalty
35
Spending Resources
Scale score: minimum 3, maximum 15, ̅ = 9,72, s = 3,31 maximum likelihood factor analysis produced one factor with all loading in excess of .84 and explaining 82,49% of the variance in the items, coefficient alpha = .89, CR = .75 ,VE = .73
In conclusion, all factor loadings are significant, and in each case the factor explains a large proportion of the variance (0.57 to 0.82). However, the items Passion2 (loading .56), Loyalty2 (loading .56), BSconnect3 (loading 0.53) and WOM1 (loading 0.57) load below advised standards of .60 (Hair et al., 2006). Furthermore, a relative high number of variables load between 0.6 en 0.7 which is low for a theoretical model. One would expect loadings higher than 0.70. This shows that the developed scales have modest validity. One question, “this brand is close to me”, on the scale intimacy is dropped because of poor loading .390. This causes the Cronbach alpha to go from .738 to .855 showing an increase in reliability of the scale. It is difficult to say why this specific question loads poorly on the factor. One possible reason may be that the question is seen as less appropriate for a consumer brand relationship. Questioning a small number of respondents show they felt that the statement of this question was more appropriate for an interpersonal relationship. This could be one of the causes for the relative low score (mean = 3.3.) compared to the score of the factor (mean = 4.0). All scales show a Cronbach alpha that is near or above the 0.6 cut of point (Hair et al., 2006). In addition, all item to total correlations are above .45 and all interim-correlation above 0.3.
36
4.3 Discriminant validity
The percentage variance-extracted and the squared correlation estimate (between constructs) are used to test the constructs on discriminant validity. In order to indicate sufficient discriminant validity the VE score should be higher than the squared correlation (SC). The squared correlation between factors should show results in line with the conceptual model.
4.3.1 Drivers of brand love
Satisfaction has a VE score of .62 and the squared correlations are .31 (brand-self connect) and .07 (hedonic value). Brand-self connect has a VE score of .43 and the squared correlations are .31 (satisfaction) and .06 (hedonic value). Hedonic value has a VE score of .69 and the squared correlations are .07 (satisfaction) and .06 (brand-self connect). As expected all construct are significant positively related to one another. However, all constructs appear to be truly distinct from one another and show VE scores well above SC estimates.
4.3.2 Components of brand love
Intimacy has a VE score of .70 and the squared correlations are .25 (commitment) and .62 (passion). Passion has a VE score of .43 and the squared correlations are .62 (intimacy) and .65 (commitment). Commitment has a VE score of .63 and the squared correlations are .25 (intimacy) and .65 (passion). It is hard to interpret these results, all correlations are positive as expected but intimacy and commitment show modest correlation. Expected was high correlation between all three variables. In addition, Passion seems to predict intimacy and commitment better than its own items which is not ideal.
4.3.3 Effects of brand love
37 In conclusion, the discriminant validity of the components of brand love and the effects of brand love are modest. The discriminant validity of the drivers of brand love is good. The construct validity of hedonic value and spending resources is remarkable high. This could be an indication that the effect of hedonic value on brand love and the effect from brand love on spending resources are different than anticipated (Carroll and Ahuvia, 2006). The low correlation of commitment with the intimacy raises some concerns about the specified conceptual model. Taking into account a number of factor loadings between 0.6 en 0.7, the overall results of the model are poor / modest for a model a priori. An overview of the discriminant validity is shown below in table 7.
38
4.4 The measurement model
First a univariate test is computed in PRELIS to check the distribution of the data. The output (see appendix table 2) shows measures of skewness and kurtosis of each variable and gives test statistics for testing the hypotheses of zero skewness and zero kurtosis (Jöreskog, 1999). The results show only ten variables showed insignificant p-values <0.05 on the skewness and kurtosis test and none of the variables passed both tests. Although only a few variables show extreme kurtosis or skewness >1.5. This is particular interesting because research on the use of ordinal variables as continuous variables show limited bias in parameter estimates when skewness and/or kurtosis is limited (Bentler, 2010; Babakus et al., 1987; Ethington, 1987). The default estimation procedure in LISREL, maximum likelihood, uses the assumption that the observed variables has a multivariate normal distribution (Fletcher and Powell, 1963). When the data appears to have a nonnormal distribution it is advised to use robust maximum likelihood (RML) or maximum likelihood mean-adjusted estimation (Schumaker and Lomax, 2004; Satorra and Bentler, 1994). RML estimation uses the asymptotic covariance matrix (ACM) produced by LISREL. The ACM adjust the 2 and standard error estimates by a scaling factor based on the non-normality in the data.
The measurement model estimation includes 21 x-variables, 11 y-variables, 32 error terms and 36 covariance terms. After 19 iterations LISREL analysis shows a 2 of 1407.07 (Satorra-Bentler Scaled Chi-Square) with a p-value of <0.01. So the overall model fit has a significant p-value which indicate that the observed and implied (estimated) variance-covariance matrices differ. Researchers argue that model complexity and a high number of respondents influence the 2 statistic (Kenny and McCoach, 2003). Or as Bentler (2010) puts it: “each df represents one way of being wrong about the model”. The chi-square statistic has a tendency to indicate a significant probability level when the measurement model is estimated with a sample size of >200 (Hair et all., 2006; Schumacker and Lomax, 2004; Kenny and McCoach, 2003; Schumacker, 1992). The calculation of the chi-square statistic shows that sample size has a positive effect on the increase of the chi-square statistic. In addition some researchers found that the 2 statistic is sensitive from departure from normality of the data (Boomsma, 1983; Muthen 1983). Although results from Ethington (1987) found little influence of skewness on the 2 statistic. Taking all in account it seems reasonable to expect a significant 2 statistic . The CFI value, the TLI and SRMR are all below the cut-off points (see table 8). An extensive overview of the fit statistics of measurement model 1 is shown in appendix table 3.
