Is it true loyalty or just random purchases?

Is it true loyalty or just random purchases?

HOW DO BEHAVIORAL AND ATTITUDINAL LOYALTY INFLUENCE PROFITABILITY AND HOW DOES THIS DIFFER ACROSS INDUSTRIES IN THE

NETHERLANDS?

by INGE MEIJER

2

Is it true loyalty or just random purchases?

HOW DO BEHAVIORAL AND ATTITUDINAL LOYALTY INFLUENCE PROFITABILITY AND HOW DOES THIS DIFFER ACROSS INDUSTRIES IN THE

NETHERLANDS?

by INGE MEIJER

University of Groningen Faculty of Economics and Business

MSc. Business Administration – Marketing Research and Management Master Thesis April 2013 Commelinstraat 36 1093 TV Amsterdam 06 41 76 56 48 I.B.A.Meijer@student.rug.nl Student number 2009781

Internal supervisor: Erjen van Nierop - University of Groningen Second supervisor: Yi-Chun Ou – University of Groningen

3

Management summary

The main issue of this thesis is how behavioral and attitudinal loyalty relate to profit and how this differs across industries. Five industries have been investigated in order to find that that there indeed are differences and that it is not allowed to generalize the outcomes of the research over these industries. Seven different characteristics have been investigated in order to find out which factors influence the differences that occur per industry.

For the Fast Moving Consumer Goods industry (pasta and soft drinks) people are behavioral loyal because they are not willing to look for alternatives and because they are afraid the quality of another brand is worse. Attitudinal loyalty is not significantly relating to profit.

In contractual settings (telecom and health insurances), neither behavioral loyalty or attitudinal loyalty is significantly related to profit. Involvement is an important influencer for behavioral loyalty and profit separately, alike risk avoidance. Furthermore, inertia leads to behavioral loyalty. Satisfaction plays a role in influencing loyalty in telecom but is not significantly contributing to loyalty in the health insurance industry.

In the tourism industry (bungalow parks), behavioral loyalty is to a higher extent related to profit than is attitudinal loyalty. Inertia is an important influencer of behavioral loyalty. Attitudinal loyalty can be enhanced by risk avoidance and ingoing word of mouth.

Some outcomes are allowed to be generalized over industries.

The more involved people are with a product the more attitudinal loyal they become, although this does not automatically lead to higher profit. This is the case in all industries. Furthermore, when people experience a lot of time, effort and money to switch to another brand they are less likely to evaluate alternatives. Satisfaction has a positive influence on both behavioral and attitudinal loyalty. Although all these factors lead to loyalty, they do not automatically lead to profit.

A segmentation shows clear and useful segments in terms of behavioral and attitudinal loyalty. The research reveals that the most loyal people are soft drink buyers and the least loyal people are in the bungalow park industry.

Key words: Brand loyalty, behavioral loyalty, attitudinal loyalty, profit, switching costs,

4

Preface

This paper is my master thesis for the master Business Administration, specialization Marketing Research and Management at the University of Groningen. I conducted a study about the differences in behavioral and attitudinal loyalty in several industries. I combined writing my master thesis with an internship at DirectResearch B.V., expert in online market research.

DirectResearch was founded eight years ago by Drs. Okkie Boot. The organization has developed into a mature research partner in both profit and non-profit industries. DirectResearch allowed me to make use of the their consumer panel; EUpanel. This panel exists of 27.000 people representative for the Dutch population. This way I was able to reach enough respondents for a reliable research.

5

Table of content

2.1 Behavioral and attitudinal loyalty... 10

2.2 Loyalty in relation to profit ... 11

2.3 Loyalty across industries ... 13

2.4 Fast Moving Consumer Goods industry ... 14

2.4.1 Involvement ... 14

2.4.2 Inertia ... 14

2.5 Contractual settings: telecom and health insurances ... 16

2.5.1 Switching costs ... 16 2.5.2 Inertia ... 17 2.5.3 Competition ... 17 2.5.4 Satisfaction ... 18 2.6 Tourism ... 21 2.6.1 Risk aversion ... 21 2.6.2 Word of Mouth ... 22 3 Research design ... 26 3.1 Research methods ... 26 3.2 Data collection ... 29 3.3 Plan of analysis ... 29

3.3.1 Designing the database ... 29

3.3.2 Econometric model ... 31

6 3.3.4 Segmentation ... 36 4 Results ... 38 4.1 Demographic criteria ... 38 4.2 Transforming variables ... 39 4.3 The model ... 42 4.3.1 Functional form ... 42 4.3.2 Pooling ... 44 4.3.3 Estimation ... 45 4.4 Hypothesis testing... 46 4.4.1 Direct effects ... 46 4.4.2 Mediating effects ... 48

4.4.3 Summarizing the results of hypothesis testing ... 53

4.5 Segmentation ... 54

4.5.1 Stage 1: objective of clustering ... 54

4.5.2 Stage 2: Designing the cluster analysis ... 55

4.5.3 Stage 3: Assumptions ... 55

4.5.4 Stage 4: Selecting a Clustering Algorithm ... 55

5 Conclusions and recommendations ... 59

5.1 Main research; testing hypotheses ... 59

5.2 Additional research; segmentation ... 62

5.3 Implications ... 63

5.3.1 Theoretical implications ... 63

5.3.2 Practical/managerial implications ... 63

5.4 Limitations and future research ... 64

6 References ... 66

7

1 Introduction

It is generally accepted that firms benefit more from long-term customer relationships than short-term customer relationships. Morgan and Hunt (1994) have given conceptual evidence for this argument. Reichheld and Sasser (1990) have demonstrated that it is far less expensive to retain a customer than to acquire a new one. Moreover, the longer the customer stays in the relationship, the more profitable this becomes for firms. Evidence for this statement is found in several industries. Reichheld and Sasser (1990) state that service organizations focus on so-called “zero defections” which means keeping every customer the company can profitably serve. This has a major effect on the bottom-line. By retaining 5% more of their customers, companies can boost profits by 25% - 100% due to economies of scale, larger market share, smaller unit costs and many other factors usually associated with competitive advantage (Reichheld and Sasser, 1990). In the banking and insurance industry it is found that a customer who has been with a bank for five years is far more profitable than a customer who has been with a bank for one year (Reichheld, 1996). This is due to lower operational costs and efficiency advantages. More specifically, a credit-card issuer found that a 5% increase in retention increased per-customer profit by 125% (Reichheld, 1996). Likewise, it has been estimated that a company can double its size in 14 years with a retention rate 5% higher than a comparable company, and even in seven years when the retention rate is increased to 10% (Sheth and Sisodia, 1995). Therefore, when money is spent on retaining customers under the relationship marketing strategy, this is likely to make marketing more efficient (Sheth and Parvatiyar, 1995). In line with the theory of Sheth and Parvatiyar, Jacoby and Chestnut (1978) suggest that brand-loyal consumers may be willing to pay more for a brand because they perceive some unique value in the brand that no alternative can provide. These studies imply that loyalty is essential in order to be profitable.

8 profitability is much weaker and subtler than the proponents of loyalty programs claim. Little or no evidence is found to suggest that customers who purchase steadily from a company over time are necessarily cheaper to serve, less price sensitive, or particularly effective at bringing in new business.

Inconsistencies in results that have been found previously (for example Reicheld and Sasser, 1990; Reinartz and Kumar, 2002) might be due to inconsistent measures for loyalty. The term ‘loyalty’ can be defined and measured in many different ways. Therefore, in this research a distinction is made between behavioral and attitudinal loyalty; concepts which are explained in more detail in chapter 2.1.

Another reason why the phenomenon of brand loyalty is interesting is because the majority of journals only investigates one specific industry. The results may be specific to that particular industry but in many cases this is not tested (Lee, Lee and Feick, 2001;Knox and Walker, 2003; Khatibi, Hishamuddin and Thyagarajan, 2002; Kandampully and Suhartanto, 2003). However some scientists have started researching the topic of loyalty across several industries (Reinartz and Kumar, 2002; Martensen, Grønholdt and Kristensen, 2000), there is still a gap in literature. This research tries to cover this gap by investigating whether there are significant differences across industries in the way loyalty contributes to profit. This leads to the following research question:

How do behavioral loyalty and attitudinal loyalty influence profitability and how does this differ across industries?

1.1 Conceptual model

9 Figure 1. Conceptual model

In this model, the dependent variable is profitability which is influenced by the two independent variables attitudinal loyalty and behavioral loyalty. The model is moderated by industry because differences are sought per industry in the extent to which behavioral and attitudinal loyalty create true loyalty. In the theoretical framework each industry is described according to specific characteristics that characterize this industry. However, many characteristics describe a certain industry, in this thesis only the most important characteristics are described and used as a predictor for loyalty in order to identify the most important relations for a specific industry.

This thesis is relevant for both theoretical and practical purposes. In literature, much is written about loyalty and its influence on performance or profitability. However, many of these researches focus on only one specific product or industry which makes it not possible to say it is generalizable over industries or not. This research distinguishes from them and adds value because differences are investigated per industry. The finding of the research is practically relevant as well, since organizations can benefit from the new insights in loyalty specific per industry. These results are presented to DirectResearch B.V. who can use them as a service to their customers, who are working in different industries.

The structure of the thesis is as follows. After this introduction chapter, the theoretical framework follows, where an extensive overview is given concerning loyalty issues. After the general part about both attitudinal and behavioral loyalty, several industries are discussed according to literature. Here it is explained which factors characterize each specific industry and which effect is expected to have on loyalty. Based on these expectations, hypotheses are

10 developed. In chapter 3, the research design can be found. In this chapter, the research project is explained in terms of research methods, data collection methods and a plan of analysis. This chapter exists of two parts, namely the research design about the hypothesis testing and the segmentation. Next, the results of the research are presented in the fourth chapter, also first the hypothesis testing followed by the segmentation. The final chapter contains conclusions and recommendations for the scientific world and the managerial implications.

2 Theoretical framework

Loyalty is defined by Schiffman, Bednall, Cowley, Watson and Kanuk (2001) as ‘the commitment of a consumer to a product or service, measured by repeat purchase or attitudinal commitment’. A more detailed definition is given by Oliver (1999) who defines loyalty as ‘a deeply held commitment to rebuy or repatronize a preferred product/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’. Loyalty has been linked to profit in different ways. Chettayar (2002) put the additional costs of acquiring customers over retaining them at ten-fold. Additionally, brand loyalty leads to a greater market share (Assael, 1998). Ehrenberg, Goodhardt and Barwise (1990) showed that brands with a higher level of purchasing loyalty exhibit greater market shares for frequently purchased products, implying gaining more profit.

2.1 Behavioral and attitudinal loyalty

11 behavioral loyalty is the only way of measuring loyalty and suggest that attitudes are not relevant to determining brand loyalty (Sharp, Sharp and Wright, 2002).

The concept of criticizing the effect of loyalty in general is not new. It was Newman (1966) who first challenged the approach of equating behavior patterns with preferences to infer loyalty. Other researchers (e.g., Day, 1969; Coulson, 1966) have highlighted the distinction between ‘‘spurious loyalty’’ as captured by the behavioral patterns, and ‘‘true/intentional loyalty’’ that extends beyond the regular purchasing of a brand (Bandyopadhyay and Martell, 2007). It can be concluded that there are many visions on the subject of loyalty. The fact that some scientists do not even distinguish between behavioral en attitudinal loyalty infers the topic needs some more research. Besides that, the positive relation between loyalty and profit is not substantiated by all researchers. This is discussed in the next section.

2.2 Loyalty in relation to profit

Previous research found different ways in which loyalty can relate to profit. First of all, Chaudhuri and Holbrook (2001) have found two ways in which loyalty can relate positively to profitability. Purchase (or behavioral) loyalty leads to greater market share whereas attitudinal loyalty leads to a higher relative price for the brand. Moreover, attitudinal loyalty also incorporates a strong positive attitude, which is likely to result in word of mouth (WOM) advertising, with customer acquisition benefits. Other benefits of attitudinal loyal customers are that they have a greater tolerance of negative experiences and lower price sensitivity, which also infers the reduced need for incentives to generate repeat purchase, which leads to more profitable customers (Fitzgibbon and White, 2004). Attitudinally loyal customers are considered to be more profitable than behaviorally loyal customers (Chaudhuri and Holbrook, 2001; Reinartz and Kumar, 2002). Dick and Basu (1994) and Amine (1998) agree on the argument that behavioral loyalty cannot exist without attitudinal loyalty by stating that however repeat purchases are an expression of customer loyalty, a positive attitude ensures that previous purchase behavior will continue.

In sum, behavioral loyalty relates to profit in the following ways:

 Greater market share

12

 Higher relative price

 Strong attitude results in WOM

 Greater tolerance of negative experiences

 Lower price sensitivity

It is necessary to measure both parts of loyalty separately to find out what drives customers in their buying behavior. For example, a customer may be driven by functional attributes, quality attributes or simply by price. Attitudinal loyalty measures will help brand mangers understand (1) why and for what reasons, customers purchase their brands as well as those of their competitors, and (2) what are the strengths and vulnerabilities of their brands (Bandyopadhyay and Martell, 2007). Moreover, consumers might have a positive attitude towards a product but are not able to buy the product so behavioral loyalty will be low. Examples are that the product is not available in the store where the person shops or a person does not have the financial resources to buy the preferred product. This means these people are “forced” to buy a brand they do not prefer and might even be behavioral loyal to it. Jones and Sasser (1995) call such customers “hostages” because they are forced to buy this brand and this suggests that they are likely to defect to another brand when the preferred alternative becomes available.

Bandyopadhyay and Martell (2007) made a classification of behavioral and attitudinal loyal customers, see table 1. A person can be strong or weakly attitudinal loyal and the level of behavioral loyalty is divided into three levels: single users who use only one brand, multiple users, who use multiple brands and non-users, who do not use your product at all. This table gives a good view about how both parts of loyalty are needed in order to create “true loyalty”. This table is referred to several times throughout the thesis and is used as a guideline in the second part of the thesis, the segmentation.

Table 1. Consumer brand loyalty classification according to their behavioral and attitudinal characteristics (Bandyopadhyay and Martell, 2007)

Attitudinal loyalty Behavioral loyalty

Single users Multiple users Non-users

Strong Brand loyal Variety seeker Potential buyer

13 After studying general literature about behavioral and attitudinal loyalty, it can be said that both types are important: Where behavioral loyalty often directly leads to profit, attitudinal loyalty is more indirectly related to profit. Next, it will be discussed how loyalty differs across industries according to literature and hypotheses are presented. Bandyopadhyay and Martell’s classification (Table 1.) will be used to describe the differences.

2.3 Loyalty across industries

Reinartz and Kumar (2002) studied organizations in four different industries: the before mentioned high-tech corporate service provider, a large U.S. mail-order company, a French retail food business and a German brokerage house. Although the general finding, that the results of (behavioral) loyalty are less than expected, is overall the same, differences can be found per company. For example, it appeared that 15% of short-term customers in grocery retail are actual profitable, whereas 36% of long-term customers in this industry are. For the corporate service provider this difference is much smaller; 20% for short-term customers against 30% for long-term customers. These results indicate differences per industry and therefore the effect of loyalty should not be generalized over industries.

A Danish research showed differences across industries in how satisfaction leads to behavioral loyalty (Martensen, Grønholdt and Kristensen, 2000). Because differences occur in this relationship it is likely that also differences occur in other types of relationships. Therefore, it is expected that differences can be found between behavioral and attitudinal loyalty in relation to profit.

Each specific industry has its own characteristics which are explained in the next section. For example, the Fast Moving Consumer Good (FMCG) industry is known for its low-involvement products (Mittal and Lee, 1989), switching costs are characterizing contractual settings (Aydin and Özer, 2005) and touristic products can be defined as high risk purchases(McKercher and Guillet, 2010). Because of these specific characteristics, it is expected that the two types of loyalty play a different role.

Taking all the above into account, the following hypothesis is composed:

H1: The relationship between loyalty and profit differs across industries.

14 2.4 Fast Moving Consumer Goods industry

Brand loyalty in grocery markets, unlike durables or the financial services, is never likely to be absolute. It will always be a relative behavior since consumers tend to purchase from a portfolio of brands (Uncles, Ehrenberg and Hammond, 1995). This may, however, differ per product category. In the FMCG industry it is expected that involvement and habitual buying play an important role in predicting loyalty.

2.4.1 Involvement

Zaichkowsky (1985) defined product involvement as ‘a person’s perceived relevance of the object based on inherent needs, values, and interests’. Involvement is proven to be strongly linked to repurchasing intention as well as higher loyalty levels (Olsen, 2007). There is a very strong relationship between brand involvement and commitment for grocery products (Knox and Walker, 2003). Therefore, involvement is an important predictor of loyalty for consumer goods. Most people are generally involved at a low to a medium level by FMCG (Mittal and Lee, 1989). This is because the risk is low and therefore people are relatively likely to switch. Although, generally speaking the risk is low, some distinction can be made. In the research of Mittal and Lee (1989) the authors have selected two groups. The first group contains products with which people are involved at a medium level. These products are newspapers and toothpaste. The second group contains products to which people are lowly involved: kitchen towels and tinned tomatoes. Kuenzel and Musters (2007) also made a distinction in the level of involvement in FMCG. According to their research, people are highly involved with milk, mayonnaise and soft drinks. Three products people are very low involved with are table salt, bouillon cubes and pasta. Thiele (2005) and Rundle-Thiele and Bennett (2001) suggest that in consumable markets characterized by high switching, low involvement and low risk, behavioral measures can predict brand loyalty levels appropriately.

2.4.2 Inertia

15 service become increasingly automated as a consequence of repeated experience with that service (Murray and Häubl, 2007). The statement of Chaudhuri (1999) suggests that behavioral loyalty is mainly causing profit boosts and attitudinal loyalty is to a lesser extent related to profit in the FMCG sector.

Taking all these arguments into account, due to the low level of involvement it is expected that behavioral loyalty is to a greater extent important in the FMCG industry than attitudinal loyalty. It is also expected that people are more loyal as they become more involved with a certain product. This leads to the following hypotheses.

H2a: For FMCG, behavioral loyalty is to a higher extent related to profitability than is attitudinal loyalty.

H2b: For FMCG, the level of involvement is positively related to attitudinal loyalty which in turn relates positively to profit.

H2c: For FMCG, inertia is positively related to behavioral loyalty which in turn leads to profit.

16 Figure 2. H2a, H2b and H2c in a conceptual model for the FMCG industry

2.5 Contractual settings: telecom and health insurances

It is expected that loyalty plays a different role in contractual settings than in the FMCG industry explained in the previous section. This is due to several factors that characterize this situation. The telecom and health insurance industry both act in contractual settings. Although they have clear distinctions, they also have characteristics in common. The most important issues for these industries are switching behavior, inertia, competition and satisfaction. These issues are discussed next.

2.5.1 Switching costs

Switching behavior is a relevant topic when discussing contractual settings. Switching (or churn) behavior is the opposite of customer loyalty (Wieringa and Verhoef, 2007). When a customer is very loyal, he is not likely to churn, whereas non-loyal customers are very likely to churn. Switching costs are the aspects that customers take into consideration before they decide to switch or not. Especially in contractual settings this is a relevant topic. When a contract is about to end, the customer has to make a decision whether to extend the contract or to switch to a different provider. Switching barriers, or switching costs, can prevent a customer from switching to a competitor. Switching costs consist of the time, money, and

17 effort that consumers expect will be required of them to ensure compatibility between new purchases and earlier investments (Klemperer 1995). Klemperer (1995) posits that switching costs arise from several sources: (a) need for compatibility with existing equipment, (b) transaction costs, (c) uncertainty about the quality of untested brands, (d) discount coupons and similar devices (e.g., loyalty program), and (e) psychological costs of brand loyalty (or noneconomic brand loyalty). Compatibility with existing equipment is expected to be more applicable for the telecom industry, whereas psychological costs or transaction costs might be more applicable for the health insurance market. Aydin and Özer (2005) found that switching costs are positively related to behavioral loyalty in contractual settings which means that when switching costs are high, consumers are likely to stay with their current provider.

Switching costs are typically not directly observed. Consumers are also typically heterogeneous in their baseline preferences for products (Dube, Hitsch and Rossi, 2009). Therefore, switching costs have to be measured individually and switching costs are referred to as perceived switching costs, since individuals can perceive switching costs differently.

2.5.2 Inertia

Whereas theory about switching costs implicitly assumes a rational, explicit decision to stay with the company, inertia diverges from this explicit decision and instead occurs because customers engage in habitual buying behaviors. That is, customers keep buying from the same supplier just because they have done so before and without explicitly considering possible alternatives (Wieringa and Verhoef, 2007). The topic of inertia was also discussed in the FMCG industry. In this industry inertia occurs simply because it is not worth time evaluating alternatives (Ehrenberg, Barnard and Scriven, 1997). In contractual settings inertia is assumed to be highly related to switching costs.

High-inertia customers are passive in contemplating switching, thereby resulting in them remaining status quo with their service providers. Low-inertia customers may be actively searching for and comparing deals, consequently leading to their defections (Colgate and Lang, 2001; Zeelenberg and Pieters, 2004). In this regard, it is expected that high-inertia consumers are staying with their current provider because of a lack of motivation to assess alternatives which causes behavioral loyalty.

2.5.3 Competition

18 Hence, when people experience high switching costs, they are less likely to evaluate alternative offers and when people perceive low switching costs, they are more open-minded to alternatives. However, both the telecom and the health insurance industry can be considered highly competitive during the last years because of privatization in the market which causes competition (van de Pol and Seumeren, 2004).

According to Seiders, Voss, Grewal and Godfrey (2005) competitive intensity describes to what extent direct competition takes place in the immediate environment of a firm. Due to the fact that mobile phones have become highly important in daily life and the various providers, the telecom industry can be considered highly competitive. Also the Dutch health insurance industry can be defined as competitive. Where the government used to influence the health care system mostly, it now has become a competitive market with a growing number of providers (van de Pol and Seumeren, 2004).

According to Jones and Sasser (1995) highly competitive markets are characterized by many alternative products or services offered, the costs of switching is low, or the product is not important to the buyer. In competitive markets, consumers are being targeted with all kinds of different marketing instruments by several organizations. This increases the awareness of services or products offered by competing service organizations (Bolton, Lemon and Verhoef, 2004).

This indicates that when people are to a higher extent price sensitive they are more open to alternatives and perceive lower switching costs. This leads to less inertia, less loyalty and less profit.

H3a: In contractual settings behavioral loyalty is to a higher extent related to profitability than is attitudinal loyalty.

H3b: In contractual settings, switching costs relate positively to inertia.

H3c: Inertia relates positively to behavioral loyalty which in turn relates positively to profit

Note that h3c hypothesizes the same relationship in contractual settings as does h2c in the

FMCG industry.

2.5.4 Satisfaction

19 (1998) in which they argued that a high degree of customer satisfaction does not always translate into behavioral loyalty. They found that even if a customer had reported a high level of satisfaction, they still possessed a predisposition to switch service supplier. It might be clear that satisfaction does not necessarily lead to loyalty but generally, satisfaction is believed to be positively related to behavioral loyalty (Mittal and Kamakura, 2001; Szymanski and Henard, 2001).

In a highly competitive environment satisfaction relates differently to behavioral loyalty than in a less competitive environment (figure 3. Jones and Sasser, 1995).

Figure 3a. The relationship between satisfaction and loyalty in a competitive environment (Jones and Sasser, 1995)

Figure 3b. The expectation of the relationship between satisfaction and loyalty in the telecom and health insurance industry.

Figure 3 shows that true loyalty can be created in a competitive environment, but only when satisfaction is very high. Hence, when a customer is not perfectly satisfied, the degree of loyalty decreases fast and exponentially. In the research of Jones and Sasser (1995) the automobile industry was considered to be highly competitive (figure 3a.). As explained in chapter 2.5.4. the telecom and health insurance industry are considered to be highly competitive due to privatization of the market. Therefore, the relation between satisfaction and loyalty is expected to develop in the same, exponential, way as the line that depicts automobiles (figure 3b).

20 positive at a declining rate. This is not consistent with figure 2. due to the fact that Helgesen did not take competition into account. According to Jones and Sasser (1995) (figure 3) especially in markets with high competition, only completely satisfied customers are truly loyal; even the slightest point of dissatisfaction has tremendous consequences for loyalty.

Jaiswal and Niraj (2011) found that attitudinal loyalty has a mediating role in the relationship between satisfaction and behavioral intentions. This leads to the following hypothesis:

H3d: Satisfaction relates positively to attitudinal loyalty which in turn leads to behavioral loyalty

Taking all hypotheses into account, the following model visualizes the research expectations for contractual settings.

Figure 4. H3a-H3d in a conceptual model for contractual settings

21 2.6 Tourism

Due to changes in lifestyle over years (including changes in work patterns, travel needs, eating habits, and the development of a cosmopolitan community), services offered by hospitality businesses are considered to be necessities, rather than luxuries (Kandampully and Suhartanto, 2003). Consequently, during the past decade, there has been an exponential growth in hospitality businesses to meet the demands of the growing market. This has provided consumers with a great variety of choices while simultaneously augmenting competition in the marketplace (Kandampully and Suhartanto, 2003). Therefore, hospitality organizations today are faced with intense competition, and the associated challenge of steady growth in a competitive environment. In this special environment risk aversion and word-of-mouth are expected to be of great importance. These topics are discussed in the next sections.

2.6.1 Risk aversion

A touristic product is something that must be purchased at a distance, without the opportunity to pretest before buying. As such, an elevated level of risk of a poor experience is associated with trying a new destination, whereas it is lessened when returning to a known one (McKercher and Guillet, 2010). Risk aversion, therefore, may induce attitudinal loyalty (Jones, Mothersbaugh, and Beatty 2000). According to Chaudhuri and Holbrook (2001), in situations of uncertainty, trust reduces the uncertainty in an environment in which consumers feel especially vulnerable because they know they can rely on the trusted brand. This would imply that consumers who trust a certain brand or organization are likely to be loyal in order to avoid risk.

In this thesis, tourism is specified to bungalow parks. Hence, the concept of tourism is too broad to generalize since the definition of tourism includes three aspects, that is destination, transport and accommodation (Cooper, 2008). With 5.6 million holidays in The Netherlands (NBTC NIPO Research, 2010), which is 41% of all domestic holidays, is the bungalow sector rather large.

However, there are theories about travel motives of wanderlust (Crompton 1979), novelty seeking and self-development (Pearce and Lee, 2005) which would induce non-loyalty, these are not likely to be applicable for the Dutch bungalow market.

The theory about risk avoidance, on the other hand, is more likely to be true: Because people want to avoid risk, they will form a positive attitude towards the bungalow park which will lead to more profit.

22

2.6.2 Word of Mouth

Word of mouth (WOM) is considered to be an important factor for all kinds of products, however, it is an especially important factor in the tourism industry. Hospitality and tourism marketers find the issue of critical importance for the following reasons:

 hospitality and tourism product offerings, as intangible goods, cannot be evaluated before their consumption, thus elevating the importance of interpersonal influence (Lewis and Chambers 2000);

 many hospitality and tourism products are seen as high-risk purchases, for which the emotional risk of reference group evaluation is an important aspect of the decision making process (Lewis and Chambers 2000);

 the hospitality and tourism industry is intensely competitive, suggesting that the use of interpersonal influence may provide important competitive advantages for early adopters; (Litvin, Goldsmith and Pan, 2005).

 and finally, considering the dearth of hospitality and tourism industry specific literature related to the issue, it would appear that the industry lags behind others in the development and discussion of strategies for managing interpersonal influence in an electronic environment (Litvin, Goldsmith and Pan, 2005).

The introduction of the internet made the influence of WOM even stronger; websites like tripadvisor.com and zoover.nl created Electronic Worth of Mouth (eWOM) in the tourism industry. Vermeulen and Seegers (2008) have proven that positive reviews improve attitudes toward hotels. This is referred to as ingoing WOM. Furthermore, Chaudhuri and Holbrook, (2001) state that attitudinal loyalty enhances WOM . This is called outgoing WOM. This implies that the relationship between WOM and attitudinal loyalty is two-sided. This leads to the following hypotheses:

H4a: In the bungalow park market, attitudinal loyalty is to a higher extent related to profitability than is behavioral loyalty.

H4b: A higher degree of risk avoidance relates positively to attitudinal loyalty which in turn relates positively to Outgoing WOM

H4c: Ingoing WOM relates positively to attitudinal loyalty which in turn relates positively to profit

23 Figure 5. H4a – H4d in a conceptual model for the bungalow park industry

24 Taking all industries together the following overall model (Figure 6) is created. This model does not show all individual relations but give an overview of all hypothesized variables that influence either behavioral or attitudinal, which is expected to relate to profit. This simplified model represents the relationships that are used for the econometric model. In order to check for pooling, all relationships should be measured the same way. How this works in practice is explained in the methodology section.

Figure 6. All variables in relation to loyalty and profit

25 Summarizing the theoretical underpinnings, the hypotheses are given schematically in table 2. Table 2. Summary of hypotheses

Industry Characterized by Loyalty Hypothesis

H1: The relationship between loyalty

and profit differs across industries. FMCG Low involvement Behavioral H2a: For FMCG, behavioral loyalty is

to a higher extent related to profitability than is attitudinal loyalty.

H2b: For FMCG, the level of

involvement is positively related to attitudinal loyalty which in turn relates positively to profit

H2c: For FMCG, inertia is positively

related to behavioral loyalty which in turn leads to profit.

Contractual settings: telecom and health insurances High switching costs, inertia, satisfaction

Behavioral H3a: In contractual settings

behavioral loyalty is to a higher extent related to profitability than is attitudinal loyalty.

H3b: In contractual settings,

perceived switching costs relate positively to inertia

H3c: Inertia relates positively to

behavioral loyalty which in turn relates positively to profit.

H3d: Satisfaction relates positively to

attitudinal loyalty which in turn leads to behavioral loyalty

Tourism – Bungalow market

High risk Attitudinal H4a: In the bungalow park market,

attitudinal loyalty is to a higher extent related to profitability than is behavioral loyalty.

H4b: A higher degree of risk

avoidance relates positively to

attitudinal loyalty which in turn leads to outgoing WOM

H4c: Ingoing WOM relates positively

26 H4d: Outgoing WOM relates

positively to indirect profit

3 Research design

In this chapter it is explained how the research is conducted in order to test the hypotheses.

3.1 Research methods

With the purpose of testing the hypotheses a quantitative, online research takes place in the form of a questionnaire. Quantitative research can especially be useful when researching a large amount of people and gives a good insight in their opinion (Malhotra, 2009).

In order to find out if the hypothesized characteristics indeed are more applicable to the specific industries, the relevant variables have to be measured for all industries. To measure a variable, one or more questions are asked which means that 18 questions are asked per industry. The questionnaire would become too long when all industries are presented to one respondent. This has the risk that respondents become bored and this affects the quality of the response. Therefore, the industries are divided into five questionnaire blocks; for every industry one block and the FMCG in two blocks, namely one for a low involvement product (pasta) and one for a high involvement product (soft drinks) according to the research of Kuenzel and Musters (2007). The blocks are designed in the same way so the results can easily be compared to each other. The questionnaire starts with a screening question about which products the consumer uses. After the screenings question the respondents get the block(s) about the products relevant to them. They get to fill out one, two or maximum three blocks in order not to become boring. The time it takes a respondent to fill out one question block is approximately five minutes per block.

Table 3. Measures of variables

Variable Relevant

for industry

Measuring instrument Questions

Behavioral loyalty (BL) FMCG, contractual settings, tourism

2 items, 7 points Likert scale (Chaudhuri and Holbrook, 2001)

Mittal and Lassar (1998) advise to use multiple item scale

• I will buy this brand the next time I buy product X (BL1) • I intend to keep purchasing this brand. (BL2)

Multiple item scale: which of the following situation is the most applicable to you? • I always buy this brand, no matter what / I never buy this brand (BL3)

Attitudinal loyalty (AL) FMCG, contractual setting, tourism

2 items, 7-points Likert scale Chaudhuri and Holbrook (2001)

• I am committed to brand X (AL1)

• I would be willing to pay a higher price for this brand over other brands (AL2)

Profit (P) FMCG and tourism

1 item.

Share of Wallet:

Based on Wirtz, Matilla and Lwin (2007)

• Estimate how often (%) you buy brand X compared to other brands of that product? (P1)

Contractual setting

Wirtz et al.(2007) • What do you pay on average on your monthly telephone bill / health insurance? (P1) Involvement (INV) FMCG, contractual settings, tourism

3 items, 7 points Likert Scale (Mittal and Lee, 1989)

• I have a strong interest in product X. (INV1) • Product X is very important to me. (INV2) • For me product X does not matter. (INV3)

Perceived switching costs (SC)

Contractual setting

3 items, 7 points Likert scale, according to the method of Jones et al. (2002)

• It would take a lot of time to switch to a new provider/brand (SC1)

• It costs me a lot of money to switch to another provider/brand (SC2)  not appropriate for FMCG

• It would take a lot of effort to switch to another provider/brand (SC3) Inertia (I) FMCG,

Contractual setting,

1 item, 7 points Likert scale (Wieringa and Verhoef, 2007)

28 tourism Satisfaction (SAT) FMCG, Contractual setting, tourism

3 items, 7 points Likert scale (Shin and Kim, 2007)

• I am satisfied with the current service (SAT1)

• The current service meets all the requirements that I see reasonable (SAT2) • The ratio between quality and price is good (SAT3)

Risk avoidance (RA) FMCG, contractual setting, tourism

2 Items, 7 points Likert scale (Mittal and Lee, 1989; Jones et al., 2002)

• It is not a big deal to buy another brand/ contract/ bungalow park. (RA1)

• I am not sure what the level of quality (FMCG)/service(contract and bungalow) would be if I switched to a new brand/contract/bungalow park. (RA2)

Ingoing WOM (IWOM) FMCG, contractual setting, tourism

5 items, 7 points Likert scale (Jager, 2011)

• I ask others for advice about product X (IWOM1) • Others advise me without asking for it (IWOM2)

• I visit comparing websites or check information from consumer organizations (IWOM3)

• I appreciate other’s advice (IWOM4)

• I prefer a product that impresses others (IWOM5) Outgoing WOM (OWOM) FMCG, contractual settings, tourism

1 Item, Net Promoter Score (Reichheld, 2003), scale 0-10

• How likely is it that you would recommend brand X to others? (OWOM)

Indirect profit (INP) FMCG, contractual settings, tourism

1 item 6 point scale (Chaudhuri and Holbrook, 2001)

3.2 Data collection

The questionnaires are distributed via the EUpanel of DirectResearch, which contains 27.000 members and is representative for the Dutch population. The norm is n=300 respondents per industry in order to be reliable. To make sure the respondents represent the Dutch population well, demographic criteria according to the ‘Golden Standard’ (CBS) are taken into account. This standard ensures people from all age groups, different genders, educational levels and living regions are represented. The data is cross-sectional since the data is obtained at one point in time and contains multiple subjects.

3.3 Plan of analysis

Next, the plan of analysis is described. This contains designing the database, and discussing the statistical tests and analyses that have to be performed in order to reject or accept the hypotheses.

3.3.1 Designing the database

In order to perform the necessary analyses, the raw data has to be transformed to usable variables.

3.3.1.1 Transforming variables

The aggregation level of the obtained data is on case level instead of respondent level. Hence, a respondent filled out maximum three blocks, which equals three cases. In order to transform the data, the measures of all items should be comparable. The variable Profit (P) is measured differently in contractual settings compared to the non-contractual settings. A transformation has taken place for the contractual settings in order to make the variables comparable. The values now all range from 0 to 100.

Another action that has to take place is that several items are inverse measures. This contains the variables INV3, BL1 and RA1 in which a value of 1 indicates a high level of that variable and 7 a low level. The values of these items are converted so they all measure on the same level.

3.3.1.2 Transforming multiple items into one variable

30 be taken. In order to measure the reliability of the items Cronbach’s alpha is used. This measure ranges from 0 to 1. With a value of .7 or higher, the items are found to be internally consistent (Hair et al. 2010). When α < .7 the alpha could be increased by deleting one of the items. If this is the case and α can become > .7 the item will be deleted and not be taken into account when creating the new variable.

When a Cronbach’s alpha is not sufficient (<.7) a factor analysis is conducted in order to see if deleting the item is the best option. The goal of the factor analysis is reducing data. The best option for a factor analysis with this goal is a Principal Component Analysis (PCA). PCA is used when the objective is to summarize most of the original information (variance) in a minimum number of factors for prediction purposes (Hair et al., 2010). There need to be done two separate factor analyses, one for the independent variables and one for the mediating variables. Hence, these variables have a different purpose in the analysis and correlation can affect the outcomes. The PCA with the independent variables should create five components; one for each variable. The same is applicable for the PCA for the mediating variable BL and AL which should create two components. The eigenvalue of each component should exceed 1.0. In order to better fit the components a rotation method is selected. The effect of rotation is to redistribute the variance from earlier factors to later ones to simpler, theoretically more meaningful factor pattern (Hair et al., 2010). The most appropriate rotation method is oblique factor rotation. This rotation method turns axes not retaining the 90 degrees angle. With this method the obtained components do not need to be orthogonal which means correlation between components can occur. This is likely to be the case since the variables are expected to be correlated to each other. Furthermore, this rotation method is advised when outcomes want to be generalized (Hair et al., 2010). The outcomes of the PCA will show which items are allocated to which component. For a sample size of n=300 a factor loading of .35 is necessary to be significant (Hair et al., 2010).

31

3.3.2 Econometric model

Once the database is designed, the actual content of the database can be analyzed. The relationship between the independent variables, loyalty and profitability will be measured using an econometric model in which profitability is the dependent variable, the loyalty scores are the mediating variables and the specific characteristics are the independent variables. 3.3.2.1 Functional form

The functional form can be linear or multiplicative. Using multivariate regression analysis it can be tested if the independent variables relate positively or negatively to profitability.

It is likely that there is interaction between the independent variables. In the past, many research has been done that indicate relationships between variables (Anderson, 1996; Lee, Lee and Feick, 2001; Yang and Peterson, 2004).

Since there is expected to be a lot of interaction between attitudinal and behavioral loyalty a multiplicative model is expected to be most appropriate. When this truly is the case, the model will look as follows:

Where:

= Profitability for industry i for consumer c

= behavioral loyalty for industry i for consumer c = attitudinal loyalty for industry i for consumer c = Involvement for industry i for consumer c

= Switching costs for industry i for consumer c

= Inertia for industry i for consumer c

= Satisfaction for industry i for consumer c

= risk avoidance for industry i for consumer c

= Ingoing Word of Mouth for industry i for consumer c = Outgoing Word of Mouth for industry i for consumer c

= Error term

32 regressions are compared on two aspects. First, the R squares (R2) are analyzed. The R2, which is the squared correlation coefficient, is the value that indicates the percentage of total variation of the dependent variable that is explained by the regression model of independent variables (Hair et al.,2010). The model with the highest R2 explains the most of the variation en is likely to be the best model. Second, the predictive validity is tested by running a regression based on approximately two third of the data. The estimates of these parameters are used to predict values for the third part. The predictive validity of both models is tested by using the Mean Absolute Percentage Error (MAPE) test for predictive validity. This test is dimensionless, easy to relate to, and potentially useful if one wants to make comparisons on forecast accuracy across different settings (Leeflang et al., 2000).

∑

|

̂

|

̂

=

Because this measure reports the error relative to the mean error, the values can be compared over different levels. Hence, the scale of the multiplicative model will be different from the additive model. The MAPE test makes it possible to compare the accuracy of the predictions. The model that predicts best will have the lowest MAPE value and determines the functional form.

3.3.2.2 Pooling

The model is aggregated over the different industries which allows to compare different outcomes. It is expected that the model cannot be pooled since differences are expected according to the studied literature, which is the main research topic of this thesis. A Chow test is used to test whether pooling is appropriate (Chow, 1960). This test tests for the hypothesis that the betas differ too much across industries in order to allow pooling according to the OLS method. Hence, when it appears that the model can be pooled over all industries, H1 cannot be accepted. An advantage of not pooling is that specific data per industry is used.

33 observations per parameter. With n=300 this will not cause any problems. The Chow test looks as follows: Where ∑

= Sum of Squared Residuals = Degrees of Freedom

= Standard Error of the Estimate = Number of observations = Number of parameters

= Number of cross-sectional units

3.3.2.3 Model assumptions

General model criteria are that the model must be simple, evolutionary, complete, adaptive and robust. Furthermore, estimates of parameters (bias) or estimates of the variance of the parameters can be wrong (Leeflang et al., 2000). Wrong estimates of the variance of the parameters relate to the disturbance term assumption. The conditions the residuals should satisfy are:

34 reformulating the model or creating new variables. This might be the case when variables are measures approximately the same phenomenon. If this occurs, the choice can be made to combine these variables.

 The residuals should be homoscedastic which means that the residuals have the same variance for all possible values of a predictor variable. Heteroscedasticity is the event of obtaining unequal variances of the disturbance term. When this occurs the result is a wrong estimate of the variance of effects and re-estimation with GLS has to take place.

 The residuals are normally distributed.

 The mean of the residuals is zero.

Wrong estimates of parameters can cause the following problems:

 Omitted variable bias. When an important variable is not estimated, the other variables will capture this effect and therefore these parameters are biased. It is, however, hard to overcome this bias since it is almost impossible to build a perfectly complete model.

 Wrong functional form. This can be the case when a linear model is predicted but the variables are not linearly related to the dependent variable.

 Endogeneity. There should be no relation between the error term and the independent variables. When this is the case this is called exogeneity.

 Non-constant parameters can be obtained whenever is chosen for a pooled model when this was not allowed. The Chow-test, therefore, needs to be carefully performed.

To make sure these problems will not occur, the assumptions need to be checked carefully. How the problem will be solved when one of the assumptions is not met, is discussed in the result chapter.

3.3.3 Hypothesis testing

35 between the industries which means H1 can be rejected. As described before, a Chow-test is

conducted to test this.

H2a, H3a and H4a are all build up the same way: they state that either behavioral or attitudinal

loyalty is more positively related to profitability for each industry. This can be tested by performing a t-test with the betas of the variables:

In which:

β1 = Estimate for behavioral loyalty

β2 = Estimate for attitudinal loyalty

SE = Standard Error of the estimate

In order to perform calculations with this equation, it needs to be rewritten. This is explained in appendix A.

H2b assumes a positive relationship between involvement and attitudinal loyalty,

36 Figure 6. The mediation chain Baron and Kenny, 1986)

The way to test whether this is indeed the case, several regressions have to be performed (Preacher and Hayes, 2004). First, a simple regression is conducted between involvement (independent) and profit (dependent); relation C in the figure 6. Next, a simple regression is conducted between involvement (independent) and attitudinal loyalty (dependent); relation A in figure 6. The third step is conducting a simple regression with attitudinal loyalty (independent) and profit (dependent); relation B in the picture. When one of these regressions show a non-significant relation, mediation cannot be accepted. When all regression show a significant relation, the last step can be performed. The final step is conducting a multiple regression with both involvement and attitudinal loyalty as independent variables and profit as dependent variable. If involvement is no longer significant when attitudinal loyalty is controlled, the findings support full mediation. When involvement is still significant when controlling attitudinal loyalty partial mediation is supported (Preacher and Hayes, 2004).

The same is done for hypotheses 2c, 3c, 3d, 4b and 4c, which all assume a mediating

effect on several variables.

Hypotheses 3b and 4d hypothesize a direct effect which is tested by means of a simple

regression.

3.3.4 Segmentation

A second part of the research will try to identify segments according to the extent to which consumers are loyal in the different industries. The segments are based on the classification of Bandyopadhyay and Martell (2007) who identify six segments: Brand loyal, Variety seeker, Potential buyer, Constrained buyer, Deal prone and Indifferent (see table 1). It

37 will be analyzed whether certain segments are represented by a certain industry, for example: In which industry are the most people Brand Loyal and in which are they Indifferent?

The method to perform the segmentation is cluster analysis. The primary goal of cluster analysis is to partition a set of object into two or more groups based on the similarity of the object for a set of specified characteristics (Hair et al., 2010). A cluster analysis is a technique that is used to classify cases into relatively homogeneous clusters. The cases in each cluster tend to be similar to each other and dissimilar to the other cases in the cluster. Cluster analysis is often used to understanding buyer behavior, identifying new product opportunities, selecting test markets and for reducing data (Malhotra, 2010).

One general cluster analysis over all industries takes place with only behavioral and attitudinal loyalty as clustering variables. The others are not taken into account. This way, the chance is the largest that the clusters of Bandyopadhyay and Martell are obtained. The goal of the segmentation is to identify which target group is mostly applicable to what industry. In order to find this out, industry is the most important describing variable.

By creating clusters, organizations can be advised which customers are most loyal in which industry. According to background variables the clusters can be described.

3.3.4.1 Distance measure

Because the measures of the variables are continuous a distance measure is most appropriate. There are several distance measures that can be appropriate; Euclidean distance measure, Squared Euclidean distance measure and City-Block are often used. The best way to find the most appropriate measure is by trying several options (Hair et al.2010).

3.3.4.2 Clustering procedure

38

4 Results

In this chapter, the results of the analyses are discussed. 4.1 Demographic criteria

2350 people have been invited to take place in the research of which 1151 actually completed the questionnaire. This is a response rate of 48%. Table 4 shows the demographic criteria of the sample. Although there are some slight differences per industry, this is not worth mentioning. The demographics correspond approximately with the Golden Standard of the Netherlands which is developed by the Dutch organization of statistics (CBS, 2012). Only educational level is to some extent higher in the sample compared to the Golden Standard. However, this is not expected to be a problem since all groups are represented to a good extent.

Table 4. Demographic criteria

The aggregation level of the data is over cases instead of over respondents. The number of cases differs per respondent; some filled out three and some only one. The number of cases does therefore not equal the number of respondents. The number of cases is n=1577, see table 5.

Sample Golden Standard (source CBS)

Gender Male 52% 50% Female 48% 50% Age 12 - 20 years 3% 5% 21 - 30 years 12% 16% 31 - 40 years 18% 18% 41 - 55 years 31% 30% 56 - 65 years 19% 17% 66 years and older 18% 14%

Education Primary school 1% 10%

39 Table 5. Number of cases per industry

Industry subject Number of cases (n)

FMCG Pasta 303 Soft drinks 354 Telecom 302 Insurances 304 Tourism Bungalow park 313 Total 1577

The first step in analyzing results is analyzing the database for any ‘wrong’ values that might disturb the results which might be caused by unmotivated respondents or wrong interpreted questions. These cases are removed from the database in order to keep the database ‘clean’ and results are not biased by these respondents.

4.2 Transforming variables

As described in the previous chapter the variables and items have to be transformed into variables that can be used for the analyses. First the Cronbach’s alpha is used to measure the internal consistency of the items measuring the relevant variable, see table 6.

Table 6. Cronbach’s alpha

Total Pasta Soft drinks

Telecom Insurance Bungalow

INVOLVEMENT .737 .880 .877 .770 .570 .867 BEHAVIORAL LOYALTY .812 .769 .832 .755 .820 .815 ATTITUDINAL LOYALTY .726 .707 .770 .617 .681 .799 SWITCHING COSTS .778 .548 .548 .804 .761 .930 SATISFACTION .807 .792 .712 .830 .832 .825 RISK AVOIDANCE .542 .709 .476 .544 .487 .436 INGOING WOM .836 .852 .830 .738 .794 .812

Because the value does not always exceeds .7 a factor analysis is performed to see if the items can be used to measure the variable. Two separate factor analyses are performed to not confuse the independent variables with the mediating variables.

40 (PCA) for the industries telecom and insurances gives a very good output, using an oblique rotation method. The analysis created five components, based on an eigenvalue of at least 1, a KMO value of .716 and a significant value for Bartlett’s test of Sphericity (p =.000). The items are divided over the components as shown in table 7. This factor analysis shows that the items can be grouped together in the relevant variables.

Table 7. Factor scores independent variables for telecom and Insurances (includes Switching Costs 3)

Component

Ingoing WOM Satisfaction Switching Costs Involvement Risk Avoidance

Involvement1 0.22 0.03 -0.08 -0.75 -0.11 Involvement2 0.11 0.16 -0.01 -0.81 -0.01 Involvement3 0.08 0.02 -0.01 -0.70 0.13 SwitchingCosts1 0.18 -0.08 0.87 -0.03 -0.06 SwitchingCosts2 0.14 -0.05 0.90 0.03 0.13 SwitchingCosts3 -0.05 -0.07 0.70 0.09 0.27 Satisfaction1 -0.15 0.89 -0.04 -0.16 0.13 Satisfaction2 -0.16 0.84 -0.06 -0.06 0.12 Satisfaction3 -0.17 0.86 -0.07 -0.02 0.11 RiskAvoidance1 -0.19 0.25 0.10 0.05 0.71 RiskAvoidance2 -0.07 0.16 0.22 0.03 0.74 IngoingWOM1 0.82 -0.23 0.16 0.02 -0.05 IngoingWOM2 0.75 -0.15 0.09 -0.08 0.03 IngoingWOM3 0.73 -0.11 -0.01 -0.29 -0.08 IngoingWOM4 0.83 -0.15 0.05 -0.26 0.09 IngoingWOM5 0.38 -0.13 -0.04 -0.15 0.54

Table 8 shows the factor scores for the industries pasta, soft drinks and bungalow parks, also using an oblique rotation method, based on an eigenvalue of >.1. The KMO value is .729 which is high enough to be considered an appropriate factor analysis. Table 8. however, shows some contradicting scores for risk avoidance and involvement3.

Table 8. Factor scores independent variables for pasta, soft drinks and bungalow parks (without Swichting Costs3)

Component

Ingoing WOM Satisfaction Switching Costs Involvement Risk Avoidance

Involvement1 .12 .14 .00 -.95 -.00

Involvement2 .10 .18 .10 -.94 .04

Involvement3 .02 .03 -.04 -.03 .79

41 SwitchingCosts2 .09 .09 .85 -.06 .28 Satisfaction1 -.13 .78 .04 -.18 .17 Satisfaction2 -.10 .88 .01 -.14 .07 Satisfaction3 -.11 .85 .05 -.10 .01 RiskAvoidance1 -.17 .18 .48 .01 .63 RiskAvoidance2 .09 .15 .54 .01 .58 IngoingWOM1 .85 -.16 .06 -.05 -.06 IngoingWOM2 .79 -.08 .02 -.06 .10 IngoingWOM3 .79 -.06 .02 -.17 -.06 IngoingWOM4 .89 -.13 .05 -.11 -.05 IngoingWOM5 .73 -.12 .19 -.07 -.03

Additionally, another factor analysis is conducted in which all industries are measured separately. The output of these analyses can be found in appendix B. These factor analyses are, however, not perfect as well for several reasons. First, there is the KMO value which is lower than .7 for the industries telecom and bungalow. Second, the factor scores for risk avoidance do not fit appropriately to one specific component in all industries. Because of these reasons, it is chosen to not use the factor scores but to create weighted summated scales. These weighted scales are based on the correlation of the Cronbach’s alpha value which takes the mutual importance into account. This way five new independent variables are created.

Also, a factor analysis is conducted for the mediating variables behavioral loyalty and attitudinal loyalty. The factor analysis for these variables does not give a sufficient outcome either. When creating components based on an eigenvalue >1 only one component is extracted, both in the overall analysis and split per sector. Therefore, a fixed number of components is set to extract, namely two. Doing this, the items are placed in the right component, although it is not satisfactory since the items in some cases have high scores for both components (appendix C.). Because the outcomes of the factor scores are not perfect but they do show a certain tendency about grouping the variables it is chosen to use the weighted summated scales to create the new variables.