An empirical study into why brand equity helps when service fails

(1)

(2)

An empirical study into why brand equity helps when service fails

Master Thesis

Rob Armand Philips Jacob van der Doesstraat 54B

2518 XP Den Haag

S#: S1975919 R.A.Philips@gmail.com

0627621728

1st_{supervisor: Dr. J. van Doorn} 2nd_{supervisor: Dr. J.A. Voerman}

20 / 02 / 2012

(3)

Preface

The thesis is a topic that has eluded me twice, as I transitioned from International Business and Languages at the Hanzehogeschool to a short-track pre-master programme, in which I did not have to do a bachelor’s thesis. For me, this has proven to be a mixed blessing. On the one hand, I found myself very motivated to find a topic that interested me, one that I could invest in. On the other hand, I had never made a thesis, which sometimes proved challenging.

When looking into topics for my thesis, I found that my research interests lay in the field of brand equity. In addition, I have always had an interest in transportation, and the culmination of these two things is this master thesis. I was particularly fortunate that I could use an existing data set of a survey that was conducted among customers of a large Dutch transport provider.

I would therefore like to thank Dr. Jenny van Doorn, my supervisor, for making it possible to use this resource. I would also like to pay tribute to her continued help and vision during the entire process. In addition, I would like to thank my second supervisor, Dr. Liane Voerman, for reviewing my thesis and lastly, my family and close friends, for supporting me during the final phase of my study.

(4)

Management Summary

The main goal of this thesis is to find out whether there is a direct effect of brand equity on the perception and evaluation of a service failure by the customer. This thesis looks at three subjective measures that portray well how a customer evaluates a failure, which are also used often in service failure literature: attribution of blame towards the firm, the severity of a failure and the recovery by the company. Brand equity is looked at by means of brand connection, which is an important part of brand equity, and can be described as the psychological manifestation of brand equity. A direct effect of brand connection on a customer’s evaluation of a failure would mean that a customer with a high brand connection attributes less blame to the company, rates an incident as less severe and gives higher scores on recovery actions by the firm. These effects are expected because the strong, favourable and unique associations of the established brand equity could conflict with the negative information of a failure. A second important goal of this thesis is to see what happens to the hypothesised effect in case of multiple service failures.

(5)

(6)

1| Introduction

Unfortunately, service failures are to be expected in some encounters between customers and companies, both in the delivery process and in the core service itself (Smith, Bolton and Wagner, 1999). The transportation sector can provide an interesting example, as the delivery process and completion of a service can take some time, allowing for failures to occur. The Economist recently wrote an article (Anon., 2011) in which it described a flight delay caused by over-fuelling by the pilot. 37 customers were asked to disembark the plane voluntarily. Of course, no one was keen to give up their flight, and 30 passengers had to be ordered off the plane (the whole situation was exacerbated by the fact that it took place on Boxing Day), making for a terrible customer experience. This extreme case clearly illustrates the severe impact service failures can have on customers, and that the mitigation of these incidents by companies is critical.

1.1 Problem background and relevance

Service failures are likely to have a negative effect on the customer. A dissatisfied customer, as opposed to a satisfied customer, may sooner switch to a competitor (Keaveney, 1995), engage in negative word-of-mouth (Hoyer and Macinnis, 2008) or compound the negative experience in his or her attitude towards the company, ultimately affecting outcome measures such as customer satisfaction and loyalty (Bitner, Booms and Tetrault, 1990), and retention (Sloot, Verhoef and Franses, 2005). It is important for researchers and managers to understand customer behaviour during service failures, what the potential effects of failures are, and what can be done to rectify situations in which customers are dissatisfied due to failures. The academic research world has been aware of this importance for quite some time, and the service failure and recovery literature is ample (e.g.: Bittner, Booms and Mohr, 1994; Smith et al, 1999; Maxham and Netemeyer, 2002; Ringberg, Odekerken-Schröder and Christensen, 2007; Roehm and Brady, 2007; Brady, Cronin, Fox and Roehm, 2008).

(8)

The second perspective argues that people with strong, favourable and positive associations (Keller, 2008) have higher expectations than people with lower brand equity (Roehm and Brady, 2007). Additionally, people with a high affective commitment to the firm showed higher expectations which has the potential for greater disappointment and feelings of betrayal if a firm fails (Mattila, 2004).

Studies on the effects of brand equity’s influence on service failures and its antecedents are still relatively new. So far, it has been established that there is indeed a brand equity effect. This effect holds that people who have high customer-based brand equity, provide more favourable evaluations and have more positive behavioural intentions (Brady et al, 2008). The nature of the effect however is not completely clear, as Brady et al (2008) find that despite the advantage, initially a more drastic decline in customer ratings will result for a high equity brand. Additionally, Roehm and Brady (2007) report that recovery must follow as soon as possible after a failure and that severity and distraction must be high for a high equity brand to have a prevailing advantage. In these studies on the effects of brand equity in case of service failure, only one failure (Brady et al, 2008) or two service failures (Maxham and Netemeyer, 2002) have been looked at. This thesis includes a third service failure by the same company, which provides an extended insight in the stability of the effect of brand equity.

1.2 Contribution and goal of this thesis

The contribution of this thesis is to more clearly define what the effect is of brand equity on the perception of multiple service failures by a customer. As discussed above, it has already been established that brand equity moderates the relationship between a failure and satisfaction. This thesis will therefore not look at the moderating role of brand equity, but will investigate if brand connection, an important part of brand equity, has a direct influence on how a customer perceives a service failure. In order to achieve this, subjective measures of the evaluation and rating of a failure by a customer are needed.

(9)

Previous literature with regard to severity has only looked at incidents as being low/high (Roehm and Brady, 2007; Smith et al 1999). This study investigates the severity of an incident more comprehensibly, by including a range of both extreme, medium, as well as smaller failures and is evaluated directly by the customer who experienced the failure. Lastly, recovery is any action that companies undertake to rectify a failure. Recovery research is also ample (e.g.: Bitner et al, 1990; Keaveney, 1995; Smith et al, 1999; Maxham and Netemeyer, 2002), as service recovery is a powerful tool for companies to restore a customer’s satisfaction and loyalty (Smith and Bolton, 1998).

The goal of this thesis is to find out whether the reason that customers with high customer-based brand equity experience failures as less severe, is thanks to their more favourable view on the attribution, severity and recovery of a failure. In order to do so, the following research questions are drawn up:

1 Is there a direct, positive effect of brand equity on attribution, severity and recovery?

2 Does this effect increase, remain stable or wear off after the first service failure, when looking at people who experienced two or three service failures?

Longitudinal data from a large Dutch transportation provider will be used in this thesis. Such a service firm is particularly suitable, given the common occurrence of multiple service failures like a delay in travel time, and should yield interesting results.

1.3 Structure of the thesis

(10)

2| Literature review

This section will review literature on how customers perceive and evaluate service failures and what the company can do to recover service. Additionally, brand equity and an important part of it, namely the self-brand connection, will be described.

2.1 Service failures and the customer’s subjective perception of failure

In order to understand how customers perceive failures, it is important to define what a service failure is and to discuss its context. Before service fails, a company has agreed to provide a service, often in exchange for money from the customer. This mutual agreement leads to a period of time during which a customer directly interacts with a service until it is completed. This is the definition of a service encounter, which was defined in the eighties by Shostack (1985). The article he wrote about service encounters demonstrates that the stage was already being set years ago, attributing to the importance of failures in the service encounter. As the term ‘encounter’ suggests, both the customer and the company and its employees play a role. During this encounter, things can go wrong which will lead to the interaction between the customer and the company to become dissatisfying. Things can go wrong during the delivery of the service, or with the outcome of the service. Concurrently, in literature, two types of service failures have been identified: process failures and outcome failures (Smith et al, 1999). To measure a service failure, the term ‘Critical Incident’ is often used to define a specific interaction between customers and service firm employees that is especially satisfying or especially dissatisfying (Bitner et al, 1990). It is important to note that not only employees or companies are responsible for service failures; Bitner et al (1994) demonstrated in their research that in 22% of dissatisfactory incidents, problem customers were the cause.

(11)

2.1.1 Attribution

Attribution indicates the extent to which a customer holds the seller responsible for a failure (Maxham and Netemeyer, 2002). Attribution is linked to attribution theory, which contends that if a problem’s cause is seen as unstable, purchase intentions are unharmed because the problem is considered unlikely to persist (Folkes, 1988; Brady et al, 2008).

When someone makes a causal attribution, he or she links an event to its cause (Taylor, 1994). Why do people make attributions? The answer to this question roots in psychology and has been well researched within this domain, as the causal attributions that people make have consequences for behaviour, affect and expectancies (Taylor, 1994). The main reason is that every person has a need for prediction and control of one’s environment (Harvey and Weary, 1984; Taylor, 1994). The reason that someone wants to control his or her environment, is that the more the maker of attributions beliefs the other party (a company in the setting of a service failure, for example) has control over rewarding or punishing, the more important it is for the attribution maker to understand the target person’s behaviour and the more he or she is motivated to engage in causal analysis (Harvey et al, 1984). Based on the attribution made, uncertainty for the individual is reduced. How heavily someone will attribute blame to a company depends on the stability and controllability of the failure. In this context, stability refers to the extent to which the cause of a failure is stable and permanent, and could thus occur again. Controllability is the extent to which a company has control or choice over the failure. This means that when a failure is clearly the fault of the company, customers will increase their desire to complain and warn others against the company. A study conducted by Smith and Bolton (1998) confirmed this by finding that stable attributions of failures to the firm negatively affect cumulative satisfaction and repatronage intentions.

(12)

2.1.2 Severity

The severity of a service failure can be described as the magnitude of a failure episode (Roehm and Brady, 2007). Maute and Forrester (1992) find that the more severe a failure is, the more the customer will want to ‘exit’ or distance themselves from the company, voice their dissatisfaction or simply suffer in silence and hope things get better. Customer evaluations decline as service failures become more severe, and within the context of multiple failures, the second failure is rated more severely than the first (Maxham and Netemeyer, 2002). Maxham and Netemeyer’s (2002) results also show that people will see incidents as more severe especially when they were satisfied before the incident. This is because they have higher expectations and the gap between their last positive experience and their current bad experience is likely substantial. People who were already dissatisfied to start with also report increasing severity due to the magnitude of multiple incidents, but not as much as people who were first satisfied.

Severe failures require greater recovery efforts on the part of the firm. The explanation for this effect is that, as the loss for the customer grows larger, they will view the exchange as more inequitable (Smith et al, 1999) and perceptions of unfairness intensify (Seiders and Berry, 1998). At this point, the consequence of a severe incident is that the company must either deliver an appropriate recovery solution or the customer’s perceived unfairness will manifest in a certain behaviour, such as defecting to a competitor or contacting management. How a customer will behave, depends on many factors. Seiders and Berry (1998) suggest to take two endpoints on a continuum that show how a customer could react; retaliatory responses (emotional) on one side and deterrence responses (cognitive) on the other.

2.1.3 Service recovery

(13)

In a best case scenario, a good recovery leads to even higher satisfaction with the company than before the incident, which is called the recovery paradox. This phenomenon has received ample attention in literature (e.g.: Hart, Heskett and Sasser, 1990; Smith and Bolton, 1999; McCollough, Berry and Yadav, 2000), although it remains hard to draw up generalizable situations in which the paradox exists. The consensus appears to be that the recovery paradox effect on satisfaction does exist, albeit with a few notes:

• The effect is only obtained at the very highest levels of customers’ recovery ratings, above 5 on a 7-point Likert scale (Smith and Bolton, 1998).

• Paradoxical increases in satisfaction diminish after more than one failure (Maxham and Netemeyer, 2002).

• The effect is most likely to occur in cases where the customer holds strong emotional ties with the company (Ringberg et al, 2007).

Recovery cannot be used as the only weapon when dealing with failures, nor can companies expect to be able to ‘plan’ every recovery to be perfect (as every service failure case is typically heterogeneous in nature). Additionally, satisfactory recoveries might be harder to create, as customer expectations regarding treatment in failure contexts are higher than they are in standard service encounters (Tax, Brown and Chandrashekaran, 1998). Even if, in spite of this, a satisfactory recovery is achieved, this will most likely result in higher subsequent expectations of the customer, which highlights a potential pitfall in recovering well (Maxham and Netemeyer, 2002). In the end, error-free service delivery is the better option, given the volatile conditions under which the recovery paradox exists and the fact that customers will have more confidence in a company that reliably delivers service without failures (McCollough et al, 2000).

(14)

Ringberg et al (2007) have come up with a new approach based on customer cultural models, which sums up well the different kind of reactions a negative encounter evokes within a customer and what the best-practice solution that the company should use is.

Cultural Model Customer Reaction Best Practice Company Reaction

Relational Emotional Sincere apology, respect and explanation: only here is a recovery paradox likely

Oppositional Aggressive, angry Provide recovery options for customer to choose from (additionally; Mattila (2009) finds that especially women prefer getting recovery options)

Utilitarian Sunk cost, inconvenienced

Acknowledge problem, compensate for time and energy and offer refund

Table 2.1: Adopted from Ringberg et al, 2007. Each cultural model comprises of customers that have different reactions in dealing with failures. Each of these segments require a different approach.

2.2 Brand equity and brand connection

Brand equity has emerged as an important topic for researchers and indeed as a central business concept over the last few decades (Leone, Rao, Keller, Luo, Mcalister and Srivasta, 2006). In this time period, many models of the concept were drawn up which over time resulted in two main models, that of Aaker and that of Keller. Where Aaker’s model focuses on brand assets and knowledge, Keller’s model also includes the emotional side to brand equity. Keller called his concept Customer-Based Brand Equity. CBBE is defined as the differential effect that brand knowledge has on consumer response to marketing of that brand (Keller, 2008).

(15)

High customer-based brand equity can have many positive effects, such as increased customer loyalty (Roehm and Brady, 2007, Keller, 2008), repatronage intentions of customers and the ability for the firm to charge a premium price (Roehm and Brady, 2007). However, it is important to make a critical note here that all these favourable effects are assumed to occur under acceptable brand performance (Roehm and Brady, 2007). It is of course not realistic to assume that this static situation always exists. In service failure settings, brand equity can offset negative effects of unacceptable brand performance, among other options like acting quickly, offering explanations, providing fair treatment, complaint procedures and fair compensation (Vázguez-Casielles, Belén del Río-Lanza and Díaz-Martín, 2007).

There are few studies that have examined brand equity within the context of service failures. The most noticeable studies are those performed by Roehm and Brady (2007) and Brady et al (2008). In their first study (2007), Roehm and Brady find that brand equity might provide mixed blessings. Their results hinge on two moderating factors, distraction and severity. High equity brands have the advantage only when a customer evaluation was made immediately after learning of a severe failure or when there was a distraction. In the case of a distraction, customers were too busy coping with the failure. With resources devoted to coping issues, the high equity brand was found to be temporarily safe from negative revaluation. In a related second study, Brady et al (2008) find that high brand equity leads to more favourable satisfaction evaluations and behavioural intentions. Most importantly, this advantage spans the entire failure and recovery sequence, which means that high brand equity provides an advantage in pre-failure, failure and post-failure circumstances, giving it an enduring, long-term characteristic. With regard to timing, high equity brands show a steeper decline in ratings than low equity brands, which can be explained by higher expectations for the brand with high CBBE.

Brand connection

(16)

The self-brand connection is defined as the degree to which consumers have incorporated a brand into their self-concept (Fournier, 1998; Escalas and Bettman, 2003). The self-concept contains one’s identity, values and goals (Fournier, 1998), and has two dimensions. The first dimension refers to an individual level and the second to a group level. People have two general motivations to form connections with brands and their equity; to enhance their own self-esteem and to verify and be consistent with the self (Escalas and Bettman, 2003).

The concept of self-brand connection is that how someone sees one’s self, is co-defined by the set of brand associations that belong to a certain brand. When these brand associations are used to construct the self or to communicate the self-concept to others, a strong connection is formed between the customer and the brand. As people ideally want to create and communicate a positive self-image, they can show a bias towards accepting and including information that is negative or inconsistent. The explanation for is that people are more likely to attribute positive outcomes to aspects of self and negative outcomes to circumstances unrelated to self (Miller and Ross, 1975).

Self-brand connection is a powerful concept, as people will form strong relationships with those brands that have values and personality associations that are congruent with their self-concept (Swaminathan, Page and Gürhan-Canli, 2007). They will base their decisions on which brand to buy on this congruence, and have been shown to influence members of their own reference groups with their brand choices (Escalas and Bettman, 2003). As the connection between a person and a brand is based on the associations this person holds, brand connection is an important element of brand equity (Keller, 1993). Additional benefits of a strong self-concept connection include a more durable relationship and the tendency for people to want to discount and counter-argue negative information (Swaminathan et al, 2007). How strongly someone will tend to discount negative information, depends on the commitment of the customer (Ahluwalia, Burnkrant and Unnava, 2000). Ahluwalia et al (2000) show that highly committed customers showed extensive counter argumentation.

(17)

3| Hypotheses and conceptual model

The literature review showed that high customer-based brand equity could help companies during service failures. The question that must now be turned to is why this is. This will be done by investigating if brand connection directly influences a customer’s evaluation of attribution, severity and recovery. This chapter draws up the hypotheses, which are based on theories in the field of marketing and (social) psychology, and concludes with the conceptual model in which the hypotheses are visually presented.

3.1 The direct effect of brand connection on the customer’s perception of service failure A direct, positive effect of brand connection on how a customer perceives a failure would mean that a customer who holds a company in high esteem, attributes failures less to the brand, rates failures less severe and is more satisfied with recovery efforts. However, before being able to conclude whether the main effect of brand connection on severity, attribution and recovery is positive or negative, two views need to be looked at, that provide contrasting expectations and theories.

The first view would suggest a negative effect. The rationalisation for this is as follows. Firstly, expectancy-disconfirmation theory (Oliver and Burke, 1999) would postulate that having higher expectations leads to higher disappointments when service fails (the disconfirmation). Customer’s expectations are based on many sources, such as prior experience, vicarious experience, and interpersonal and commercial communications (Oliver and Burke 1999). The link that can be established here is that someone with high customer-based brand equity will have higher expectations, resulting from positive prior experience and exposure to commercial and personal communications. This could mean that customers attribute the failure to the company in a stable manner, rate the incident more severely and rate recovery attempts lower. Further explanation of this lies in forward assimilation theory (Oliver and Burke, 1999). This theory holds that consumers will interpret outcome information to be consistent with their expectations. In effect, expectations become consistent with satisfaction, meaning that when expectations are not met, the customer will not be satisfied. Therefore, the effect of brand equity should mostly be negative when a failure occurs.

(18)

This positive effect is based on motivated reasoning, in which people are more likely to arrive at the conclusion that they want to arrive at (Kunda, 1990), as this saves cognitive resources (Fennis and Stroebe, 2010). This last point is corroborated by the second theory; the anchoring and adjustment model. According to this model, prior positive associations (which effectively serves as high brand equity) provide favourable data points upon which current beliefs are based. When assuming that there are numerous favourable data points, which is a likely condition in order to have developed high brand equity, a few new negative observations might not combine immediately to form a negative attitude, creating a ‘buffering’ effect of brand equity.

Having discussed the possibility of both a positive and negative effect of brand connection, which view should be adopted? It is believed that a positive effect is more likely, for the following reasons. Firstly, attribution theory (Folkes, 1988) suggests that people might see service failures as an unstable occurrence, and thus will not start attributing it to the company immediately. This expectation is laid down in hypothesis h1a. Secondly, it is more likely for a customer with a strong brand connection to want to defend a brand’s position than to immediately turn to attack it, as this would undermine the enduring foundation of brand equity. When a customer with high CBBE perceives a service failure, the experience is loaded with negative information which is inconsistent with the strong, favourable and unique associations that are in memory. This preexisting cognitive structure will guide the interpretation and integration of any new information and will lead to attitude-consistent processing (Petty and Cacioppo, 1986), and therefore customers would be more likely to positively look at the magnitude of failures and the recovery actions undertaken by the company. Additionally, the negative information of a service failure is inconsistent with the positive anchor established in the mind of the customer. Together, the negative and conflicting information a service failure brings with it will lead to customers wanting to defend their current beliefs. Consistent with defensive processing, a customer will give a failure less weight and perceive it as less diagnostic than the brand equity that has already been established. When a customer thus evaluates the severity and also the recovery of a service failure, he or she will be inclined to base the evaluation on their previously established brand equity, leading to hypotheses h1b and h1c.

H1a Brand connection has an effect on the attribution of blame; a stronger brand connection leads to a lower probability of customers attributing blame.

H1b Brand connection has a positive effect on service failure severity ratings; a stronger brand connection leads to more positive evaluations of service failure severity.

(19)

3.2 The direct effect of brand connection in case of multiple service failures

Finding out whether a positive direct effect exists would be new and relevant information in the service failure literature. But without knowing the effects over time and multiple failures, a large insight would be missed. By researching the effects of multiple failures, an important gap in existing literature is also addressed. More service failures will not make the experience of the customer better, and the question is what will happen over a stretch of time in which multiple service failures occur. Even though brand equity has the potential to provide an enduring advantage and protection from a failure, it will not be able persist forever. With each additional failure, there are a number of theories at play that explain why a customer can only take so much from a company and why brand equity cannot protect a brand forever.

When looking at the effects of multiple failures on attribution, attribution theory (Taylor, 1994) outlines that the extent to which a customer will attribute blame in a stable manner to the company, depends on the controllability and stability of a failure. The first failure will most likely be seen as an unstable occurrence and not be stably attributed to the company. With more failures, the customer will start connecting the failures and see them as one big, cumulative failure. In the case of multiple service failures, the customer will think of the company as having high controllability (the company apparently is unable to prevent failures) and therefore attribute more extensive blame. It can thus be hypothesised that after the first failure, people will start to more extensively attribute blame to the company, leading to a direct effect of brand connection on attribution that should get weaker with each following service failure. This expectation can be found in hypotheses h2a.

(20)

However, with more service failures come more negative data points, which will pull the attitude of the customer in the opposite direction. This will happen more quickly than positive information, as negative performances by companies have a greater influence on the customer than positive ones (Mittal, Ross and Baldasare, 1998). This way, the negative information will become the most attitude-consistent way to process, which will lead to a customer who rates service failures more severely and rates recovery actions by the company as less positive. Therefore, the direct effect should be strong after the first failure has occurred, but will get weaker for people who experienced two failures and for people who experienced three failures. When a customer evaluates the severity of a failure, it is likely that, as the negative information contained within each additional failure increases, the customer will have an increasingly negative outlook. The negative information is weighed more heavily, and the positive anchor established in memory is being shifted with each negative experience. Therefore, the effect of brand equity on severity should get weaker with each additional failure after the first failure. This hypothesis is formulated in h2b.

Remaining are the expectations with regard to the customer’s evaluation of a firm’s recovery. When more failures occur, the customer with high CBBE will have a less positive anchor. In addition, the negativity effect that will occur as the negative information of multiple failures is weighed more heavily, might make it hard for a customer to rate a recovery positively with each additional failure. Therefore, when a customer rates the recovery effort of the company, there should also be a decrease in the direct effect with each additional service failure, as is hypothesised in h2c.

H2a The effect of brand connection on attribution will get weaker with each failure after the first service failure.

H2b The positive effect of brand connection on severity will get weaker with each failure after the first service failure.

(21)

3.3 Conceptual model

In this section, the conceptual model is presented that follows from the previous section in which the hypotheses were formulated. As can be seen from the figure, each service failure is looked at independently to be able to identify the hypothesised direct effect of brand connection, which is the psychological manifestation of brand equity.

Figure 3.1: The conceptual model of this research.

(22)

4| Methodology

This chapter will offer a description of the data set used, and will outline and validate the scales and measures used for the data analysis. Descriptive statistics are also presented, and the chapter concludes with the plan of analysis, in which the statistical techniques used will be explained.

4.1 The research project and research design

The data sample that is used in this thesis consists of data on customers of a Dutch public transport provider. Topics brought up in the survey pertained to service failures and subsequent recoveries (up to three), satisfaction, brand equity, switching costs, commitment, loyalty and share of wallet. The research project was carried out in 2007 and 2008 by means of an online survey. One measurement was conducted in each year to be able to gain an insight in any changes in customer ratings. Getting an accurate insight in changes is one of the main advantages of a longitudinal study (Malhotra, 2010). The first survey was sent out to approximately 10.000 customers, and yielded a response rate of 59.9 % (5994 customers), whereas the second survey had a response rate of 32.2 % (3221 customers). The final data set used contains only participants that have participated in both rounds. The research design used can be classified as a between-participants design, in which three groups are created. The first group contains people with one service failure experience, the second group contains people who experienced two failures and the third group consists of people who have had three services failures. The groups can be best described as three static groups, as measurements on the groups are made after the treatment and customers are not assigned at random (but on the basis of the number of service failures experienced) (Malhotra, 2010).

4.2 Scales and measures

(23)

Multi-item scales used to measure customer satisfaction and brand connection

Satisfaction is measured using a multi-item scale, which is capable of capturing more information than can be provided by a single-item measure (Churchill 1979; Bergkvist and Rossiter, 2007). The scale in this thesis is similar to the one used by Smith and Bolton (1998), which consisted of an index of four questions that asked for the level of satisfaction, happiness, and whether customers were pleased and convinced the company did a good job. The scale used in this thesis looks at whether a customer experiences positive feelings, thinks the company is a good choice and feels at home with the company. The second multi-item scale in thesis is used to measure brand connection. The measurement is based on the brand connection scale developed by Escalas (1996) and was utilised in their subsequent research (Escalas and Bettman; 2003). Escalas’ scale allows to measure the extent to which a person has made a connection with a brand, and whether the brand is being used for the creation of the self-image and the communication of this image to others.

Validity customer satisfaction scale and brand connection scale

The two multi-item scales used are tested for validity by means of a Cronbach alpha’s test for internal consistency and by means of factor analysis to see whether both the customer satisfaction scale and the brand connection scale load on separate factors.

• Cronbach’s alpha

The test for internal consistency of the satisfaction scale showed a Cronbach’s alpha which was above 0.6 (∝ = 0.76), indicating satisfactory internal consistency reliability (Malhotra, 2010). The same procedure was performed for the brand connection scale. After internal consistency tests with Cronbach’s alpha, it was decided to remove one of the questions originally part of the scale (I feel company X helps me to be who I want to be) as this yielded an increase of internal consistency, going from ∝ = 0.76 to ∝ = 0.87. The scales can be found in table 4.1.

• Factor analysis

(24)

The rotated component matrix shows that all brand connection scale items indeed strongly load on a single dimension. Two of the four questions for customer satisfaction strongly load on a second dimension, with the other two loading less strongly on this same dimension. Overall there is no strong discrepancy and it can therefore be concluded, on the basis of both the factor analysis and the Cronbach alpha’s test, that both scales are suitable for use in this analysis. In the table below, both multi-item scales and single-item questions are summarised.

Construct _∝ Questions Measurement Source

Satisfaction 0.76 • Overall satisfaction with the service • The company is a smart choice • I feel at home with the company • The company invokes positive feelings

All items: 10-point Likert scale

Based on Smith and Bolton, 1998 ; Aaker, Fournier and Brasel, 2004

Brand Connection

0.87 • I can identify with company X • I use brand X to tell others who I am • I feel a connection with company X • Company X fits within my lifestyle

All items: 10-point Likert scale

Escalas and Bettman, 2003

Attribution - Was company X responsible for the service failure?

Dichotomous: yes / no

Maxham and Netemeyer, 2002 Severity - How severe do you rate the service

failure?

10-point Likert scale

Recovery - How well did company X help you in rectifying the service failure?

10-point Likert scale

Smith, Bolton and Wagner, 1999

Table 4.1: Overview of scales and measures used.

4.3 Descriptive statistics

(25)

Variable Group 1 (1 SF)Group 1 (1 SF) Group 2 (2 SF)Group 2 (2 SF) Group 3 (3 SF)Group 3 (3 SF)

Mean SD Mean SD Mean SD

Satisfaction T-1 5.95 1.49 5.58 1.58 5.35 1.64 Satisfaction T 6.24 1.49 5.79 1.58 5.45 1.66 Brand Connection T-1 4.68 1.94 4.34 1.96 3.99 1.95 Severity 7.33 2.06 7.39 2.27 7.55 2.40 Recovery 2.92 2.11 2.45 1.84 2.35 1.91 Switch costs 5.05 1.57 5.18 1.61 5.20 1.58

Variable Group 1 ( 1 SF)Group 1 ( 1 SF) Group 2 ( 2 SF)Group 2 ( 2 SF) Group 3 ( 3 SF)Group 3 ( 3 SF)

Yes No Yes No Yes No

Attribution 85% (1129) 15% (195) 90.7% (528) 9.3% (54) 91.7% (317) 8.3% (28)

Table 4.3: Frequencies for attribution (N is given between brackets). * SF = Service Failure(s)

From table 4.2, it can be deducted that the mean scores for both satisfaction (at T and T-1) and brand connection (at T-1) show a decrease over the groups as customers experience more service failures. Scores for satisfaction at both measurements are similar, although ratings for satisfaction at T are slightly higher. When looking at the mean scores given on severity and recovery, and the frequencies for attribution (table 4.3), a few observations can already made. Regardless of whether people have experienced one, two or three failures, the attribution of blame is quite high across all static groups. Severity also has high mean scores for all failures. Interestingly, the mean scores for recovery are not very high. Of some interest is the fact that the mean score of the evaluations of severity and the number of customers who attribute blame seem to increase from group one to group two to group three, even though it cannot be said whether the difference is statistically significant. Likewise, the mean scores of the evaluations of how well a company has recovered show a decline over the three groups, as more service failures are experienced by the customers in successive groups.

(26)

4.4 Plan of analysis

To test for the effects of brand connection on attribution, severity and recovery, two techniques will be used: seemingly unrelated regression (SUR) and logistic regression. The reason for using logistic regression, in addition to SUR, is that attribution is a binary variable, and can therefore not be included in the SUR as dependent variable. The statistical methods and their equations are outlined in the section below.

Logistic regression: Attribution

To analyse the effects of brand connection on attribution, three logistic regression equations will be estimated, for each of the three static groups containing people with one, two or three failures. Each equation will have a different sample size, as not everyone has experienced two or three failures. The first equation will have the largest sample size (N = 1276) as most people have experienced one failure, and the third equation will have the smallest sample size (N = 339). Sample size for the second equation is 568. This should not pose any problems, as the smallest sample size, for the third equation, still contains more than three hundred people. Attribution is the dependent variable, and brand connection is the independent variable. The brand connection scale is taken at the first measurement (T-1), as this is the moment before any failures occurred. In the regressions, switching costs are controlled for. Switch costs have been shown to affect behavioural intentions (Jones, Reynolds, Mothersbaugh and Beatty, 2007), and Gustaffson, Johnson and Roos (2005) suggest that higher switching costs might have a positive effect on intentions and behaviour. Therefore, it could be of influence on the dependent variables in this research. Switching costs are measured based on the calculative commitment scale (Verhoef, Franses and Hoekstra, 2002). The three equations can be found below:

Logistic regression static group 1 (1 service failure) (1)

Logistic regression static group 2 (2 service failures) (1)

Logistic regression static group 3 (3 service failures) (1)

(27)

SUR: Severity, Recovery and Customer Satisfaction T

Severity and recovery will be tested by using the statistical technique of seemingly unrelated regression (SUR) (Zellner, 1962). SUR is different from normal linear regression, in that it can accommodate multiple regression equations in one model. Each equation has its own dependent and independent variables, that can be different for each equation in the model. Additionally, SUR allows variables to be both dependent and independent (Beasley, 2008) and is therefore, in combination with the potential for multiple equations, a flexible analytic tool. By using SUR in this thesis, the results are more efficient as information on different equations is combined (Revankar, 1974) and, additionally, the correlation of variables and error terms are accounted for.

Each of the three SUR models will have a different sample size due to the fact that not all customers have experienced the same number of failures. The first and second models have sample sizes of 1215 and 547 respectively. The smallest sample size (for the third static group) is also still sufficiently large (N = 321). For each static group, there are three equations. The first equation regresses brand connection and switch costs on severity. The second is of brand connection and switch costs on recovery (switching costs are controlled for in the equations of severity and recovery). Lastly, an equation is estimated for each group that tests for the effects of customer satisfaction at T-1, brand connection at T-1, attribution, severity and recovery on customer satisfaction at T. The equations are represented visually below:

(SUR) Static group 1 (1 service failure) (1)

(2) (3)

(SUR) Static group 2 (2 service failures) (1)

(2) (3)

(SUR) Static group 3 (3 service failures) (1)

(28)

5| Results

In this chapter, the results of the regression analyses will be discussed. The results of the logistic regressions for attribution will be discussed first. Afterwards, the outcomes of the seemingly unrelated regressions for severity, recovery and customer satisfaction are presented.

5.1 Results logistic regression

Three logistic regressions were performed, one for each of the three static groups. Significant coefficients are flagged by a star (*). As table 5.1 shows, for the first and second static group, a significant model was established (p = 0.047 and 0.001 respectively), meaning that there is predictive capacity in the equation. Brand connection was significant for both the first group that experienced one service failure (β = -0.099, p = 0.016) and the second group that contains people who experienced two failures (β= -0.209, p = 0.005). Therefore, there is indeed a direct effect of brand connection on attribution, consistent with hypothesis h1a. As the beta coefficient is negative, it means that higher levels of brand connection lead to a decrease in the probability that someone will attribute blame to the company.

Dependent Variable

Regression model significant (Omnibus Tests

of Model Coefficients)?

Independent variables

Coefficient Sig Constant

Attribution static group 1, 1 SF

Yes, P = 0.047 Brand Connection, t-1 Switch costs -0,099* -0,028 0,016 0.579 2.383 Attribution static group 2, 2 SF

Yes, P = 0.001 Brand Connection, t-1 Switch costs -0,209* 0.213* 0,005 0.028 2.198 Attribution static group 3, 3 SF

No, P = 0.096 Brand Connection, t-1 Switch costs -0,130 0,224 0,197 0,094 1.862

Interestingly, the coefficient is more negative for people who experienced two failures. This finding is inconsistent with the expectations of hypothesis h2a, which expected a direct effect of brand connection on attribution that would weaken for people who experienced multiple failures. It is however important to note that it cannot be said that the effect is thus stronger, as the two coefficients were not tested for significant differences between them. This is due to the fact that the coefficients are in two separate models. For people who experienced three failures (static group 3), the direct effect of brand connection disappears, as the third regression did not yield a significant model (p = 0.480). In sum, hypothesis h1a is accepted. Hypothesis h2a cannot be accepted as the coefficient of brand connection gets stronger instead of weaker for the second group. The effect disappears after the third failure.

(29)

5.2 Results seemingly unrelated regression

The results of the three SUR analyses are found in tables 5.2 through 5.4. The outcomes will be discussed per element (severity, recovery and customer satisfaction at T, respectively), as this facilitates a more clear overview of the discussion with regard to the results over the multiple static groups and failures. Significant coefficients are flagged by a star (*).

Dependent Variable

Regression model significant (ANOVA)?

R-sq Independent variables Coefficient Sig Constant

Severity static group 1, 1 SF

Yes, P = 0.006 0.0082 Brand Connection, t-1 Switch Costs -0.094* 0.012 0.002 0.744 7.734 Recovery static group 1, 1 SF

Yes, P = 0.000 0.0298 Brand Connection, t-1 Switch Costs 0.185* 0.039 0.000 0.311 1.857 Customer Satisfaction T, static group 1

Yes, P = 0.000 0,6179 Customer Satisfaction, t-1 Brand Connection, t-1 Attribution SF 1 Severity SF 1 Recovery SF 1 0.657* 0.109* 0.047 -0.004 0.051* 0.000 0.000 0.539 0.722 0.000 1.567

Table 5.2: Results of the SUR for the first static group, containing people who experienced one service failure Dependent

Variable

No, P = 0.496 0.0025 Brand Connection, t-1 Switch Costs -0.051 -0.039 0.306 0.533 7.875 Recovery static group 2, 2 SF

Yes, P = 0.010 0.0246 Brand Connection, t-1 Switch Costs 0.133* 0.079 0.001 0.112 1.433 Customer Satisfaction T, static group 2

Yes, P = 0.000 0.6418 Customer Satisfaction, t-1 Brand Connection, t-1 Attribution SF 2 Severity SF 2 Recovery SF 2 0.656* 0.129* 0.014 -0.015 0.050* 0.000 0.000 0.916 0.378 0.028 1.458

Table 5.3: Results of the SUR for the second static group, containing people who experienced two service failures Dependent

Variable

No, P = 0.848 0.0010 Brand Connection, t-1 Switch Costs -0.038 -0.012 0.572 0.892 7.840 Recovery static group 3, 3 SF

No, P = 0.071 0.0162 Brand Connection, t-1 Switch Costs 0.122* 0.017 0.022 0.803 1.739 Customer Satisfaction T, static group 3

(30)

Severity

When looking at the results of the SUR in table 5.2 for the group of people who experienced one failure, a significant regression model was estimated (p = 0.006). The negative significant coefficient of brand connection ( β = -0.095, p = 0.002) is in line with hypothesis h1b, which states that there should be a positive direct effect of brand connection leading to lower severity ratings by high-equity customers. The negative coefficient indeed signals that the higher brand connection, the lower the severity rating of a customer, supporting hypothesis h1b. Consistent with Maxham and Netemeyer (2002), the second service failure is rated more severely by customers than the first (mean SF 1 = 7.39, mean SF 2 = 7.33). This research further demonstrates that the third failure is also rated more severely by people who experienced three failures than people who experienced only two failures (mean SF 3 = 7.55).

According to hypothesis h2b, the direct effect should get weaker for people that experienced two or three service failures. The analysis for the second and third static group support this prediction. As can be deducted from table 5.3, no significant regression model with significant coefficients of brand connection can be established for customers who experienced two failures (p = 0.496). Therefore, h2b is accepted. For people with three failures, table 5.4 shows that there is also no significant effect ( p = 0.848).

Recovery

Hypothesis h1c states that there should be a positive direct effect of brand connection on recovery, meaning that people with higher brand equity should give higher recovery ratings. Both for the first and the second static group, a significant regression model is estimated (p = 0.000 and p = 0.001 respectively). Table 5.2 shows that for the group of people who experienced one failure, there is a significant, positive effect (β = 0.185, p = 0.000) of brand connection on the ratings of recovery actions. Customers with higher levels of brand connection thus view recovery actions by the company in a more favourable light, and hypothesis h1c is accepted.

(31)

Customer Satisfaction T

For all three static groups, the equation for customer satisfaction was significant (p = 0.000 for group one, group two and group three). Additionally, the majority of the variance in customer satisfaction at T is explained in each group, ranging from 61% for group one to 67% for group three. The most interesting result is that brand connection and satisfaction at T-1 are positively significant throughout all groups, which means that the more satisfied someone was in the past and the stronger the connection he or she has formed with the company, the higher that person’s satisfaction after multiple service failures. When looking at the change of these coefficients, it can be seen that they remain relatively stable. The coefficient for customer satisfaction T-1 changes from 0.657 to 0.656 to 0.697. Brand connection’s coefficient changes from 0.109 to 0.129 to 0.111. Apparently, the higher someone’s pre-failure satisfaction and brand connection, the higher someones satisfaction three failures. This effect remains present after multiple failures and is relatively stable.

Of the other factors, only recovery is significant in explaining customer satisfaction for groups one and two (recovery of the third failure, in addition to attribution and severity for all three failures, are not significant in explaining customer satisfaction at T).

Concluding remarks

In sum, four out of six hypotheses have been accepted, as can be seen in table 5.5. For all of the elements that make up the subjective experience of a service failure by the customer, it was established that there is indeed a significant influence of brand connection. All of the effects were in the hypothesised direction, except attribution. The direct effect of brand connection on attribution of blame showed the opposite effect over time, and its coefficient got more negative for the customers who experienced two failures as opposed to one failure. The next chapter will provide a discussion based on the results, and will offer potential explanations for the effects over multiple failures that have been found here.

Hypotheses Supported

(32)

6| Discussion

The main finding of this thesis is the support found for the hypothesised direct effect of brand connection on the customer’s perception of a service failure, as measured by attribution, severity and recovery. This finding provides additional information in that it proves that brand equity protects a brand from the negative consequences of a service failure, via less negative evaluations of attribution, severity and recovery. Where other studies (Roehm and Brady, 2007; Brady et al, 2008) find that a person with high brand equity shows more favourable satisfaction ratings and behavioural intentions, the findings of this thesis show that high brand equity leads to a less negative subjective experience of a service failure. The results show continuing support for the notion that the building of brand equity is a valuable strategy in counteracting the negative effects of service failures.

Another important finding of this thesis is the insight provided in how the customer’s subjective experience of a failure and the effect of brand equity change, as a result of multiple failures. However, a remark that has to be made is that, in spite of many of the regression coefficients being statistically significant, the explanatory power of the models was quite low (as indicated by the R(-squared) statistic). A possible explanation for the low R(-squared) values might be that there are other independent variables that explain the ratings for attribution, severity and recovery, not included in the analyses. This is not unexpected, as it is likely that the customer will experience and invoke a wide variety of thoughts and emotions that contribute to how he or she will evaluates a service failure. Given the complex nature of consumer behaviour, it is unlikely that a consumer would only invoke his or her connection towards a brand in forming evaluations of service failures. Having addressed this, there are a number of observations that can be made in light of the results, and a separate discussion of attribution, severity and recovery will follow below. This discussion will also compare the findings to existing literature.