39 specification search (Leamer, 1978). A number of possible causes are found using the Modification Indexes (MI) and extensive output from LISREL. Three variables, resp. Passion2, Loyalty2 and WOM1 are selected to be deleted from the measurement model because of relative low loadings (<0.60) estimates and/or high error variances (>0.90) and/or high MI values (>20). MI values in this case indicate possible cross-loadings.
The adjusted measurement model (see figure 6) shows improvement in fit statistics. All fit indexes are near but still below their rules of thumb as shown in table 8. In addition, all factor loadings and error variances are significant at the 99% significant interval. The t-values of the model are computed by dividing the path coefficient estimate by its standard error, t-values with a value greater than 2.576 are considered statistically significant at a 99% significance interval (Jöreskog, 2003). In addition, seven factors show statistically significant correlation at a 99% significance level. The factor Hedonic value correlates insignificant on brand-self connect, loyalty, commitment and hedonic value and the factor Spending resources has insignificance correlation with satisfaction, hedonic value, intimacy and passion. As mentioned before, this may imply that the effect of hedonic value on brand love and the effect from brand love on spending resources is not as expected. An extensive overview of the fit statistics of measurement model 1A is shown in appendix table 4.
Figure 6. Finalised measurement model without error and covariance terms
Fit index Cut-off point Measurement model 1 Measurement model 1A
2 significant p-values can be expected 1407.07 p-value >0.01 907.91 p-value >0.01 - 398 314 CFI >0.95 0.92 0.94 TLI >0.95 0.90 0.93 SRMR <.08 0.09 0.08
Table 8. Fit indices for the measurement models of brand love
satis faction
brand-self connect
hedonic
value intimacy passion
40
4.5 The structural model
The second order factor model had difficulties converging and needed 164 iterations before being estimated. This is beyond the iteration limit of fifty that LISREL has as default. To estimate the model the AD=OFF (Admissibility check after n iterations) setting had to be added to the syntax. The results show that the model has a poor fit with the data. Furthermore, two estimates were insignificant (brand-self connect on brand love and hedonic value on brand love). And the construct brand-self connect caused negative error variance in the equation. In literature this is called a Heywood case (Schumacker and Lomax, 2004). There are number of possible solutions to deal with a Heywood case: Fixing the offending estimate to a very small value, adopting model parameterization that ensures positive error variance or using a model with equality constrains that ensure nonnegative error variance estimates (Dillon et. al, 1987). If none of the options have satisfying result it is advised to delete the violating estimate. In this case the offending estimate was fixed to a small value (0.05) in order to successfully estimate the model. An extensive overview of the fit statistics of structural model 1 is shown in appendix table 5.
The second structural model with brand love as a first order factor containing the variables from the construct Intimacy, Passion and Commitment showed slightly better fit values. The scale score of brand love is reasonable (all loadings >.612, coefficient alpha, .881), keeping in mind that adding variables to a factor causes an increase the Cronbach alpha by default. However, only one relationship in this measurement model is not significant at the 99% significant interval. Making this model more useful for interpretations. An extensive overview of the fit statistics of structural model 1A is shown in appendix table 6. Table 8 shows the scores of both models.
Table 8. Fit indices for the structural models of brand love
Fit index Cut-off point Structural model 1 Structural model 1A
41 Unfortunately this model also suffered from the same Heywood case. The offending estimate was again fixed to a small value (0.05) in order to successfully estimate the model. Although the model fit suggests that conclusions must interpreted with care. Figure 7 shows the (standardized) path coefficients.
Figure 7. Structural model 1A with brand love as an one order factor, * indicates p-values <0.01, **indicates p-values <0.05
Although model fit suggest that the results are not very solid all hypotheses are statistically significant. An overview is shown in table 9.
Drivers
Satisfaction’s effect on brand love H1 .64* Brand-self connection’s effect on brand love H2 .38* Hedonic value’s effect on brand love H3 .11**
Effects
Brand love’s effect on WOM+ H4 .92*
Brand love’s effect on brand loyalty H5 .68* Brand love’s effect on spending resources H6 .29*
Table 9. Hypotheses testing of the drivers and effects of brand love, * indicates p-values <0.01, **indicates p-values <0.05
Table 9 shows that satisfaction has the biggest impact on brand love and hedonic value the lowest. In addition, it shows that brand love has the biggest impact on WOM+ and the smallest on spending resources.
X9 X10
Brand-Self
connection Brand Love
42 These path estimates can be used to predict a value based on any observed value. For example an estimated value for brand love would be computed using the following equation:
̂
BL = (Satisfaction)+ + 0.11(Hedonic value)An observed value for the drivers would be represented by a multiple equation prediction. For example an estimated value for WOM+ would be computed using the following equation:
