Find your perfect match!

(1)

MASTER THESIS

Larissa Hendriks

Find your perfect match!

A study about the effects of personality traits on

relationship quality, relationship evaluation and

(2)

Master Thesis

Final version

20-06-2016 Larissa Hendriks s2071983 Mathenesserdijk 283a +31623602970 l.m.e.hendriks@student.rug.nl

Master thesis MSc Marketing Intelligence and Management University of Groningen Faculty of Economics & Business Department of Marketing PO Box 800, 9700 AV Groningen (NL) First supervisor: dr. ir. M. J. Gijsenberg (m.j.gijsenberg@rug.nl) Second supervisor: dr. J. T. Bouma (j.t.bouma@rug.nl) External supervisors: dr. L.H. Teunter (l.teunter@metrixlab.com) &

(3)

Management summary

(4)

1. Introduction

Since a few decades marketing is changing from transactional to relational orientations (Dwyer, Churr, & Oh 1987; Jackson, 1985). Marketing is supposed to give attention to the connection between the firm and its customers (Reibstein, Day & Wind, 2009). In this view Customer Relationship Management (CRM) has emerged as one of the dominant mantras in business strategy circles according to Palmatier, Dant, Grewal and Evans (2006). Both in business practice and as a focus of academic research, CRM has experienced explosive growth in the past decades (Srinivasan & Moorman, 2005). Morgan and Hunt (1994, p. 22) define CRM as “all marketing activities directed towards establishing, developing, and maintaining successful relational exchanges”. One of the most common outcomes expected from CRM is increased loyalty (Palmatier et al., 2006).

Customer loyalty refers to the strength of the relationship between customers and a firm. Li and Green (2010) state that loyal customers provide a constant stream of revenue and ensure cost reductions, hence, increasing profitability. Moreover, customers who have a long-term relationship with the firm are less costly and buy more products (Reichheld, 1996; Ganesh, Arnold & Reynolds, 2000) compared to the costs of acquiring new customers (Bhattacharya, 1998; Colgate & Danaher, 2000; Gupta, Lehmann & Stuart, 2004).

Customer loyalty can be improved by successful CRM efforts through the creation of stronger relational bonds (Palmatier et al., 2006). Many firms invest heavily in CRM to create these relational bonds, assuming that close customer-firm and customer-employee relationships lead to positive financial outcomes (Palmatier et al. 2006). To create positive financial outcomes, one could state that firms should allocate their resources to customers who are likely to be receptive to CRM (Palmatier, 2008) and profitable to the firm. Anderson and Narus (1991) propose that organizations need to pursue both transactional and relational marketing simultaneously because not all customers want the same customer-firm relationship. Due to little knowledge about differences in these relationship preferences it is hard to determine which customers are likely to be receptive to CRM. Consequently, this makes it hard to optimize the budget allocation for CRM (Palmatier et al., 2006). This shows the relevance of researching customer-firm relationships.

(6)

Page 2 relationship, namely relationship quality and relationship evaluation. Both these concepts can explain how and why people (dis)engage in close relationships.

Next to these relationship elements, it is important to include personality traits to clarify individual differences and to examine the customers who are likely to be receptive to CRM. Personality traits describe people in terms of emotions and thoughts (Parks-Leduc, Feldman & Bardi, 2015). In the present research personality traits are covered by three psychological theories, namely the Five Factor Model (FFM), the general attachment theory and the business-specific attachment theory. The FFM describes a personality based on five dimensions. The general attachment theory of Bowlby (1973) provides a framework for understanding why people form close emotional bonds with others. An extension of this general theory is the business-specific attachment theory. This theory describes the attachment style of people towards a firm in a specific customer-firm relationship (Klohnen et al., 2005). Research about the effect of personality traits on relationship quality, relationship evaluation and ultimately loyalty is limited so far. Especially the general and specific attachment theory provide new ways to look at customer-firm relationships in the marketing field.

The present study aims to provide and expand current knowledge about the impact of personality traits on customer-firm relationships. Customer attachment styles and the FFM expand this body of knowledge by providing new theoretical explanations for differences in relationship quality, relationship evaluation and ultimately loyalty. Besides, this study provides insights about the impact of attachment styles across different industries to see to what extent previous outcomes of research about attachment styles are generalizable.

(7)

Page 3 budget allocation among customers. In order to provide these insights, we give an answer to the following research question:

“To what extent do personality traits influence relationship quality, relationship evaluation and

ultimately loyalty?”

(8)

Page 4

2. Theoretical Framework

To provide an answer to the research question “To what extent do personality traits influence

relationship quality, relationship evaluation and ultimately loyalty?” we explain the variables and their

relationships in the same order as described in the research question. First, we discuss the effect of personality traits on relationship quality, followed by the effect on relationship evaluation and loyalty. The underlying reasoning behind the research question is namely that the effect of personality traits on loyalty is a sequential process. To make the structure clear, we will shortly introduce all variables here first followed by a discussion of the relationships between the variables.

The basic idea is that personality traits affect loyalty and that this effect is mediated by relationship quality and relationship evaluation. The direct effect of personality traits on loyalty has already been researched by Mende, Bolton and Bitner (2013). So, to add new knowledge to this research area, this study only focuses on the direct impact of personality traits on relationship quality and relationship evaluation and not on loyalty.

The personality traits included in the present study are the Five Factor Model, the general attachment styles and the business-specific attachment styles. These concepts and its effect on relationship quality are described in section 2.1. We describe relationship quality in terms of trust and commitment according to De Wulf, Odekerken-Schröder and Iacobucci (2001). In short, trust involves the appraisal of partners as reliable and predictable (Rempel, Holmes & Zanna, 1985) and commitment is the emotional or psychological attachment to a brand or firm (Verhoef, 2003). Section 2.2 explains trust and commitment more detailed. Furthermore, to describe relationship evaluation we use the framework of De Haan, Verhoef and Wiesel (2014). This framework includes three Customer Feedback Metrics (CFMs): Customer satisfaction (CSAT), Net Promotor Score (NPS) and Customer Effort Score (CES). In short, the first CFM describes to what extent a customer is satisfied with (a part of) the firm. NPS reflects to what degree a customer is willing to recommend the firm. CES describes the amount of effort it takes for a customer to handle an issue with the firm. So far, there is not much research done about the impact of CES on customer behavior (De Haan, Verhoef & Wiesel, 2014), so this has an explorative nature. We include all three CFMs to create a great and more complete view of relationship evaluation. Section 2.3 explains the CFMs more detailed. Loyalty is the final outcome and is measured by the retention intention and churn intention, which we discuss in section 2.5.

(9)

Page 5

2.1 Personality traits

The area of personality traits is worth a second look according to Bloemer et al. (2003) De Wulf et al. (2001) and Odekerken-Schröder et al. (2003). Ajzen (1987) mentioned that personality traits play an important role in predicting and explaining human behavior. For example, attachment styles, a form of personality traits, are crucial determinants of relationship quality in romantic relationships (Mikulincer 1998; Simpson 1990). Moreover, Furst et al. (1996) reported that consumer purchase behaviors can be associated with personality traits.

This study takes into account the personality traits in three ways, namely by the Five Factor Model (FFM), general attachment styles and business-specific attachment styles. In short, the FFM describes the personality of a person according to five dimensions. The attachment theory explains several styles that describe why and how people have relationships with others. Attachment styles can be general towards other persons or they can be specific towards a target in a specific interpersonal relationship or customer-firm relationship (Klohnen et al., 2005). Specific attachment styles may or may not be congruent with the person’s general or higher-level attachment styles (Mikulincer & Shaver, 2007).

We chose to include the attachment theory, because both the general and the specific attachment styles provide us a new way to look at customer-firm relationships. So far, the attachment theory is mainly used in psychological research, but it is interesting to see if the impact of attachment styles in interpersonal relationships is generalizable to customer-firms relationships. Knowledge is limited about the predictive value of the general and specific attachment styles in terms of loyalty and there is no consensus yet about which theory is better in explaining consumer behavior. This is also apparent from the following contradicting researches. At the one hand Mende, Bolton and Bitner (2013) stated that relationship-specific attachment styles are better predictors of relationship outcomes than general attachment styles. On the other hand, Verbeke, Bagozzi and van den Berg (2013) state that attachment styles are general instead of relationship-specific, because it is connected to the biological system and brains of a human. Namely, the regulation of dopamine can be associated with attachment styles.

(10)

Page 6 Next to the outline of the psychological theories, we discuss the effect of these theories on relationship quality in terms of trust and commitment (De Wulf, Odekerken-Schröder & Iacobucci, 2001). It is likely that the perceived quality of the relationship is influenced by the personality of the persons involved in the relationship. For example, some people trust others easily, while others are more reserved in relationships with others. Mikulincer (1998) and Simpson (1990) have shown that personality traits are crucial determinants of relationship quality.

2.1.1 General attachment styles

The attachment theory of Bowlby (1973) provides a framework for understanding why people form close emotional bonds with others. This theory results in different attachment styles. An attachment style is the systematic pattern of relational expectations, needs, emotions, and social behaviors that results from the internalization of a particular history of attachment experiences (Mikulincer & Shaver, 2007). Attachment theory is a major foundation for research in psychology that studies interpersonal relationships (Hazan & Shaver, 1994).

In general, attachment styles are measured along two dimensions ‘attachment anxiety’ and ‘attachment avoidance’ (Brennan, Clark & Shaver, 1998). The following definitions are used in the current research to describe these dimensions, based on Brennan, Clark and Shaver (1998). First, attachment anxiety is the extent to which a person worries that the other might not be available in times of needs, has an excessive need for approval, and fears rejection and abandonment from this firm. Anxiously attached people tend to ‘‘be involved in affectively unpleasant relationships’’ (Simpson 1990, p. 972). Second, attachment avoidance is the extent to which a person has an excessive need for self-reliance, fears depending on others, distrusts relationship partners’ goodwill, and strives for emotional and cognitive distance from partners. If a person has a low level of anxiety and a low level of avoidance, one could be called secure. Secure people are very happy, satisfied and desire close relationships with others. They basically have a positive view of themselves, their partners and their relationships. To summarize, one could have an anxious, avoidant or secure attachment style.

(11)

Page 7 According to Simpson (1990) attachment styles interfere with the development of trust. Individuals with a low level of avoidance and anxiety (secure persons) find it easy to trust and forgive others, because they believe that others have good intentions and generally expect positive outcomes from relationships (Bowlby 1973; Jang et al. 2002; Kachadourian et al.. 2004; Mikulincer et al., 2001). Anxious persons are insecure about the stability of their relationships which restrains them from having relationships with a high level of trust, because insecurity plays a critical role in the formation and development of trust (Berdscheid & Fei, 1977). People with a high level of attachment avoidance have a ‘‘mistrusting bias’’ which means that they do not consider it a relational goal to become comfortable with relying on and trusting a partner (Zhang & Hazan 2002, p. 228). Simpson (1990) also confirmed that high levels of attachment anxiety or attachment avoidance have a negative influence on trusting a partner. In this view, it is expected that anxiety and avoidance have a negative impact on trust.

In addition, individuals with a low level of avoidance and anxiety (secure persons) desire close and loyal relationships with others (Brennan, Clark & Shaver, 1998). One of the necessary aspects of a close relationship is commitment (Rusbult & Buunk, 1993). Commitment exists when the individual consumer identifies with and is attached to their relational partner (Fullerton, 2003; Gruen et al., 2000). Anxious persons can sometimes show extreme levels of commitment, but are unwilling or unable to commit to permanent relationships (Simpson, 1990). Their insecurity about the stability of the relationship restrains them from developing relationships characterized by high levels of commitment (Simpson, 1990). Avoidant persons are suspicious, skeptical and unreliable, which restrains them from committing to others (Simpson, 1990). In line with these researches, we expect that anxiety and avoidance have a negative impact on commitment.

2.1.2 Business-specific attachment styles

(12)

Page 8 similar, but that the specific attachment style is a better predictor of relationship outcomes in terms of relationship quality and relationship evaluation. Now, we will discuss the limited literature about the impact of specific attachment styles on relationship quality in terms of trust and commitment. The effects on relationship evaluation are explained in section 2.4.

In line with the effects of the general attachment styles, Mende and Bolton (2011) have shown that low levels of business-specific attachment anxiety and business-specific attachment avoidance both have a positive effect on the firm employee and the firm itself in terms of trust. In this view, we expect that specific anxiety and specific avoidance have a negative impact on trust. Also in line with the effects of the general attachment styles, Mende and Bolton (2011) have shown that business-specific customer attachment styles are likely to interfere with the development of commitment in service relationships because they influence a person’s affect regulation (Mende & Bolton, 2011). Mende and Bolton (2011) have shown that customers with a high level of specific attachment anxiety and specific attachment avoidance have a lower level of commitment. In line with this research, we expect that anxiety and avoidance have a negative impact on commitment.

2.1.3 Five Factor Model

In marketing research so far, personality traits have been described by the Five Factor Model (FFM), also known as the Big Five. The FFM is introduced by McCrae and John (1992). This model consists of five dimensions, namely Extraversion, Agreeableness, Conscientiousness, Neuroticism and Openness to experience (see Table 2.1) (McCrae & John, 1992; Sadowski & Cogburn, 1997; Thoms, Moore & Scott, 1996; Parks-Leduc et al., 2015).

Table 2.1. Five Factor Model of Personality traits.

Dimension A persons’ degree of…

Extraversion Assertiveness, sociability and energy

Agreeableness Friendliness, cooperativeness, being courteous, flexible and trusting. Conscientiousness Dependability, achievement orientation and perseverance

Neuroticism Emotional stability, anxiety, self-confidence, self-consciousness Openness to experience Imaginativeness, curiosity, open-mindedness and being artistic.

(13)

Page 9 relationship quality and with elements of the FFM measures such as warmth and assertiveness (Extraversion), feelings (Openness to experience), trust and altruism (Agreeableness), and achievement-striving and self-discipline (Conscientiousness). Furthermore, relationship quality was negatively related to depression (Neuroticism). In line with these researches, we expect that individuals with a high level of Extraversion, Agreeableness and Conscientiousness have a higher level of trust and commitment to the firm, while Neuroticism is expected to have a negative effect on relationship quality.

2.1.4 Comparison FFM and attachment styles

We expect that the general and specific attachment theory and the FFM would share some variance, but would also offer unique explanatory power because of their special relevance to the domain of close relationships. On the one hand these theories are similar. Namely, recent research supports the positive correlations between secure attachment and Extraversion on the assertiveness aspect. Furthermore we expect a positive correlation between the secure attachment and Conscientiousness, because both aspects are affected by self-discipline and competence (Noftle & Shaver, 2006). Moreover, the secure attachment correlates positively with Agreeableness, because trust and altruism facets are important for both aspects (Gosling et al., 2003). The secure attachment correlates negatively with Neuroticism, because anxiety, avoidance and neuroticism are strongly related to insecurity. Most researches did not found a correlation between the attachment style and Openness to Experience.

(14)

Page 10

2.2 Relationship quality

Relationship quality is affected by personality traits as we have shown in the previous paragraphs and moreover, relationship quality affects the way the relationship is evaluated in terms of customer satisfaction, NPS and CES (De Haan, Verhoef & Wiesel, 2014). Previous research shows that among others, trust and commitment are differentially related to relationship drivers and outcomes, which implies that the constructs are distinct (Andaleeb, 1996; Anderson & Narus 1990; Morgan & Hunt, 1994). The literature about CES is limited, so drivers and outcomes of this variable are of explorative nature.

2.2.1 Trust

Trust involves the appraisal of partners as reliable and predictable, the belief that partners are concerned with one’s needs and can be counted on in times of need, and feelings of confidence in the strength of the relationship (Rempel, Holmes & Zanna, 1985). Trust reflects the appraisal of others as dependable and predictable, and the belief that a partner is concerned with one’s needs (Mikulincer 1998). Trust is necessary for successful customer relationship building (Berry, 1995).

It is expected that trust has an effect on CFMs. In line with Anderson and Narus (1990, p.46) and Geykens, Steenkamp and Kumar (1998), satisfaction is conceptualized as a consequence of trust. Satisfaction is a global evaluation of fulfilment in the relationship (Dwyer et al, 1987). According to Geykens, Steenkamp and Kumar (1998) building trust is a very effective way to increase satisfaction and long-term orientation. In line with these researches, we expect a positive effect of trust on satisfaction.

Moreover, trust is positively related to customer referrals in a service context (Gwinner et al., 1999). People with more trust in a firm, will recommend more and will have more positive recommendations (Gwinner et al., 1999). These customer referrals produces the greatest benefit for firms according to Brown et al. (2005) and Johnson and Selnes (2004). The referral value is reflected by the NPS in the framework of CFMs. So in this view, we expect that trust has a positive influence on NPS.

(15)

Page 11 a good outcome of the way the firm deals with the specific issue. Being suspicious takes effort and in this way it could be expected that a lower level of trust results in a higher CES.

2.2.2 Commitment

Verhoef (2003) states that commitment is a relevant variable in customer relationships. Commitment is the emotional or psychological attachment to a brand or firm that develops before a customer would be able to determine that their repeat purchase behavior was derived from a sense of loyalty (Beatty & Kahle, 1988: p4). It measures consumer attitudes of attachment to a brand. Morgan and Hunt (1994) describe commitment as an enduring desire to continue a relationship. Furthermore, commitment is the most prominent perception representing the strength of the relationship (Moorman, Zaltman, & Desphandé, 1992; Morgan & Hunt, 1994).

Previous research has shown that commitment is highly positively associated with satisfaction and loyalty (Dimitriades, 2006). A customer who is more committed will be more satisfied, because it can build strong relationships. In line with Palmatier et al., (2006) strong relationships lead to positive relationship outcomes such as customer satisfaction. In this view, we expect that commitment positively influences satisfaction.

Moreover, Dick and Basu (1994) suggested that a potential consequence of commitment may include word of mouth communications. Mayer and Schoorman (1992, p.63) found that an individual who is high in commitment is motivated to actively engage in behaviors that would help the employing organization achieve its goals. Furthermore Harrison-Walker (2001) shows that commitment is positively related to word of mouth activities. One of these word of mouth activities is praise, which could be interpreted as recommendation. In this view, it is expected that commitment has a positive effect on NPS.

(16)

Page 12

2.3 Relationship evaluation

Relationship evaluation is affected by relationship quality as we have shown in the previous paragraph and moreover, relationship evaluation affects loyalty. In case a customer has a positive perception of the relationship, a customer is more likely to be loyal. The three CFMs that describe the relationship evaluation, customer satisfaction, NPS and CES are extremely popular, in particular with the development of the NPS (De Haan, Verhoef & Wiesel, 2014). Customer Satisfaction and CES are about customer perceptions looking backwards, while NPS and loyalty intentions are about customer perceptions looking forwards. Furthermore, the use of more than one metric provides significantly better predictions (De Haan, Verhoef & Wiesel, 2014). In this view they state that the metrics satisfaction and NPS strengthen each other and that both of them are good indicators of future customer behavior and firm performance.

2.3.1. Satisfaction

Satisfaction is important with respect to the evaluation of a supplier’s offerings (Bolton & Lemon, 1999). Customer satisfaction can be described as the consequence of the expectancy disconfirmation paradigm. This paradigm is the outcome of a customers’ comparison between perceived and expected service performance (Churchill & Surprenant 1982). Furthermore, satisfaction can be defined as the emotional state that occurs as a result of customer’s interactions with the firm over time (Anderson, Fornell, & Lehmann, 1994; Crosby, Evans & Cowles, 1990). This construct is an often used CFM that measures to what extent a customer is satisfied with (parts of) the firm.

Bearden and Teel (1983) and Oliver (1980) show a positive relationship between customer satisfaction and loyalty intentions. This relationship is confirmed by Heskett et al. (1994) who introduced the service-profit chain. They state that customer satisfaction is a driver of customer loyalty and customer loyalty drives revenue growth and profitability. Furthermore, customer satisfaction has a positive effect on retention and repurchase behavior (Yi, 1990; Mittal & Kamakura, 2001; Gustafson, Johnson & Roos, 2005; Gupta & Zeithaml, 2006). To conclude, in line with previous studies (e.g. Bearden & Teel, 1983; Oliver, 1980; Heskett et al., 1994), we expect that satisfaction has a positive impact on loyalty.

2.3.2 NPS

(17)

Page 13 Customer recommendations are important in the service industries. Earning positive word of mouth communication from customers can be a powerful force augmenting a company’s marketing efforts, especially in today’s “connected customer” contexts (Mardsen & Kirby, 2005).

We use NPS to capture the recommendation facet and measure the referral aspect of attitudinal loyalty as mentioned by Bijmolt et al. (2011). NPS is a popular CFM introduced by Reichheld (2003). It measures to what extent customers are willing to recommend the firm to friends or family on an 11-points scale. Based on these scores, customers can be put in categories ‘detractors’ (score 0-6), ‘passives’ (score 7-8) and ‘promotors’ (score 9-10). The Net Promotor Score is calculated by the percentage promotors minus the percentage detractors.

Reichheld (2003) introduced the NPS as “the one number you need to grow”. The underlying reasoning is that measuring satisfaction and retention does not help firms achieve growth but that word-of-mouth in the form of a NPS score is a better predictor. The NPS score is criticized in several researches. The superior performance of NPS as predictor for customer retention and firm performance is not proven over all industries (De Haan, Verhoef & Wiesel, 2014), because it is shown that customer satisfaction are better predictors in some industries. Research about NPS show contrary and unstable results, but despite this fact, a lot of companies use NPS as a performance metric (De Haan, Verhoef & Wiesel, 2014. As stated before, we will take into account all three CFMs to create a complete view which is comparable across the industries.

Previous studies show that customers who provide positive word-of-mouth and positive referrals are more likely to be loyal customers (Dick & Basu, 1994; Hagel & Armstrong, 1997; Srinivasan, Anderson & Ponnavolu, 2002). According to Reichheld (2003), as long as someone refers the company they validly can be labeled a loyal “customer” whether they purchase or not. To conclude, a high NPS reflects positive word-of-mouth and recommendation behavior, which results in loyal customers. So, we expect that a higher NPS results in greater loyalty.

2.3.3 CES

(18)

Page 14 than delighting customers. According to these authors the predictive power for repurchasing and increased spending is lowest for customer satisfaction, higher for NPS and highest for CES.

Contrary, de Haan, Verhoef and Wiesel (2014) have shown that CES has a limited overall incremental value because of several shortcomings. One of the shortcomings of the CES is that it only focuses on one aspect of the service, while NPS and Customer Satisfaction have a broader scope. Besides, the CES-question is only applicable for a part of the customers, namely the ones who contacted the firm with an issue. Furthermore, de Haan, Verhoef and Wiesel (2015) show that CES only has a significant predictive value in two sectors, in which it is the best predictor in comparison with satisfaction and NPS. These results are contradicting with the study of Dixon, Freeman and Toman (2010), who stated that CES has the highest predictive power. Thus, due to these contradicting results of previous results, it interesting to see what the impact is of CES in the current research.

2.4. Personality traits and relationship evaluation

It is likely that the effect of personality traits on loyalty is a sequential process which is mediated by relationship quality and relationship evaluation. But in fact, it is not possible to exclude a direct effect of personality traits on relationship evaluation and loyalty. The effect of personality traits on loyalty is already researched by Mende, Bolton and Bitner (2013). They show that both anxiety and avoidance have a negative impact on loyalty. To add knowledge to this research area, this study only focuses on the impact of personality traits on relationship evaluation, so the CFMs.

2.4.1 Attachment styles and relationship evaluation

A strong relationship leads to positive relationship outcomes (Palmatier et al., 2006). Individuals with a low level of avoidance and anxiety create strong relationships with others (Brennan, Clark & Shaver, 1998). So, overall we expect that anxiety and avoidance have a negative impact on the relationship evaluation concepts. We discuss these effects more detailed.

(19)

Page 15 In line with the effects of attachment styles on satisfaction, the impact of avoidance and anxiety on NPS is expected to be negative. Customers with a high level of anxiety fears rejection of other customers and rejection will be more likely if you do a recommendation. Moreover, attachment anxiety can lead customers to overreact to critical incidents, for example being dropped by a firm (Mende, Bolton & Bitner, 2013). This overreaction can lead to negative word of mouth and thus a low NPS. In addition, anxious persons are often involved in unpleasant relationships (Simpson 1990, p. 972), which can result in a low NPS. Furthermore, individuals with a high level of avoidance do not want to have close relationships with others, are skeptical and have an excessive need for self-reliance (Brennan, Clark & Shaver, 1998). These aspects make it reasonable that they are less willing to recommend the firm to others. Thus, in this view, we expect a negative effect of anxiety and avoidance on the NPS.

Besides, we predict attachment styles have a similar effect on CES as on the other relationship evaluation measures. An anxious customer has an excessive need for approval, and fears rejection and abandonment from the firm (Brennan, Clark & Shaver, 1998). We expect that a certain customer needs a lot of interaction moments and confirmation of the company. It costs a lot of effort to achieve this, which could be perceived by an anxious person as even more effort. People with a high level of avoidance distrust relationship partners’ goodwill, fears depending on others and are skeptical and suspicious (Brennan, Clark & Shaver, 1998). To let a firm handle your issue, you should depend on others which costs extra effort for avoidance people. Taken this into account, an issue will be perceived as much effort for an individual with a high level of avoidance. To conclude, we expect that anxiety and avoidance have a negative impact on the CES.

2.4.2 FFM and relationship evaluation

(20)

Page 16 Second, Organ and Lingl (1995) show that agreeableness "involves getting along with others in pleasant, satisfying relationships" (p. 340). In line with this research one could state that individuals with a high level of agreeableness are likely to get involved in close relationships. It is plausible that people in close, pleasant and satisfying relationships are more likely to recommend a firm to others. Furthermore, it is likely that people with a high level of neuroticism are likely to do recommendations, because they are very self-confident. Besides, extraverted people are more assertive, so individuals with a high level of extraversion will be more likely to recommend the firm to others. In line with this reasoning, we expect that Agreeableness, Neuroticism and Extraversion have a positive influence on NPS.

Finally, neurotic individuals experience more negative events than other individuals, because of their essentially negative nature (Magnus etl al., 1993). Handling an issue can be perceived as a negative event by neurotic customers due to their negative nature. Thus, we expect a positive influence of Neuroticism on CES. Furthermore, we predict that Agreeableness has a negative impact on CES, because agreeable customers have a high level of cooperativeness, flexibleness and trust which results in a relatively low perception of the effort.

2.5 Loyalty

The last step in the process of the current study is loyalty. Loyalty can be described as “a deeply held commitment to rebuy or repatronize a preferred product or service consistently in the future, thereby causing repetitive same-brand or same brand-set purchasing, despite situational influences and marketing efforts having the potential to cause switching behavior” (Oliver, 1999). In this view, van Looy (2003, p. 59) defines loyal customer behavior as: “Customer behavior characterized by a positive buying pattern during an extended period [...] and driven by a positive attitude towards the company and its products or services.”

2.6 Sector

(21)

Page 17 Furthermore, these industries have big revenues, the biggest companies have a turnover of over a few billion a year. Thus, researching the customer-firms relationships in these industries has extreme managerial relevance. Moreover, these industries have different characteristics which can impact the customer-firm relationships and thus the results. We will discuss these industry characteristics below.

The research of among others, Mende, Bolton and Bitner (2013) state that insurance services provide security to customers in times of need, which fits with the attachment theory. Furthermore insurance firms have a strong focus on CRM (Crosby, Evans, & Cowles, 1990; Verhoef 2003). To investigate differences and similarities between industries and make a comparison with previous studies possible, the insurance industry is included.

Papasolomou and Vrontis (2006) state that more banks are now beginning to look at their customer database with a view to implementing cross-sell or up-sell programmes and create loyal customers to extract more value per existing customer as this reduces customer acquisition costs. Furthermore they state that consumers are reluctant to switch because they perceive the differences to be negligible and not worth the disruption (Papasolomou & Vrontis, 2006). It is interesting to explore if we can explain this reluctance to switch in the present study.

Walsh, Groth and Wiedmann (2005) state that energy industries in most European countries undergo dramatic changes, for example the liberalization of the energy sector. This results in more competition and an increasing expenditure by energy supply companies in customer relationship management concepts, which aims at retaining existing and attracting new customers (Krafft et al. 2002). At the one hand, Zeithaml and Bitner (2010) state that energy can be regarded as an intangible, low-involvement service. On the other hand, many individuals have adopted a more socially responsible stance by moving beyond mere compliance and engaging in more environmental behavior (Williamson et al., 2006; Reeves, 2011). This is visible in investments in energy efficient houses and cars. It could be possible that these developments, especially the engaging and commitment aspects result in more word of mouth and referrals due to a high engagement.

Telecom seems to be more and more important in everyday life due to a strong desire of consumers to be connected, anywhere and anytime. This development goes hand in hand with the shift of telecom from phone calls to internet. The telecom industry faces intense competition (Lai, Friffin & Babin, 2009). In most European telecom industry the service quality, brand image and switching costs are the most important drivers of customer loyalty (Gerpott et al., 2001; Lee et al., 2001).

(22)

Page 18 To conclude, it is interesting to see to what extent the effect of personality traits on relationship evaluation and ultimately loyalty is similar over industries and to what extent it differs per industry. Due to the explorative nature of researching it across industries, we did not extend our expectations in the theoretical framework with differences per industry. This study aims at examining these similarities and differences and in this way expanding the existing literature about personality traits and loyalty.

2.7 Conceptual model

The visual representation of the conceptual model is shown in Figure 2.2. The basic idea is that personality traits affect relationship quality, relationship evaluation and ultimately loyalty. The effect of personality characteristics on loyalty is likely to be a sequential process, but we can not exclude a direct effect of personality traits on relationship evaluation. So, personality traits can impact relationship evaluation directly and indirectly through relationship quality.

In summary, the four main constructs are personality traits, relationship quality, relationship evaluation and loyalty. The big boxes represent the general concepts and the smaller boxes represent the variables used to describe these general concepts. This study aims to provide a cross-industry view, so that is why sector is included as variable as well.

(23)

Page 19

3. Methodology

3.1 Procedure

To conduct this research we have used a self-administrated questionnaire. The survey started with a welcoming page and questions about gender, age and household. The next part consisted of a general part about personality traits, so not specific for a firm. This part contained the FFM and general attachment style statements. The second section consisted of business-specific questions. This section started with business-specific attachment style items, followed by relationship quality statements. Next, the CFMs were questioned and in the end the loyalty intention measures were included. Per variable, the statements were randomized to avoid order effects.

To collect the data, we used an online panel of market research company MetrixLab. This channel had several advantages, for example it was cheaper and faster to execute an online survey than an offline survey. Besides, the online panel gave us the opportunity of randomization. In this way, questions could be randomly assigned to respondents and this improves the reliability. Moreover, the panel received survey invites via email and they could have participated voluntarily. Next, the social desirability bias was lower for online surveys than for offline surveys, because participants feel more anonymous online. This was very important in the current study, because we researched personality traits, which are sensitive to socially desired behavior. Furthermore, we had real-time access to the results and could track the type of respondents that had participated, so we could adjust the respondents who were invited, for example if a certain characteristic was not equally represented. The ultimate goal was to create a representative sample of the Dutch population according to Statistics Netherlands (statline.cbs.nl).

3.2 Subjects

The sample was a rather representative reflection of the Dutch population in terms of age structure (18 till 70 years) and gender structure according to StatLine (the electronic databank of Statistics Netherlands). The division deviated with a maximum of 7%. The sample consisted of 798 respondents of which 44.6% were male and 55.4% were female. The average age of the participants was 46.6(SD=14.3) years old. Regarding the household composition, 25.4% of the respondents were single and 67.4% were married or living together.

3.3 Measures

(24)

Page 20 To explore the explanatory effect of personality traits in this model, we used the Big Five Inventory-10 (BFI-10) to describe the FFM. The BFI-10 is a measurement instrument to measure a persons’ personality in one minute (Rammstedt & John, 2007). The BFI-10 contains ten questions and is a derivative of the BFI-44, which contains of 44 statements (John, Donahue & Kentle, 1991) and the NEO-PR-I containing 60 statements (Costa & McCrae, 1992). Rammstedt & John (2007) have shown that the BFI-10 correlates part-whole with the BFI-44 and has high reliability and validity with NEO-PR-I. The BFI-10 statements should be answered on a five-point Likert scale from totally disagree to totally agree. We use the Dutch translation of the BFI-10 provided by Denissen et al. (2008).

The general attachment style is often measured by the Experiences in Close Relationships (ECR) scale (Brennan, Clark, & Shaver, 1998) and its revised version, the ECR-R (Fraley, Wallerr, & Brennan, 2000). These two measurement instruments have 36 items to assess the attachment styles. A shorter measurement instrument is developed by Verbeke, Bagozzi and van den Berg (2013). They came up with a scale of eleven items, of which six items to measure anxiety and five items to measure avoidance. They used a scale of Shaver, which he developed most recently and received this scale through personal communication as stated in Verbeke, Bagozzi and van den Berg (2013). They also prove the reliability and validity of this short attachment style measurement instrument. We included this short version in the survey, to overcome frustration or irritation for a participant.

The business-specific attachment style measurement is developed by Mende, and Bolton (2011). They used the ECR and ECR-R as the fundament for their instrument and conducted three scale development studies in different service contexts. They came up with four items for the two dimensions anxiety and avoidance. They demonstrated high reliability and validity of the instrument. This is confirmed in the research about business-specific attachment styles of Mende, Bolton and Bitner (2013). Thus, we adopted this measurement in the current study.

Relationship quality consisted of the constructs trust and commitment. Both concepts has been measured by three items. These measures were also used in previous studies about attachment styles, so it makes it more reasonable and valid to compare the results. The three items to measure trust were developed by Doney and Cannon (1997) and also used in Mende and Bolton (2011) and Mende, Bolton and Bitner (2013). Furthermore, the three commitment measures were used in the same two researches but are founded by Coulter, Price and Feick (2003) and Gruen, Summers, and Acito (2000).

(25)

Page 21 am satisfied with [firm X]” on a 5-points Likert-scale according to Aaker, Foumier, and Brasel (2004). Furthermore, Reichheld (2003) showed that the referral value can be addressed by the question “how likely are you to recommend [firm X] to a friend, relative or colleague on a scale from 0 to 10?” whereby 0 is extremely unlikely and 10 is extremely likely. Finally, CES was introduced by Dixon, Freeman and Toman (2003) with the question “How much effort did it take for you to ensure that [firm X] handled your issue”. Reviews about this measure stated that it can be improved on three aspects. Both the scale and wording can cause misinterpreting, the word ‘effort’ does not translate neatly into all languages and it has lack of benchmarking capabilities. To deal with these problems CES 2.0 was developed. CES 2.0 is not a question, but a statement, namely: “[firm X] made it easy for me to handle my issue”. A 5-points scale is used, from 1 (strongly disagree) to 5 (strongly agree). We included CES 2.0 and prior to the CES 2.0 question, we asked the respondents if they had submitted a question or request in the past year at [firm X], because the CES is only applicable for customers who had an issue.

Loyalty has been measured by two types of behavioral intention. Behavioral intention is a good predictor of real behavior according to Fishbein and Yzer (2003). In a service-context the behavioral intention is measured by an intention to switch and an intention to stay, because the customer-firm relationship in the service industry is mostly contract-based and for example a yearly or a two-yearly agreement. In fact, intention to stay reflects retention, while intention to switch reflects churn. These kinds of intentions are the most widely used indicators of customer loyalty in firms’ customer feedback systems (Chandon et al. 2005) and even primary customer loyalty metrics (Kamakura et al. 2002; Mittal et al. 1998). One could state that the retention intention and churn intention are inverse intention measures. If one scores high on retention intention, it is likely that he scores low on churn intention. However, consumers are not always consistent in their answers and behaviors. In this view, we used both measures in the study.

(26)

Page 22

3.4 Plan of analysis

To investigate the effect of personality traits on relationship quality, relationship evaluation and ultimately loyalty, several statistical tests were used.

The first step after collecting the data was to test the validity. The selected items are mainly adapted from prior studies to ensure content validity, as most have been previously tested for internal consistency and validity (Heere et al., 2011). We conducted a Cronbach’s alpha reliability test to check the internal consistency of the different items per variable. According to Malhotra (2009, p.319) a minimum value of 0.6 is required to qualify the internal consistency to be sufficient. If the items can be combined, one construct was used in the further analyses. It was important to check if the churn intention and retention intention are inverse and could be combined to one loyalty measure or that these measures should have been interpreted as two separate measures, namely churn intention and retention intention. Furthermore, the correlation of different variables were checked.

The second step of the analysis was to perform regression analyses to investigate the effects of personality traits on relationship quality, relationship quality on relationship evaluation and personality traits on relationship evaluation. A regression analysis investigates if there is a relationship between the independent variable and the dependent variable based on correlation. This analysis makes use of continuous independent and dependent variable. When developing a regression model, it is important to test several assumptions (Leeflang, Wieringa, Bijmolt & Pauwels, 2015, p. 100). These assumptions for the disturbance term are independence of the disturbance term, homoscedasticity, normality and lastly regression assumes that the mean of the disturbance term is zero. So, we checked these assumptions in the present study. In addition, these regression analyses also provided insights about which personality trait variable has the greatest explanatory and predictive power. The explanatory power was addressed by comparing the R-squares and significant effects of the models of the different personality traits. The R-squares entailed the percentage of the model that was explained by the variance of the independent variables (Leeflang et al., 2014).

(27)

Page 23 with the mediator (M), reflected in model 2: M = b2 + a X + e2. This estimation (a) was the first part of the indirect effect. The second part of the indirect effect (b) showed that the mediator affects the dependent variable, shown in model 3: Y = b3 + b M + e3. Finally, the last step was to establish the degree of mediation, reflected in model 4: Y = b4 + c' X + b M + e4. Due to multiple mediators, the mediation macros of Hayes (2013) has been used.

Furthermore, we used a bootstrapping method. Bootstrapping is a non-parametric method based on resampling with replacement and directly tests ab (Kenny, 2011). Bootstrapping allows researchers to take a subsample from the original sample in order to generate an empirical sampling distribution (Hayes, 2009). This distribution can be used, along with the standard error of the bootstrap estimates to compute a Z-statistic or confidence interval (Hayes, 2015). According to Preacher and Hayes (2008), the bootstrap with 5000 replications is the best method, so we perform this type of bootstrapping. We used the original sample as population and sample with replacement a sample of size N. Next, we estimated ab for that sample. In the end we examined the distribution of the obtained estimates, and determined the 2.5th_{and 97.5}th_{percentiles with α=0.05. The advantage of bootstrapping is that a}

confidence interval, a p-value, and a standard error can be determined (Kenny, 2011). The confidence interval was computed in order to check if zero was in the interval. If zero was not in the interval, we could have been fairly sure that the indirect effect was different from zero.

Figure 3.1. Mediation analysis

(28)

Page 24 test developed by Breusch and Pagan (1979). This analysis tested whether the estimated variance of the residuals from a regression are dependent on the values of the independent variables. In that case, heteroscedasticity was present and the SUR was the appropriate method.

Fourth, we conducted a robustness check to assess the robustness of our results. According to Little (1970) model criteria a model should be robust. Therefore, we recoded the continuous loyalty variables into dichotomous variables with a one if the event (retention/churn) happens and a zero when the event does not happen. Two binary choice regressions are considered, the logit model and probit model (Malhotra, 2009, p.620). They have similar cumulative distribution functions and the results are very similar. We chose to perform a logit model, because the probabilities are easier to calculate, the interpretation of the parameters is easier and the logit is more commonly used (Malhotra, 2009, p.620). The logit estimates the parameters in such a way that it maximizes the likelihood or probability of observing the actual data. If p is a probability, then p/(1 − p) is the corresponding odds (Malhotra, 2009). The logit of the probability is the logarithm of the odds. The results of the standard regression and logit were compared and can be called robust when similar results appear for both analysis in terms of direction and relative size.

The final step was a meta-analysis to get overall insights, next to the already gained insights of the separate industries. We only performed the meta-analysis when significant differences between industries were observed. To get these insights an Added Z method was used. Stouffer et al. (1949) developed the Z-transform test. This method combines p-values from the separate industry analyses to test whether collectively they can reject a common null hypothesis (Whitlock, 2005). The Z-transform test takes advantage of the one-to-one mapping of the standard normal curve to the p-value of a one-tailed test. By Z we mean a standard normal deviate, that is, a number drawn from a normal distribution with mean 0 and standard deviation 1. Any value of p will uniquely be matched with a value of Z. Whitlock (2005) shows that the Z-transform test converts the one-tailed p-values, Pi, from

each of k independent tests into standard normal deviates Zi. The sum of these Zi’s, divided by the

square root of the number of tests, k, has a standard normal distribution if the common null hypothesis is true. Furthermore, we generated the weighted response parameter which belongs to the overall p-value. The formula of the Added Z and the Beta that belongs to it are presented below in figure 3.2.

Figure 3.2. Formulas meta-analysis.

𝐴𝑑𝑑𝑒𝑑 𝑍 = ∑ 𝑍 − 𝑣𝑎𝑙𝑢𝑒𝑠

√𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑑𝑢𝑠𝑡𝑟𝑖𝑒𝑠 𝑊𝑒𝑖𝑔ℎ𝑡𝑒𝑑 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 =

(29)

Page 25

4. Results

4.1 The data

4.1.1 Data cleaning

After collecting the data, we cleaned the data which included consistency checks and treatment of missing responses (Malhotra, 2009). Two respondents were screened out, because they answered “I do not know” to the questions about of which company they are a customer for all four sectors. Furthermore, we did not find any extreme outliers or inconsistencies.

4.1.2 Reliability

The first step after collecting and cleaning the data was to conduct a Cronbach’s Alpha reliability test to check the internal consistency reliability of the different items per variable. According to Malhotra (2009, p.319) a minimum value of 0.6 is required to qualify the internal consistency to be sufficient. Before performing this reliability test, the items of general avoidance were recoded, so that a high score means a high level of avoidance. For general anxiety we already used the items in a way that a positive score reflects a high level of anxiety. The four items measuring the general attachment style anxiety have a Cronbach’s Alpha of 0.81 and the items measuring general avoidance have a Cronbach’s Alpha of 0.76. So, we created a variable for anxiety by averaging the items measuring anxiety and a variable for avoidance by averaging those items.

Next, we also tested the reliability for the business-specific attachment styles, of which the Cronbach Alpha values are shown in Table 4.1. All values were above 0.6 and thus we combined and averaged the four statements per dimensions into one reliable item per dimension.

Table 4.1. The reliability tests of the business-specific attachment styles.

Bank Insurance Energy Telecom

Anxiety 0.85 0.86 0.84 0.83

Avoidance 0.81 0.80 0.81 0.79

(30)

Page 26 The statements of the relationship quality constructs can also be combined to one item for trust and one item for commitment. The Cronbach’s Alpha values of trust for bank, insurance, energy and telecom are respectively 0.85, 0.82, 0.80 and 0.79. The Cronbach’s Alpha values of commitment for bank, insurance, energy and telecom are respectively 0.89, 0.86, 0.87 and 0.88. The relation evaluation concepts and the churn intention and retention intention were only measured with one item, so a Cronbach’s Alpha test was not applicable.

The second step was to check the correlations between the general attachment style and the business-specific attachment style. We checked this correlation by performing a bivariate correlation analysis, which tests the strength of association between two metric variables (Malhotra, 2009). According to Cohen (1988) correlations below (-) 0.29 are small, correlations between (-) 0.30 and (-) 0.49 are medium and from 0.50 onwards the correlations can be seen as large. The general attachment style anxiety correlated significantly with the business-specific attachment style anxiety in all industries, but the correlation was low: 0.25, 0.25, 0.32 and 0.27 for respectively bank, insurance, energy and telecom. The general avoidance attachment style only correlated significant with the insurance avoidance attachment style, namely 0.14 (p<0.01), which was also a low correlation value. Thus, in contradiction with the expectations, the general and specific attachment styles correlations were low or absent. This means that both concepts measured a different aspect of personality which were not congruent. Instead of interpreting the specific attachment styles as an extension of the general attachment style, we have to interpret it as a separate construct.

Next, we investigated the correlations between the general attachment style and the FFM. Both anxiety and avoidance correlated significant with Extraversion, Agreeableness, Conscientiousness and Neuroticism, but not with Openness. Anxiety correlated highest with Neuroticism -0.44 (p<0.01) and avoidance highest with Extraversion, namely -0.49 (p<0.01). These correlations can be labeled as medium according to the thresholds of Cohen (1988). The other significant correlations were very low. This means that the constructs FFM and attachment styles measured a similar part of personality traits to a certain extent, but also showed some differences. They had more similarities with each other than with the business-specific attachment styles.

(31)

Page 27 correlation then would be -1.00, but they were inverse to a great extent. To conclude, the measures still differ which shows that it is relevant to take both measures into account when assessing loyalty.

4.1.3 Descriptives

The respondents had to fill in the name of the company of which they are a customer and had the option “do not know”. Per customer two industries are randomly chosen of which the “do not know” answers are excluded to make sure participants can be questioned about a specific firm. 6 participants did not knew their bank, 64 their insurance company, 60 their energy company and 12 their telecom company. In the end, 425 respondents were assigned to bank, 368 to insurance, 397 to energy and 403 to telecom. Thus, it was harder for participants to fill in their insurance and energy company than their bank or telecom firm. The participants are rather equally distributed across the industries.

The descriptives in terms of averages and standard deviations of the personality trait FFM are presented in Table 4.2. The general attachment styles and the business-specific attachment styles are shown in Graph 4.3. More detailed information can be found in Table B1 in Appendix B.

Table 4.2. Descriptives personality traits. Graph 4.3. Descriptives attachment styles.

In order to gain more insights about the general attachment styles, we developed several distributions to see what part of the sample has an anxious, avoidant or secure attachment style. We based these divisions on the new attachment variables that we created, which are the averages of the statements belonging to that variable. We tried to describe the respondents in terms of attachment styles by recoding the anxiety and avoidance variables into a categorical variable whereby a respondent has the specific style or not or in some cases can be neutral. The divisions are similar for anxiety and avoidance.

The first division stated that a respondent with an average score of 3.1 and larger is anxious/avoidant. A score of 2.9 or lower reflects a non-anxious or non-avoidant attachment style. Between 2.9 and 3.1 the respondents are called neutral, because they stated that the statements are not applicable to them, but they also did not say the statements are not not applicable to them. The second division applies strong thresholds, namely a score of 4 and larger reflects an anxious/avoidant attachment style while a score of 2 and lower reflects a non-anxious or non-avoidant attachment style. This results in a

0 1 2 3 4

(32)

Page 28 large group of respondents with a neutral attachment style. The third division is similar to the first division, from 3.1 onwards a respondent has an anxious or avoidant attachment style, below 3.1 a respondent is non-anxious or non-avoidant. In this way, no people are categorized as neutral. Graph 4.4 shows the results of these three divisions.

Graph 4.4 shows that the first division results in 17% being anxious and 28% being avoidant. Furthermore, three out of four people are anxious and almost two out of three people being non-avoidant. A small percentage of the sample is neutral. The second division shows that strong thresholds result in a large part of the sample being neutral. The percentage of people being anxious becomes very small and the percentage of people being avoidant decreases as well. The third division ensures that a very large part of the respondents is non-anxious and/or non-avoidant.

To conclude, we decided to continue with division 1, because in this way we take into account respondents who are neutral. It is important to deal with the neutral group, because they did not answered in a way we can state they are anxious/avoidant, but we can also not state that they are non-anxious or non-avoidant. Moreover, the strong thresholds ensure that more than half of the respondents are labeled as neutral, which has low face validity and makes it hard to perform further analyses. Thus, we describe the next distribution based on division 1.

