Relationship Marketing: Insight in the Structure and Outcomes in a B2B Context

Relationship Marketing: Insight in the

Structure and Outcomes in a B2B Context

Herm Jan Mateboer

Student number: 1461605

Relationship Marketing: Insight in the

Structure and Outcomes in a B2B Context

Master Thesis

Department: Marketing Management and Market Research

Completion date: 09 April 2009

Herm Jan Mateboer

Supervisors

Meester Kosterweg 10

Prof. Dr. Peter C. Verhoef

7914RG Noordscheschut

Drs. Ernst Osinga

0528-343494

hermjanmateboer@hotmail.com

External supervisor

Management Summary

Table of Content

1

Introduction and Problem Statement ... 5

1.1 Introduction ... 5

1.2 Motivation ... 7

1.3 Problem Statement and Research Questions ... 7

1.4 Organization ... 7

2

Literature Review ... 9

2.1 Behavioural outcomes of relationship marketing ... 9

2.2 Mediators of a relationship in an contractual B2B context ... 10

2.2.1 Commitment ... 10

2.2.2 Trust ... 12

2.2.3 Satisfaction ... 13

2.3 Antecedents of a relationship in an contractual B2B context ... 14

2.3.1 Customer-focused antecedents ... 15 2.3.2 Seller-focused antecedents ... 17 2.3.3 Dyadic antecedents ... 19

3

Method ... 22

4

Results ... 26

4.2.2 Breadth of the relationship ... 30

1 Introduction and Problem Statement

1.1

Introduction

According to Grant and Schlessinger (1995) there is an enormous gap between a companies’ current and full-potential profitability. This is due to the fact that organizations only focus on acquiring new customers. To close the gap, companies should not focus only on acquiring new customers, but also on enhancing the profitability of existing customers and extending the duration of customer relationships. This may explain the dramatic growth of relationship marketing (RM) in the past decade (Srinivasan and Moorman 2005; Palmatier et al. 2006). The definition of RM is defined by Morgan and Hunt (1994) as “all marketing activities directed towards establishing, developing, and maintaining successful relational exchanges.” Most researchers assume that RM investments will lead to stronger customer relationships which will eventually lead to improved seller performance. RM is aimed at the benefits of long-term relationships versus short-term relationships. Research with respect to the effectiveness of RM has yielded different outcomes; some studies show empirical support for the improved effectiveness of seller performance through RM efforts (Crosby, Evans, and Cowles 1990; Morgan and Hunt 1994), while other studies show no support. This may imply that the effects of RM are context dependent (Crosby, Evans, and Cowles 1990).

Relationships can be characterized along three dimensions: length, breadth and depth. Length refers to the retention of customers, while the depth of the relation refers to the deepening of the customer’s relationship with the firm by increased usage or upgrading. The depth of the relation is closely related with customer share, which is defined by Peppers and Rogers (1999) as “the ratio of a customer’s purchases of a particular category of products or services from supplier X to the customer’s total purchases of that category of products or services from all suppliers.” However, customer share is an overall measure of a customer relationship. In this paper the share relates only to one product and can therefore be defined as product share. De Wulf, Odekerken-Schröder and Iacobucci (2001) found strong support for a positive influence of relationship marketing on behavioural loyalty. Verhoef (2003) confirmed that customer relationship management instruments lead to increased retention and customer share. At last, the breadth of the relationship refers to the expansion of the customer relationship through cross-buying. Verhoef, Franses and Hoekstra (2001) recognized the fact that creating value by expanding the range of articles purchased is an important aspect of customer relationship management. Prior literature suggests that relationship marketing can be an effective strategy to increase customer-value along the three dimensions (e.g. Verhoef, van Doorn and Dorotic 2007).

relational outcomes is according to much research (e.g. Palmatier et al. 2006; Morgan and Hunt 1994; Garbarino and Johnson 1999) mediated by variables such as trust, commitment, relationship quality and relationship satisfaction. These mediators have different influences on the dimensions of a customer relationship. These mediators can be important to gain insight in RM and some mediators will therefore be investigated in this paper.

The outcomes of RM effort can be context dependent (Crosby, Evans, and Cowles 1990). Therefore, the context of this paper will be addressed. The context of this paper is a B2B contractual context. The outcomes of the study of Palmatier et al. (2006) indicate that RM researchers may need to take a multiple mediator when they measure customers’ relationships to capture their impact on performance. However, there is not much academic literature of the relative importance of these mediators across industries, and subsequently there outcomes. One could expect that in a contractual situation, commitment and trust should be more important than in a non-contractual setting. With respect to trust, it plays an important role before signing a contract or when a contract is almost expired. In the period during the contract the service is legally guaranteed by the contract so trust may not play an important role then. This is comparable to the distinction between discrete transactions and relational exchanges. Discrete transactions have a “distinct beginning, short duration, and sharp ending by performance,” and relational exchanges which “traces to previous agreement *and+ ... is longer in duration, reflecting an ongoing process” (Dwyer, Schurr and Oh 1987). A contractual setting seems to fit to the definition of relational exchanges. Successful relational exchanges create barriers to switch and therefore have a positive influence on relationship building. A contractual setting has much in common with the description of relational exchanges and it can be expected that this context is suitable for RM.

1.2

Motivation

The organization which is researched is a mid-size insurance company. The organization aims at a niche market, namely all insurances with respect to the professional transport industry. As a consequence, it can be concluded that the organization is a multi-service provider.

The motivation of this study is the fact that when an organization wants to change its focus from sales driven to market driven it has little knowledge about the needs of the customer with respect to relationships. As Day (1994) stated, the marketing concept of “stay close to the customer” has been more an article of faith than a practical basis. Therefore, instead of “learning from defections” (Reichheld 1996), Bolton (1998) suggests that service companies should act proactive and learn from customers before they defect. Currently there is scarce knowledge about the antecedents of relationships and their effect on the dimensions (length, breadth and depth) of a relationship. As a consequence, the organization has not identified the important antecedents of their relation with its customers and has no knowledge which antecedents and mediators have an effect on the effectiveness of the seller performance outcomes. The focus of this study will be on small businesses and entrepreneurs (SB&E), in this case it means customers with ten or less vehicles.

1.3

Problem Statement and Research Questions

In order to be able to answer the problem stated in the motivation, the problem statement of this study is:

In order to answer the specific components of the problem statement, three research questions are formulated:

1. Which antecedents are most effective with respect to RM in a B2B contractual context? 2. What is the effect of RM efforts on the dimensions of a customer relationship?

3. Do variables, such as trust mediate in the relationship between antecedents and behavioural outcomes of RM?

1.4

Organization

The organization of this paper is as follows. In section 2 the conceptual model and the hypotheses are presented. The research design and the data are presented in section 3. The results are described

2 Literature Review

In this section, the behavioural outcomes of relationship marketing will be discussed. Thereafter, the effect of the mediators on these outcomes will be discussed. The choice of relationship antecedents will be discussed and hypotheses are proposed to assess the effect of these antecedents on the mediators.

2.1

Behavioural outcomes of relationship marketing

In this study, the effect of the antecedents of a customer-supplier relationship on the behavioural outcomes of this relationship will be discussed. In this section the behavioural outcomes will be established and discussed.

The outcome which is most associated with RM is customer loyalty. Loyalty is a primary goal of RM and is sometimes even equated with the relationship marketing concept itself (Sheth 1996 in Hennig-Thurau, Gwinner and Gremler 2002). Day (1969) stated that loyalty can be divided in a rational and an affective component. The rational decision is made after an evaluation of competing brands, and once made the buyer feels that such a decision is not longer necessary, unless competitive or other circumstances change. Purchasing the brand becomes habitual which creates a strong affective orientation. This result in a narrowing of his perceptual view and reduce the effect of competitive promotional activity. The effect of loyalty and profitability has been shown by Reichheld and Sasser (1990), who found steep upward sloping profit curves when defections decreases. In a contractual setting, organizations can build a relationship with the customer for at least the contract time. This may be favourable for a relationship because a relationship exists when an individual exchange is assessed not in isolation but, as a continuation of past exchanges likely to continue into the future (Czepiel 1990). This is in contrast to discrete transactions which are evaluated individually, without any reference to past or future transactions (Bendapudi and Berry 1997).

The effectiveness of cross-buying seems therefore context dependent and more effective in a contractual setting.

The last dimension of a relationship which will be discussed is the depth of a relationship. The effect of RM on the purchased percentage of product X at a company will be investigated. The effect of satisfaction on the share of wallet and product usage has been researched by several authors (Mägi 2003; Bowman and Narayandas 2001; Bolton and Lemon 1999). However, the direct effect of relationship marketing is less apparent but the effect of relationship marketing on satisfaction is more obvious. Palmatier et al. (2006) found that relationship antecedents have a positive influence on the relationship satisfaction.

2.2

Mediators of a relationship in an contractual B2B context

For decades, the most important metric for predicting consumer behaviour was overall satisfaction. However, the shifting emphasis on building relationships with customers has added other factors which predict intentions of customers. These metric are mediators such as trust and commitment. The effects of commitment, trust, relationship satisfaction and relationship quality are extensively researched (Morgan and Hunt 1994; Garbarino and Johnson 1999; Palmatier et al. 2006). With respect to these mediating variables, there is not much agreement about the importance of these variables. Some research indicated that both trust and commitment are important to predict exchange performance (Morgan and Hunt 1994), while other research suggests that trust (Doney and Cannon 1997), commitment (Jap and Ganesan 2000) or relationship satisfaction (De Wulf, Odekerken-Schöder and Iacobucci (2001) are key predictors. Palmatier et al. (2006) suggested that different dimensions of a relationship may be synergistic. Superior performance can only be expected if all mediators are sufficiently strong for all dimensions.

2.2.1 Commitment

the study may become too complex or ambiguous. Commitment is regarded as an essential ingredient for successful long-term relationships (Dwyer, Schurr and Oh 1987). They investigate buyer-seller relationships and identified commitment as a phase in relationships. It is the most advanced phase of buyer-seller interdependence. Scanzoni (1979) distinguishes three criteria to measure commitment: (1) inputs, (2) durability and (3) consistency. Inputs consist of the fact that both exchange parties provide relatively high levels of inputs to the association (Blau 1964 in Morgan and Hunt 1994). Durability of the relationship is the extent to which environmental change is necessary to ensure a relationship and the willingness of exchange partners to make such adjustments (Scanzoni 1979). The last criterion is the consistency with which the inputs are made to the association (Dwyer, Schurr en Oh 1987). These criteria have much in common with the essence of commitment, described by Anderson and Weitz (1992), namely stability and sacrifice. Exchange partners are willing to make short-term sacrifices in order to realize long-term benefits (Dwyer, Schurr en Oh 1987). The effect of commitment on the length of the relationship is extensively researched. Prior research has shown that there is a positive relationship between commitment and relationship duration. Verhoef (2003) found a positive relation between (affective) commitment and customer retention. The positive effect of commitment on customer loyalty is established by several studies (Hennig-Thurau, Gwinner and Gremler 2008; Pritchard, Havitz and Howard 1999). With respect to calculative commitment, this type of commitment can positively influence the relationship’s length due to high termination or switching costs. Therefore, I expect a positive influence of commitment on the length of the relationship.

Hypothesis1a: Commitment has a positive relationship with the length of a relationship.

With regard to the breadth of the relationship, Bolton, Lemon and Verhoef (2004) treated commitment separately in calculative and affective commitment. They propose no effect of calculative commitment on the breadth of the relationship, since this is based on economic arguments. In contrast, customers with a high affective commitment have warm feelings towards the supplier and tend to purchase additional products or services. Verhoef, Franses and Hoekstra (2002) found no effect of commitment on cross-buying. I will follow their findings and expect no relation between commitment and breadth of relationship. Therefore, this relation will not be researched.

of high levels of affective commitment. Therefore I will not investigate the relation between commitment and depth of relationship.

It may be noticed that trust and affective commitment seems closely related when considering the definitions of both constructs. This may be the reason why the relations between commitment or trust and the dimensions of the relationship are tested in other research. However, in contrast to expectations, the relations between commitment and the dimensions of the relationship were found non-significant. This lack of evidence and, in addition, expert interviews within the firm has lead to the decision that the relations concerning commitment will not be tested.

2.2.2 Trust

Trusting parties must be vulnerable to some extent for trust to become operational (Doney and Cannon 1997). This means that outcomes must be uncertain and important to the trustor. When focusing on distribution channels, trust is an important determinant, due to the high degree of interdependence and the high switching costs. Trust is defined by Moorman, Desphandé and Zaltman (1993) as “the willingness to rely on an exchange partner in whom one has confidence”. Morgan and Hunt (1994) define trust as the perception of “confidence in the exchange partner’s reliability and integrity”. The key aspects of trust are highlighted by both definitions: confidence and reliability. The literature suggests that confidence leads to the firm’s belief that the trustworthy party is reliable and has high integrity. This is associated with several qualities, identified by Morgan and Hunt (1994): consistent, competent, honest, fair, responsible, and helpful. These qualities give a general view of trust and are more accessible to consider. With respect to a B2B context, Anderson and Narus (1990) define the perceived outcomes of trust as “the firm’s belief that another company will perform actions that will result in positive outcomes for the firm as well as not take unexpected actions that result in negative outcomes”. This is in line with the statement of Berry and Parasuraman (1991; in Morgan and Hunt 1994): “Effective service marketing depends on the management of trust because the customer typically must buy a service before experiencing it.” With regard to relationships, trust seems, similar to commitment, an essential factor for successful long-term relationships.

Morgan and Hunt (1994) investigated the effect of trust on the propensity to leave. They found strong support for a negative relationship. When the trust in a firm increases, propensity to leave decreases.

Trust has been considered to have a beneficial influence on the development of positive customer attitudes, intention and behaviour (Soureli, Lewis and Karantinou 2008). This is indeed found by Soureli, Lewis and Karantinou (2008) and Crosby, Evans and Cowley (1990), who suggests that trust has a positive effect on cross-buying intentions. Bendapudi and Berry (1997) developed a model with variables which lead to cross-buying. According to them, trust influences customers’ dedication to the provider and their willingness to enhance the relationship.

Hypothesis2b: Trust has a positive relationship with the breadth of a relationship.

At last the effect of trust on the depth of the relationship will be considered. The effect of trust on the depth of the relation is less obvious. Therefore, the definition of trust is considered as well as the definition of perceived outcomes by Anderson and Narus (1990). Trust is the willingness to rely upon another party. When buyer increase the number of products of services purchased, their dependency on the supplier increase. In conclusion, I expect trust to be a positive determinant of the depth of the relationship. This is line with Anderson and Narus (1990) who stated that when the buying party relies more on the supplier, he become more dependent and therefore should trust the supplier.

Hypothesis2c: Trust has a positive relationship with the depth of a relationship.

2.2.3 Satisfaction

According to Palmatier et al. (2006): “Relationship satisfaction reflects exclusively the customer’s satisfaction with the overall exchange”. Anderson and Narus (1990) define satisfaction as “a positive affective state resulting from the appraisal of all aspect of a firm’s working relationship with another firm”. Satisfaction is here defined as an affective state; this is in contrast to Thibaut and Kelley (1959 in Anderson and Narus 1990) who considered satisfaction as a more rational summary assessment of outcomes by comparing with other outcomes. The primary determinant of satisfaction is perceived quality or performance.

element of loyalty; however particularly after loyalty bas been established, it is difficult to entertain loyalty development without satisfaction. Therefore, I propose:

Hypothesis3a: Satisfaction has a positive relationship with the length of a relationship.

The research of the effect of satisfaction on cross-buying has been studied by Verhoef, Franses and Hoekstra (2001) who found no empirical support for this relation. They found that higher satisfaction will not lead to the purchase of additional services while less satisfaction will lead to the abundance of services already purchased. If customers perceive a company’s service offerings as similar, satisfaction of the current used service will transfer to the expected satisfaction levels of potential products (Bolton, Lemon and Verhoef 2004). This, in turn, will positively effect cross-buying. When customers expand their range of products, they become more dependent of the supplier and have higher switching costs. Therefore, I expect that dissatisfied customers will not broaden their range of products. In contrast, satisfied customers are more eager to purchase additional products.

Hypothesis3b: Satisfaction has a positive relationship with the breadth of a relationship.

With respect to the depth of the relation, Bolton and Lemon (1999) and Bolton, Kannan and Bramlett (2000) found support for the positive relation between satisfaction and service usage. Bolton and Lemon (1999) developed payment equity, which is defined as the perceived fairness of the price/usage trade-off. Payment equity is effecting customer satisfaction which in term is positively related to product usage. In addition, Reynolds and Beatty (1999) show a significant effect of satisfaction on the share of purchases. Bolton, Kannan and Bramlett (2000) suggest that “customers make repatronage decisions on the basis of their prior repatronage intentions or behaviour, updated by comparisons of their prior satisfaction levels with the company versus their satisfaction with a competitor.” Thus, several aspects influence the usage of a product or service through the mediator satisfaction

Hypothesis3c: Satisfaction has a positive relationship with the depth of a relationship.

2.3

Antecedents of a relationship in an contractual B2B context

In order to determine which factors are relevant for a contractual B2B context and a company, this study will discuss these antecedents and their relevance. Palmatier et al. (2006) give an overview of antecedents, which are identified in the literature and which have an impact on the relationship mediators. They included only antecedents in their research if at least ten effects emerge to support its empirical analysis. This seems a solid principle to select antecedents and I will select these antecedents to include in the model. The antecedents of a relationship can be divided in three categories: customer-focused, seller-focused and dyadic antecedents. Customer-focused antecedents require the involvement of the customer and the extent in which these antecedents are relevant for the customer is customer dependent. This can be for instance increased convenience because customers are used to the company. This is in contrast with seller-focused antecedents, which come from the active involvement of the seller. One may think of the knowledge level of salespersons. At last, dyadic antecedents need active involvement from both parties and the value of these antecedents can be determined for both parties.

2.3.1 Customer-focused antecedents

The first customer-focused antecedent is relationship benefits. Relationship benefits refer to the perceived value of customers in a relationship when they receive benefits from an exchange partner, such as time saving, convenience, and companionship. Gwinner, Gremler and Bitner (1998) distinguish relationship benefits in three different benefits. Social benefits focus on the relationship itself rather than on the outcome (or result) of transactions. Secondly, special treatments benefits are relationship marketing programs such as loyalty cards. Their aim is to build switching cost in order to increase loyalty. The third benefit is confidence benefit which is merged with trust by Hennig-Thurau, Gwinner and Gremler (2000). Gwinner, Gremler and Bitner (1998) describe confidence benefits as “feelings of reduced anxiety, trust, and confidence in the provider”. This is indeed closely related with the definition of trust by Moorman, Zaltman and Desphandé (1992). This antecedent is relevant to al relational exchanges because benefits such as time saving and convenience arise when parties exchange information which enables them to act more effective and efficient. In order to keep the conceptual model clarified, relationship benefits are considered as one antecedent in this paper.

The relation between relational benefits and trust has not received attention. An explanation of the absence can be the fact that confidence benefits have much in common with trust. As a consequence, I expect a positive relation between relational benefits and trust.

The last effect is the effect of relational benefits on satisfaction, which is investigated by several authors with mixed results. Hennig-Thurau, Gwinner and Gremler (2002) found only a significant relation between confidence benefits and satisfaction, while Reynolds and Beatty (1999) found support for relation between social and functional benefits and satisfaction. When all components of relational benefits are aggregated I expect a positive influence of relational benefits on satisfaction.

Hypothesis4a: Relational benefits have a positive relationship with trust.

Hypothesis4b: Relational benefits have a positive relationship with satisfaction.

Dependence on seller is the second customer-focused antecedent. Dependence on seller reflects the customer’s evaluation of the value of seller-provided resources for which few alternatives are available (Hibbard, Kumar, and Stern 2001). In the literature, a distinction is made between total dependence and relative dependence. Total dependence is the dealer dependence plus supplier dependence while relative dependence is the dealer dependence minus supplier dependence. As total dependence increase, both parties have greater stakes in the relationship. Higher levels of total dependence will lead to an increased awareness of the buyer that the supplier wants to keep the relation intact. Dealers, who are relatively dependent on their supplier belief, that their relationship with the supplier is essential in order to achieve their goals. With high perceived relative dependence, buyers acknowledge the dependence imbalance and behave more passively because they believe negotiating or discussion is not effective. The company which is researched is the only specialised transport insurance company and therefore one can expect that few alternatives are available which in term increase dependence on seller. Because, dependence on seller is a customer-focused antecedent, relative dependence will be examined. In case of a contractual context, parties are dependent on the supplier during the contract time; dependence on seller is therefore regarded as a relevant antecedent for a contractual setting.

who found a very small effect of dependence on satisfaction. They stated that the variance explained in satisfaction by dependence is too low to be relevant. Therefore,

Hypothesis5a: Dependence on seller has a positive relationship with commitment.

Hypothesis5b: Dependence on seller has a positive relationship with trust.

Figure 1 Conceptual model

2.3.2 Seller-focused antecedents

Palmatier et al. (2006) define relationship investment as the time, effort, and resources that sellers invest in building stronger relationship. Such investments often generate expectations of

reciprocation that can help strengthen and maintain the relationship. This antecedent enables marketers to understand when investing in relationships is more effective or less effective. This is due to the fact that not all customers appreciate intensive relationship and wants only an exchange with a minimum of hassles (De Wulf, Odekerken-Schröder and Iacobucci 2001). In a contractual setting, marketers are able to invest in relations because the relationship is, due to the contract period, more long-term oriented. Ganesan (1994) examined the effect between relationship investment and the long-term orientation of a customer. He found that relationship investment has a positive effect on the perception of vendor’s credibility and benevolence. These variables have a positive influence on the retailer’s long-term orientation.

When considering the effect of relational investment on relational mediators, creating unrecoverable resources creates social bonds that encourage customers to stay in that relationship and sets expectations of reciprocation (Smith and Barclay 1997). With reciprocation in mind, one can expect a positive influence because a committed partner wants to work for its relation (Morgan and Hunt 1994). Thus, when relationship investment increases, feelings of reciprocation will increase which in turn will lead to commitment.

When focusing on trust, Ganesan (1994) found support for the relation between vendor’s investments and credibility and benevolence, which may be regarded as dimensions of trust. This can be explained by the fact that if the supplying party is willing to invest, this will eventually lead to a long-term orientation. A long-term orientation may be necessary to perform the perceived outcomes of trust defined by Anderson and Narus (1990). In conclusion, relationship investment is an indicator for a long-term relationship in which trust is an important mediator.

Palmatier et al. (2006) found a positive influence of relationship investment on satisfaction. According to them, managers should act proactively with respect to RM.

Finally, De Wulf, Odekerken-Schröder and Iacobucci (2001) show that relationship investment will lead to increased relationship quality, which can be regarded as an overall measure of relationship mediators. Derived from these findings, I propose the following:

Hypothesis6a: Relationship investment has a positive relationship with commitment.

Hypothesis6b: Relationship investment has a positive relationship with trust.

Hypothesis6c: Relationship investment has a positive relationship with satisfaction.

services is salesperson problem-solving assistance. The study of Crosby, Evans, and Cowles (1990) was based on data of an insurance company which is especially relevant for the researched company. Palmatier et al. (2006) found that seller expertise has the greatest positive influence on relational mediators. Lagace, Dahlstrom and Gassenheimer (1991) investigated the effect of seller expertise on satisfaction and trust. Both mediators are strongly influenced by seller expertise. This is in line with the finding of Doney and Cannon (1997) who found a positive relationship between seller expertise and trust. With respect to commitment, seller expertise can increase the value of a relationship which makes the relationship more important for the customer. As a result, customers will invest efforts to maintain the relationship. Formally stated,

Hypothesis7a: Seller expertise has a positive relationship with commitment.

Hypothesis7b: Seller expertise has a positive relationship with trust.

Hypothesis7c: Seller expertise has a positive relationship with satisfaction.

2.3.3 Dyadic antecedents

Communication is defined by Mohr, Fisher, and Nevin (1996) as the amount, frequency, and quality of information shared by exchange partners. Morgan and Hunt (1994) stated that communication builds stronger relationships in an exchange by helping resolve problems, align goals and uncover new value-creating opportunities. Communication is an essential part of a relationship according to Mohr and Nevin (1990) who stated that communication is the “glue that holds together a channel of distribution”. Therefore, this antecedent will be included in this paper.

The effect of communication on all studied relationship mediators is investigated. Anderson and Weitz (1992) found positive influences of communication on manufacturer’s and distributor’s commitment. Mohr, Fisher and Nevin (1996) found a positive influence of communication on commitment and satisfaction. In addition, Morgan and Hunt (1994) found a positive relation between communication and trust. These findings are in line with Palmatier et al. (2006) who found that communication is one of most important determinants of the relational mediators trust, satisfaction and commitment. Therefore, the following is proposed,

Hypothesis8a: Communication has a positive relationship with commitment.

Hypothesis8b: Communication has a positive relationship with trust.

Hypothesis8c: Communication has a positive relationship with satisfaction.

organizations.” Doney and Cannon (1997) assess similarity “as the buyer’s belief that the salesperson shares common interests and values with people in the buying firm.” Similarity can trigger the intentionality and/or prediction process because similarity can, in goal-interdependent situations, be a cue for expecting the other party to facilitate one’s goals (Doney and Cannon 1997). Therefore, similarity is expected as an important antecedent because in a contractual setting, the payment of a service and the execution are mostly not simultaneously. As a result, buyers assess similarity as important because buyers who perceive salespeople as similar to themselves could expect such salespeople to hold common beliefs about what behaviour, goals, and policies are appropriate. Doney and Cannon (1997) examined how similarity influences the trust in a firms salesperson. They found a positive relation between these variables. Morgan and Hunt (1994) defined similarity as the shared value between exchange partners. There is a positive relation between shared values and commitment or trust. Crosby, Evans and Cowles (1990) examined the effect of similarity on relationship quality. They divided relationship quality in satisfaction and trust and found a positive influence of similarity on relationship quality. As mentioned before, satisfaction can be defined as an affective state resulting form the appraisal of all aspects of the relationship. The relationship aspect similarity seems to activate a positive affection to the working relationship between firms. As a result,

Hypothesis9a: Similarity has a positive relationship with commitment.

Hypothesis9b: Similarity has a positive relationship with trust.

Hypothesis9c: Similarity has a positive relationship with satisfaction.

Interaction frequency is the number of interactions within a unit of time. Doney and Cannon (1997) divided interaction in social and business interaction. The effect of interaction frequency is the same as those of relationship duration. Since, interaction frequency is closely related to relationship duration it will be included as well. Crosby, Evans and Cowles (1990), stated that interaction frequency is the frequency with which the salesperson communicates (face-to-face or indirectly) with the customer either for personal or business reasons.

affiliation are created. With respect to satisfaction, salespersons better assess customer needs through increased contact frequency. They gain more ‘customer knowledge’ and as a result are able to tailor products or services to those needs. This may influence the primary determinant of satisfaction: perceived quality or performance. Therefore I expect,

Hypothesis10a: Interaction frequency has a positive relationship with commitment.

Hypothesis10b: Interaction frequencyhas a positive relationship with trust.

Hypothesis10c: Interaction frequency has a positive relationship with satisfaction.

The last antecedent is a negative one, namely conflict. Conflict contents the overall disagreement between exchange partners. Kumar, Scheer and Steenkamp (1995) argued that conflict is usually viewed as composed as several stages. Two stages are frequently studied with respect to marketing channels conflicts. Affective conflict is defined as hostility, frustration, and anger toward a channel partner. These feelings of hostility can erupt into manifest conflict: behavioural responses such as open expression of disagreements or overt attempts to prevent the other from achieving its goals (Brown, Lusch and Smith 1991). Conflicts can arise in every relationship and are therefore relevant for a contractual B2B context.

Palmatier et al. (2006) found strong negative influences of conflict on relationship commitment, trust and satisfaction. They found that the negative influence of conflict is stronger than the positive influence of other relationship antecedents. Anderson and Weitz (1992) investigate the relationship between commitment and conflict and the negative influence is supported. Based on these findings,

Hypothesis11a: Conflicthas a negative relationship with commitment.

Hypothesis11b: Conflicthas a negative relationship with trust.

3 Method

To collect the data, I send questionnaires to customers with ten or less vehicles. There were 420 online questionnaires and 500 questionnaires sent by mail. This results in 256 responses from which 15 are partially completed. This means a response rate of 27.8%. The collecting method by mail is most effective with a response rate of 32.6% in contrast to a response rate of 22.9% of the online questionnaires. The questionnaire has been tested by several employees on spelling and intelligibility of the questions.

Most measures of the constructs I am examining were available in the literature (appendix 1). However, some measures are adapted to suit the specific needs of this research. The appendix contains all item sorted by their construct. In order to measure antecedents, I assessed all questions on a seven-point scale (1 = “Strongly disagree” and 7 + “Strongly agree”). When the questionnaire was tested with a five-point scale, the extreme points of a scale were almost not chosen. Therefore I increased the scale to a seven-point to have more variance.

3.1 Measures

Antecedents and Mediators

The questions which were asked in the questionnaire are shown in appendix 1 and will therefore not be discussed. Most questions are based on existing literature. However, some of these questions are adapted to have a better fit with the researched company or the study. The remaining questions are based on the output of expert interview within the researched company. Only dependence on seller will be discussed. Dependence on seller can be regarded as a multidimensional construct, because the dimensions are somewhat independent, a measure of internal consistency computed across these dimension would be inappropriate (Malhotra 2004). Dependence on seller can come from a lack of other potential suppliers or because of high switching costs. As a result, dependence on seller has been measured by a formative scale while other constructs are reflective scales. Respondents are asked to answer the following questions: “There are other insurers who can offer me the same product than X” and “It requires much time and/or money to switch to another insurer”.

Behavioural outcomes

The length of the relationship was measured by three items which reflect intended loyalty. Behavioural data can not be used because no inactive customers received a questionnaire.

The depth of the relationship is more difficult to measure because it hard to assess the exact size of a customer. In the database is not always information about the number of vehicles customers possess. To overcome this problem, I asked the customers the following question: “Which percentage of your insurable risks is insured at X?”

3.2 Reliability

In this study, a two-step approach was chosen to test the adequacy of the model. At first, the construct reliability was assessed using Cronbach’s Alpha and secondly, a factor analysis was conducted for all constructs.

The values of the Cronbach’s Alpha are shown in table 2. As mentioned before, dependence on seller is a formative scale. This may explain the low value of the Cronbach’s Alpha on this antecedent. The remaining constructs have all values above 0.6 which indicates satisfactory internal consistency reliability.

The outcome of the factor analysis was less satisfactory. A principal axis factoring with Varimax rotation was conducted. The outcomes are shown in the appendix 2. Several items have significant loadings in different factors and the constructs are not separated by factors. Therefore, I examine the data closer in order to find respondents who have a low standard deviation (< 0.5) across their answers. However, all respondents have higher standard deviations and therefore I decided to remove no respondents from further analysis. In order to come with a satisfactory factor outcome, the items which have scores on several factors were removed from further analysis. This was done by a stepwise process in which the worst item was removed at a time. Eventually, this resulted in a factor outcome which has all constructs in different factors (table 1).

Table 1 – Factor loadings antecedents C O N FLIC T SP EC IA L TR EA TM EN T B EN EF IT S P H O N E C A LL FR EQUEN C Y VIS IT FR EQUEN C Y C O M M UN I-C A TIO N SIMILAR IT Y SE LL ER EX P ER TI SE R ELAT IO N SH IP INVES TM EN T Conflict1 ,574 Conflict2 -,745 Conflict3 -,751 Spec_Ben1 ,762 Spec_Ben2 ,841 Int_Freq3 ,720 Int_Freq4 ,739 Int_Freq1 ,580 Soc_Ben2 ,571 Commu3 ,697 Simi1 ,771 Simi2 ,596 Sel_Exp2 -,450 Sel_Exp3 ,499 Rel_Inv3 ,496

With respects to the mediators, the problem was more complicated. Initially, when conducting a principal axis factor analysis with Varimax rotation, two factors were extracted from the data based on a criteria of an eigenvalue larger than one. The outcomes are shown in appendix 3. One item of trust and one item of commitment have significant loadings on both factors. Therefore, the item of trust was eliminated from the analysis. This elimination has an effect on the commitment item which is significant now on one factor. However, all satisfaction items are significant in both factors. Commitment and trust can be identified from the factor analysis while satisfaction is significant in both factors. Therefore, two options are left. First, a composite measure of all three mediators can be extracted. This will result in an overall measure of relationship quality. Secondly, I will use satisfaction/commitment, satisfaction/trust in the remainder of the study. However, because these mediators are hard to interpret, relationship quality will be used in the remainder of the study.

don’t do for most customers” and “I get faster service that most customers.” Contact frequency by phone call is the third antecedent. It is the measure of the number of phone calls, made by the supplier, to the customer and if the employees of X take time to learn the needs of the customer by phone. In contrast to interaction frequency which takes contacts (visits or phone calls) more general, interaction is now divided in two antecedents. Therefore, the fourth antecedent focuses on the number of visits from sales persons. The fifth antecedent is a mixture of communication and social benefits. It is formed from the statements “I know who I should call at X, when I need help” and “X is always reachable”. Thus, this antecedent focuses on the reachability of the supplier. Customer should know who to call and when to call. Similarity is the sixth antecedents and remains unchanged. The seventh antecedent is the seller expertise of the supplier and comes from two statements: “The employees of X know their products well” and “The employees of X know the risks in their particular market better than other insurers”. As a conclusion, this antecedent will focus on the expertise level of X. Finally, relationship investment is measured from the statement “X makes efforts for regular customers”, which is exactly what the term relationship investment stands for.

4 Results

The correlations are shown in table 2, above the diagonal while the covariances are shown below the diagonal. The Cronbach’s Alpha is shown on the diagonal of the table. Special treatment benefits has the lowest mean (3.58) followed by phone call frequency (3.74). The standard deviation is lowest for seller expertise (1.00) and communication (1.06), and therefore there is much agreement about these antecedents by the respondents. These antecedents have also a high mean so these antecedents are rated high among most customers.

The relationship between relationship quality and the antecedents is significant for all antecedents except visit frequency. The relations will be further tested by a multivariate regression analysis.

4.1 Antecedents

In order to test for a relation between the antecedents and mediators, a regression analysis was conducted. However, before discussing the outcomes of the model, the model is tested for its assumption in order to ensure that the results are valid and reliable. Thereafter, the outcomes of the model are discussed.

Table 2 – Correlation/Covariances Matrix

Construct Mean S.D CON SB PCF VF COMM SIM SE RI RQ LOYAL Conflict 5.29 1.33 .813 -0,03 -0,02 -0,02 -0,11 -0,04 -0,19 -0,11 -0,62** -0,02 Special treatment benefits 3.58 1.32 -0,02 .800 0,08 -0,06 0,04 0,02 0,01 0,01 0,26** -0,02 Phone call frequency 3.74 1.49

Model assumptions

In order to get reliable results from the analysis, the assumptions of the model are tested. At first de multicollinearity of the model was tested by evaluating the VIF values. There were no high VIF values in the regression analysis (table 5). Therefore, it can be concluded that there is no multicollinearity in the model. I analysed the scatter plot of the regression analysis’ residuals to detect heteroskedasticity. However, I could find no patterns for heteroskedasticity. Therefore, I assume that no heteroskedasticity exists in the model (appendix 6). Correlation between the residuals and the independent variables were computed in order to test for the assumption of endogeneity. All correlations have a p-value of 1.000 and are therefore the model meet the assumption of no endogeneity (table 3).

Table 3 – Endogeneity antecedents C O N FLIC T SP EC IA L TR EA TM EN T B EN EF IT S P H O N E C A LL FR EQUEN C Y VIS IT FR EQUEN C Y C O M M UN IC A TIO N SIMILAR IT Y SE LL ER EX P ER TI SE R ELAT IO N SH IP INVES TM EN T Relationship Quality Correlation 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 Sig. (2-tailed) 1 1 1 1 1 1 1 1 N 222 222 222 222 222 222 222 222

Finally, normality was tested. The results of the normality tests are shown in table 4, while the normality plot is shown in appendix 7. The Shapiro-Wilk test shows that the residuals of relationship quality are normally distributed. Because all assumptions are met, I will continue with this model.

Table 4 - Tests of normality antecedents

Kolmogorov-Smirnov(a) Shapiro-Wilk

Statistic df Sig. Statistic df Sig. Residuals 0,07 222,00 0,02 0,99 222,00 0,09

Results

The first antecedent of the regression analysis, conflict, has not changed. In the regression function, this antecedent has a significant strong negative relation with the mediators. Special treatment benefits, which is a subcategory of relationship benefits, has significant positive relations with all mediators. Despite of the finding of this relation, I will not make conclusions about the support of the hypotheses due to the fact that special benefit treatment is only a subcategory of relationship benefits. The following two antecedents are initially merged in two antecedents: phone call

frequency and visit frequency. In contrast to the expectations, visit frequency has no (strong) relation with the mediators. One possible explanation may be the fact that in some cases, salesperson visit customers because of complains or conflicts while on the other hand, customer react positively on sales visits when there are no problems. These effects may offset each other which results in non-significant relations. Phone call frequency, on the other hand, shows positive non-significant effects. Thus, in general, one can conclude that contact frequency may have a positive effect on the mediators.The content of the antecedent communication has altered, so I will not connect the findings with the hypotheses regarding communication. However, the results are significant and therefore worthwhile to discuss. This antecedent has much to do with the ease in which customers may communicate with the supplier. Customers may want to check their insurance portfolio or have some adjustments. It is important for them to know which employee they should call and to have always the possibility to call to the insurer. The antecedent similarity has not changed and focuses on the interpersonal relation between the salesperson and the customer. A good relationship with customers has a positive relation with all mediators. Because similarity has a positive relation with relationship quality, it can be concluded that, in general similarity has a positive relation on customer relationships. However, there is a significant relationship with relationship quality. Therefore, despite a strong assumption of the relation between similarity and the mediators, I can not support the hypothesis because of a lack of evidence. The same can be applied for seller expertise and relationship investment. Both show a positive relation with relationship quality but there relation with satisfaction, trust and commitment in particular is not tested and I can not make a statement of the hypothesis.

Table 5 – Regression of Antecedents

Relationship Quality Adj. R2 0.849

Std. Beta’s Sig VIF

Constant .007 ,787

4.2

Behavioural outcomes

With respect to the behavioural outcomes of relationship quality, the effect of al outcomes will be discussed separately. The factor scores of relationship quality was used as explanatory variables for the regression analysis. Due to missing values, 15 respondents were eliminated from further analyses.

4.2.1 Length of Relationship

When the relation between mediators and the length of the relationship are plotted, a C-shaped curve will appear (appendix 8). When a regression analysis is estimated this resulted in non-significant effects (table 6) as you might expect from appendix 8. However, because of the special shape of the curve, I

did a further investigation of this relationship. The special shape distribution of the data may indicate latent classes.

Model assumptions

To test for latent classes, I used LatentGold 4.5 in order to do a latent class regression. The dependent variable is loyalty while the factor scores of relationship quality are the independent variable. In order to select the most appropriate number of classes, the models are judge on BIC, AWE, AIC, R2 and classification error. In addition, the model was also tested with class independent intercepts. The outcomes of these evaluation measures are shown in table 7.

The AWE, which takes classification error in account, indicates that a two class model is the most appropriate. The R2 shows a large improvement from one to two classes. The BIC, AIC and AIC3 suggest respectively three, four and three classes. However, there are small differences between the classes while the classification error increases. Therefore, when all measures are taken in account, the two class model with dependent intercepts is chosen.

Table 6 – Regression length of relationship Loyalty

Adj. R2 _0,01

Std. Beta’s Sig VIF (Constant) -,029 ,642

Relationship

Quality ,075 ,254

Table 7 – Evaluation latent classes Number of

Classes

Intercept BIC AIC AIC3 AWE Class.Err. R²

1 Variable 658,4677 648,0762 651,0762 683,86 0 0,0056 2 Variable 488,7648 464,5179 471,5179 611,58 0,0681 0,8791 3 Variable 486,499 448,3969 459,3969 667,26 0,0731 0,9271 4 Variable 500,165 448,2075 463,2075 864,81 0,2489 0,9487 1 Fixed 658,4677 648,0762 651,0762 683,86 0 0,0056 2 Fixed 662,2481 641,4651 647,4651 917,25 0,2175 0,2467 3 Fixed 675,6729 644,4984 653,4984 1077,29 0,3164 0,3973 4 Fixed 696,525 654,959 666,959 1328,23 0,5621 0,3478

Secondly, the residuals are checked for heteroskedasticity. The scatter plot of the residuals of the model is shown in the appendix 9 and 10. However, there is no indication for heteroskedasticity.

Thirdly, the existence of endogeneity in the model is tested by correlating the independent variables with the residuals (table 8). However, there exist no significant correlations between the independent variable and the residuals, as a results it can be

concluded that there is no endogeneity in de model. Finally, I checked whether the residuals are normally distributed. The normally tests are shown in table 9, and are non-significant which indicates a normal distribution.

The estimated parameters are shown in table 10. The class which reacted positively on the mediators is about two-third (62.2%) of the respondents, which left one-third of the respondents who have a negative relation between the mediators and loyalty. Which means that relationship quality has a positive relation with the length of the relationship for two-third of the customers while one-third of the customers react negatively on the relationship quality.

4.2.2 Breadth of the relationship

The effect of relationship quality on the breadth of the relationship is examined with a multiple regression analysis. After examining the data, one control variable is put into the regression analysis.

Table 8 – Endogeneity length of relationship

Relationship Quality Residuals Pearson Correlation -,023 Sig. (2-tailed) ,733

N 221

Table 9 – Normality test for length of relationship

Kolmogorov-Smirnov(a) Shapiro-Wilk

Relationship duration is included because the supplying company gains more knowledge of a customer and therefore offers additional relevant products.

Model assumptions

The model is tested for multicollinearity, heteroskedasticity, endogeneity and normal distributed residuals. According to the VIF values, which are very low, there is no multicollinearity in the model.

The residuals are plotted against the predicted value and the independent variables and there were no indications

for heteroskedasticity (appendix 11 and 12). According to the correlation (table 11) there is also no endogeneity in the models.

Table 11 – Endogeneity breadth of relationship

Relationship

Duration

Relationship Quality Unstandardized Residual Pearson Correlation ,000 ,006 Sig. (2-tailed) 1,000 ,931

N 218 218

Finally the residuals are tested for there normal distribution. The normality test indicates that the residuals do not follow a normal distribution (table 12). However, because there are many respondents with one or two products (so the dependent variable is not normally distributed) the residuals show the same pattern. That may explain the disturbing on the lower and of the normality plot (appendix 13). Because the other assumptions are met, I will continue with this model. The results of the model are shown in table 13.

Table 12 – Normality tests breadth of relationship

Kolmogorov-Smirnov(a) Shapiro-Wilk

Statistic df Sig. Statistic df Sig. Unstandardized Residual ,115 221 ,000 ,951 221 ,000

The outcomes show that there is no relation between relationship quality and the breadth of the relationship, this was also reflected in the fact that the regression as a whole was not significant. The F-value indicated this by showing a non-significant value The first conclusion may be that RM will not affect the breadth of a customer relationship. However, the control variable relationship duration has a positive relation with the breadth of the relationship. Eventually, the breadth of the

Table 10 - Latent class regression length of relationship R2 0,879

Class 1 Class 2 Class Size 0,622 0,378

relationship will be affected by RM because an increased length of the relationship, which is the primary outcome of RM, has a significant relation with the breadth of the relationship.

4.2.3 Depth of relationship

The scale of depth of the relationship is much skewed as is shown in figure 2. After excluding respondents with missing values from the analysis, 220 respondents remain.

Figure 2 - Distribution of depth

0 20 40 60 80 100 120 0 5 _{10 25 35 50 60 75 80 90 99} Percentage insured at X N u m b er o f res p o n d en ts

Model assumptions (linear regression)

Because there is only one independent variable, there is no multicollinearity in the model (table 15). The scatter plots of residuals are shown in appendix 15, 16 and 17. Due to the skewed distribution of the dependent variable, the model performs better when the dependent variable has high values. Therefore, the assumption of heteroskedasticity is not met. With respect to the endogeneity of the model (table 14), all correlations between the independent variables and the residuals are highly insignificant (p-value = 1,000). However, because the data is much skewed, the residuals of the model are not normally distributed (appendix 18 and 19). Because two model assumptions can not be met, the model is not accepted.

Table 13 – Regression of Breadth of Relationship Adj. R2 _,163

Table 14 – Endogeneity depth of relationship

RELQUAL Residuals Pearson Correlation ,000

Sig. (2-tailed) 1,000

N 221

Because a linear regression model is not able to meet the assumptions, a logistic regression model is performed. All respondents with less than 80 percent insured at X, have a value of 0 while the respondents with more than 80 percent insured get a value of 1. This variable was set as the dependent variable while relationship quality was the independent variable. This resulted in the outcomes shown in table 16. Relationship quality shows no significant relation with the depth of the relationship.

Table 15 – Regression of Breadth of Relationship Adj. R2 _,020

Variable Std. Beta’s Sig VIF Constant 83,626 0,000

Relationship Quality

3,865 0,021 1,000

Table 16 – Logistic regression of Depth of Relationship -2 Log likelihood 282,106

Cox & Snell R Square 0,005

Nagelkerke R Square 0,007

Variable Beta’s Sig Constant 0,670 0,000 Relationship

Quality

5 Discussion

5.1

Conclusions

The goal of this study was to examine the effectiveness of RM with respect to the researched company, which is in a B2B contractual context. The study was divided in three aspects. The first aspect was to identify relevant antecedents of relationships in the B2B contractual context. Secondly, the effect of RM on the outcomes of RM was researched and finally the effect of mediators on the outcomes was investigated.

One of the main findings of this study, which altered the structure of the study, was the fact that the mediators cannot be divided into commitment, trust and satisfaction. The factor analysis show separate factors for commitment and trust. However, satisfaction is not derived from the factor analysis and is divided between the factors of commitment and trust. However, all mediators can be composed in one factor which resulted in the composed mediator relationship quality. This mediator is common in existing marketing research (for instance Palmatier et al. 2006; De Wulf, Odekerken-Schöder and Iacobucci 2001; Hennig-Thurau, Gwinner and Gremler 2008). A possible explanation for this finding may be the fact that satisfaction is dependent on both the level of trust and commitment and vice versa. When customers are not satisfied when they do not trust the firm and reversely, unsatisfied customers do not trust the firm. The same may applied for commitment. As a result, these construct do depend on each other and customers do not have distinct feelings about the supplier, with respect to the mediators.

A second conclusion from the factor analysis, were the different antecedents than proposed. Dependence on seller was not extracted from the factor analysis and was excluded from the remainder of the study. Relationship benefits, which is composed of three dimensions, was not extracted from the analysis and only the special treatment benefits construct could be extracted. Interaction frequency was divided in phone call frequency and visit frequency. Initially, it was proposed that these components were composed in one construct. The core of the other constructs remains mainly unchanged, despite the fact that some items were excluded from the analysis. These changes lead to another structure of the study than initially proposed. As a consequence, the proposed hypotheses could not be tested. However, the study revealed insights in the structure of RM which will be discussed hereafter.

frequency shows no significant relation. In contrast, Palmatier et al. (2006) found a small significant effect of interaction frequency on relationship quality. The other antecedents have significant effects in both studies. Conflict has the strongest influence on relationship quality. In fact, this is the only antecedent which is negative. Despite the fact that Palmatier et al. (2006) did not investigate the relation between conflict and trust, they found that conflict has the strongest effect on trust of all antecedents. In addition, conflict has a strong influence on commitment as well. It can be concluded that this result is in line with Palmatier et al. (2006). The magnitude of the relations between other antecedents and relationship quality differ with the outcomes of the meta-analysis of Palmatier et al. (2006), which may imply that the effects and structure of RM are context dependent. The remaining antecedents will be discussed in order of importance. Conflict is followed by communication and similarity with respect to the magnitude of the effects. Customers in the insurance industry want to be able to communicate to the supplier at any time and want to know which employee they should call. Besides, a good relationship with the salesperson is perceived important by the respondents. It can be concluded that the most important antecedents, conflict, communication and similarity are all a strong related to build interpersonal relationships. The following antecedents, phone call frequency and seller expertise, are more functional aspects of the relationship and focus on an objective evaluation of service performance. Finally, the last two antecedents, relationship investment and special treatment benefits, focus on a good relationship between the supplying company and the customer. Customers react positively on perceived relationship investment of the supplier and a special treatment of the firm towards them. These findings are in line with Palmatier et al. (2007) who identified different relational levels. They found that interpersonal relationships have the greatest impact on financial outcomes. They stated that “in situations in which the salesperson plays a central role, relationship quality with the salesperson has a significant, direct impact on financial outcomes and operates independently of relationship quality with the selling firm”.

which can be defined as an out-of-the-ordinary event during an interaction that customer perceive or recall as unusually negative (Roos 2002). However, because the current study makes no distinction between customers who face a critical incident or not this may be a possible explanation of the surprising existence of customer segments. Another argument for the mixed effect of relationship quality on the length of the relationship may be context dependent. Rauyruen and Miller (2007) focus on the B2B context and found that commitment and trust cannot predict loyalty. Therefore, it can be concluded that customers in general do not become more loyal when they have a high evaluation of relationship quality. However, the company is market leader and is the only specialized insurance company in their industry, the lack of competition may force the customers to remain loyal because switching to another company may be perceived as a cost due to an expected lower service quality.

The explanation of the negative relation of relationship quality on the length of the relationship is not obvious. The covariates (antecedents of the relationship) are not significant so there are no significant differences between the classes. A possible explanation may be the price consciousness of the customers. The researched company sells premium products with an above average price. When customers are not fully aware of the benefits of the insurances, which justify the higher price, they will not become loyal. However, this does not explain why there is a negative relationship between relationship quality and loyalty. Wong et al. (2006) find a possible explanation for the negative relationship. According to them, favour exchange is an important antecedent to build longer-term relationship. Favour exchange is defined (Yau et al. 2000) as situations in which one party provides or reciprocates favours or allowances to another and expects that similar favours or allowances will be returned by that client at a later date. Therefore as Yau et al. (2000) stated, the way to maintain longer-term relationships is to store favours and keep the other indebted. However, when customers become extremely satisfied, committed or trusted in the firm the salespersons run out of favours and the customer does not longer feel indebted. As a consequence, the relationship is not longer perceived as an obligation and the customer may feel free to leave the firm.

of the relationship. In addition, the length of the relationship increases when relationship quality increases. As a result, it can be concluded that, eventually, RM has a positive effect on the breadth of the relationship.

The depth of the relationship has no significant relation with relationship quality. Customer tends to insure almost all their insurances at one insurer instead of gradually increasing their portfolio of insurances. The conclusion may be that, at least for the insurance industry, RM has no relation with the depth of the relationship. Furthermore, Mägi (2003) investigated the effect of satisfaction on the share-of-purchase, which is related to the depth of the relationship. However, this effect was very small (the unique explained variance is 0.02). Therefore is can be concluded that mediators do not have a (strong) effect on the depth of a relationship.

As mentioned before, the structure of the tested model has changed. For clarity reasons the model is presented in figure 3.

Figure 3 - Final model

Non-significant relation Positive relation

Negative relation

Legend

Relationship Quality Length of relationship Conflict

Special treatment benefits

Phone call frequency