Graph 4.4. Three division of the general anxious and avoidant attachment style.

(33)

Page 29 and avoidant, he was labeled as anxious and avoidant. If a person scores neutral on both dimensions, he was labeled as neutral. The results are shown in graph 4.5.

Graph 4.5 shows that fewer people were secure in terms of business-specific attachment style in comparison with general attachment style. These differences can be explained by a different attachment target, namely other people for the general attachment style and a specific firm for the business-specific attachment style. As a consequence of this, different measurement instruments are used which makes it harder to compare both distributions. Another explanation is that an interpersonal relationship differs from a customer-firm relationships in interests. Whereby the interpersonal relationship the interests are more likely to be similar, in the customer-firm relationship the firm and the customer have different interests. The firm wants to make profit and the customer wants a certain service.

Next, we discuss the differences between the business-specific attachment styles across industries. People were most secure in the insurance industry and least in the bank sector. This is interesting, because research about business-specific attachment styles so far is only conducted in the insurance industry. The economic crisis can be an explanation for the low percentage of secure customers in the bank sector. Furthermore, many respondents were avoidant in the business-specific attachment style in the energy and telecom sector in comparison with the insurance and bank sector. A reasoning behind this finding is that customers care more about their financial services and want to have a stronger relationship with their bank or insurance company. The division of anxious, anxious and avoidant and neutral respondents were very similar across the industries and these customers represent a small part of the sample. Anxious and anxious and avoidant are extreme attachment styles, so it is reasonable that only a small group of people has this attachment style, probably they show this across industries and not for a specific industry or firm.

(34)

Page 30 Furthermore, the descriptives of the relationship quality construct, CFMs and loyalty per industry are shown in Table B2, B3 and B4 in Appendix B. The graphs show that trust is relatively low in the bank industry and similar in the insurance and energy sector. The average scores on commitment were very similar over the industries. Furthermore, the scores of customer satisfaction, NPS and CES were almost identical over the industries. The churn intention and retention intention differed barely between industries.

4.2 Main effect personality traits on relationship quality

In order to analyze to what extent the personality traits influence the relationship quality, we regressed trust and commitment on the general attachment style, business-specific attachment style and FFM. We chose to perform regression analyses, because regression analysis is a powerful and flexible instrument for analyzing associative relationships between a metric dependent variable and one or more independent variables (Malhotra, 2009). To compare sizes and directions of effects, we use the standardized beta coefficients. The results are presented in the following sections.

4.2.1 Main effects general attachment style on relationship quality

Table B5 and B6 in Appendix B show the results of the regression analyses of the general attachment style on trust and commitment. The results show that anxiety has a significant negative influence on trust in the bank industry (Beta=-0.13, p=0.02) and that avoidance has a significant negative influence on trust in the insurance industry (Beta=-0.18, p=0.00). Furthermore, we can conclude that avoidance has a significant negative impact on commitment in the insurance industry (Beta=-0.21, p=0.00). The other dimensions and industries do not show a significant effect of general attachment style. To conclude, the effect of general attachment styles on relationship quality is negligible. The highest adjusted R² is 5%, which means that the general attachment styles at most explain 5% of the variance, which is a very low percentage. So, the general attachment styles are not a good way to explain differences in relationship quality.

4.2.2 Main effects business-specific attachment style on relationship quality

(35)

Page 31 Table 4.6. The effect of business-specific attachment styles on trust.

Industry Adj. R² F (sig.) Predictor Beta

Bank 0.57 279.46 (p=0.00) Anxiety Avoidance -0.38*** -0.61*** Insurance 0.55 227.25 (p=0.00) Anxiety Avoidance -0.39*** -0.57*** Energy 0.43 150.92 (p=0.00) Anxiety Avoidance -0.31*** -0.52*** Telecom 0.48 185.62 (p=0.00) Anxiety Avoidance -0.33*** -0.61*** * < 0.10, ** < 0.05, *** < 0.01

Table 4.7. The effect of business-specific attachment styles on commitment.

Industry Adj. R² F (sig.) Predictor Beta

Bank 0.51 224.74 (p=0.00) Anxiety Avoidance -0.29*** -0.74*** Insurance 0.50 185.98 (p=0.00) Anxiety Avoidance -0.28*** -0.72*** Energy 0.53 219.42 (p=0.00) Anxiety Avoidance -0.21*** -0.80*** Telecom 0.43 152.75 (p=0.00) Anxiety Avoidance -0.27*** -0.75*** * < 0.10, ** < 0.05, *** < 0.01

4.2.3 Main effects FFM on relationship quality

Table B7 and B8 in appendix B show the results of the regression analyses of trust and commitment on FFM. The results show that Extraversion has a significant positive effect on trust in the bank industry (Beta=0.14, p=0.01) and a marginal significant effect on trust in the insurance industry (Beta=0.09, p=0.09). Besides, Agreeableness has a significant positive effect on trust in the bank sector (Beta=0.12, p=0.03), insurance sector (Beta=0.11, p=0.04), marginal significant effects in the energy sector (Beta=0.09, p=0.07) and telecom sector (Beta=0.10, p=0.08). Neuroticism has a marginal significant effect on trust in the insurance industry (Beta= 0.07, p=0.06). Thus, Extraversion, Agreeableness and Neuroticism are the three most important FFM dimensions. Next, Agreeableness has a positive significant effect on commitment in the insurance sector (Beta=0.15, p=0.02) and Extraversion has a marginal significant positive effect on commitment (Beta=0.11, p=0.08).The other dimensions and industries do not show a significant effect of FFM. Thus, Agreeableness and Extraversion are important FFM dimensions if we look at commitment.

(36)

Page 32 which is very low. This means that the FFM is also not a very good explanatory variable for relationship quality relative to the explanatory power of business-specific attachment style.

4.3 Main effects personality traits and relationship quality on CFMs

To see to what extent the CFMs are influenced by personality traits and relationship quality, we regressed the CFMs on all of these variables. To compare sizes and directions of effects, we use the standardized beta coefficients. The results are shown in Table B9, B10 and B11 in Appendix B.

First, we discuss the effect on customer satisfaction. In all industries the effects of trust, commitment and business-specific anxiety are significant. The first two concepts have a positive effect on customer satisfaction, while business-specific anxiety has a negative effect on customer satisfaction. Furthermore, there is one remarkable effect of the FFM in the insurance sector. Namely, the FFM dimension Conscientiousness has a negative effect on customer satisfaction in the insurance industry. This means that if someone is has a high dependability and perseverance, he is less satisfied with his insurance company. Overall, we can state that trust has most positive impact on customer satisfaction, followed by commitment. Contrary, business-specific anxiety has a negative impact on customer satisfaction. So, relationship quality is an important predictor of customer satisfaction which can be complemented with business-specific anxiety as personality trait variable.

Find your perfect match!

MASTER THESIS

Larissa Hendriks