Attribution

(33)

In addition, attribution theory would suggest that in case of multiple failures, a stable pattern of failures will be seen by the customer which will lead to the customer increasingly attributing blame to the company (Folkes, 1988; Taylor, 1994). How can the contradicting finding of this thesis be explained? A possible explanation follows from a note that Maxham and Netemeyer (2002) make. They stipulate that the increasing attributions of a customer, in case of multiple failures, depend on the degree to which a customer thinks failures represent a pattern that is stable to the firm. It could be that for customers who experienced two failures in this research, the second failure’s cause could not be attributable to the company, thereby explaining why customers did not more extensively blame the company for the second failure as opposed to the customers who experienced one failure.

Another possible explanation for the opposite effect lies in the negative information that is contained within a service failure which conflicts with the positive anchor in the mind of the customer with high CBBE. It could be that the first failure is seen as an unstable occurrence, as is consistent with attribution theory (Taylor, 1994). People will therefore not think too much about it. The second failure, however, is forcing the customer to think about the failures in more detail, leading to attitude-inconsistent processing in customers who have high CBBE. In this thought process, the customer is drawn towards the positive anchor that is established in memory thanks to previous experiences, and the new evaluation is pulled in the direction of the existing evaluations. The result is that the customer will want to resist blaming the company. This expectation is consistent with the bias people show towards positive information (Hogarth and Einhorn, 1992). The observation that the effect is no longer present for the customers who had three failures shows that for these customers, something changes that makes three failures the limit, after which brand equity no longer provides a positive anchor and the effect of brand equity is no longer statistically significant.

Severity

(34)

For the people with two service failures, the two failures that both have high severity outweigh any established brand equity in the mind of the customer. This would also be consistent with prospect theory, which states that losses are weighed more heavily than gains (Kahneman and Tversky, 1979). Maxham and Netemeyer (2002) similarly suggest that customers consider ‘failure history’ in case of multiple failures, which means that they will not look at the individual failure, but take together all the failures in evaluating severity. Taking these considerations into account, it is suggested that the results found in this research could be due to the high severity of the incidents, combined with the customer paying more attention to the combined failure history and negative nature of the failures. In sum, the results show that brand equity’s effect on how a customer evaluates the severity of a failure is strongly influenced by multiple failures with high severity, and add to the notion in academic literature that failure magnitude is a very important concept, which can negatively affect satisfaction (Hoffman, Kelley and Rotalsky 1995; Smith et al, 1999).

Recovery

(35)

Concluding remarks with regard to existing literature

On a macro-level, how do the results of this thesis relate to the findings in literature with regard to the effects of brand equity on service failures? It is believed that the results found for severity and recovery do not provide any major contradictions. Firstly, Roehm and Brady (2007) conclude that brand equity provides a buffering mechanism against the negative effects of a failure, that is of finite nature. This thesis supports this and provides additional insight into the range of consequences that may be attributed to brand equity, by demonstrating that a direct effect of attribution, severity and recovery on brand equity exists. Given that the direct effect was not found to be present for all failures and declined in case of multiple failures for severity and recovery, the results are in line with Roehm and Brady’s (2007) notion that brand equity’s effect will show a diminishing tendency, especially for high CBBE customers, and eventually disappear. The only exception to this is attribution. As discussed above, the findings with regard to attribution go against what was hypothesised. It also contradicts the finding in literature that, generally, attributions of blame increase with additional service failures (e.g.: Taylor, 1994; Maxham and Netemeyer, 2002). Secondly, the results for severity and recovery also have important similarities to those found by Maxham and Netemeyer (2002). They find that in case of multiple failures, customers discount the effects of one failure when the firm has typically provided satisfactory performance. The second failure, however, leads to a strong intolerance on the part of the customer for bad recoveries and increased severity ratings. This increasingly less favourable perception of a service failure by the customer fits the results found here well.

(36)

7| Managerial implications

Performance failures are inevitable, maybe even more so in the service landscape. Therefore, it is important for brand managers and service managers to know how to prepare the company for failures and how to deal with the aftereffects. First of all, they can more specifically build brand equity by investing in building a strong connection with customers. They can do this by communicating those brand associations that appeal best to the customers the company is targeting. These brand associations are used by the customer in building the connection between the self and the brand. The power of this concept lies in that customers will base what products to buy on their connection with a brand, and will actively communicate this preference for the company to others. Additionally, self-brand connections will lead to more robust brand attitudes, that are difficult to change and imitate by competitors, and can result in a more forgiving customer in case of service failures. When a service failure does occur, brand and service managers should have the knowledge that customers who have a strong connection, will experience a failure less negative than customers with a weak connection.

Another implication is that firms should be aware that multiple service failures can threaten the relationship with a customer. In case of multiple service failures, the effect that brand equity has on how a customer evaluates severity and recovery, reduces and eventually is no longer significant. Even though the probability of evaluations of attribution will get smaller for people who experience a second failure, the positive effect that brand equity has, is no longer significant when three failures have occurred. What can be taken away from this is that, in case of multiple service failures, companies that enjoy high customer-based brand equity have an advantage over companies that have low CBBE. This advantage translates in a less negative perception of a service failure by the company, which will get weaker with additional service failures (with the exception of attribution, as discussed in chapter 6), and eventually disappears.

(37)

8| Limitations and future research possibilities

As with any academic study, there are some limitations that can be identified, and perhaps help to improve future research. Firstly, there are some limitations in the data set. All data comes from only one company, and the post measurements were not made directly after a service failure, which could cause some deviation in the ratings by customers, since some time has passed since the failure occurred. Future research could improve upon this by including multiple companies in studies with regard to the effects of brand equity in unsatisfactory performance settings. Additionally, attribution was measured on a nominal scale, limiting the potential for analysis. Future research would benefit by measuring attribution on an interval scale.

A second limitation is that, even though attribution, severity and recovery are considered to be an accurate description of the subjective experience of a failure, and an extensive literature review was conducted to come to these three elements, there may be other factors at play. An idea for future research could for example be to also include failure type in the analysis.

(38)

9| References

Aaker, J, Fournier, S. and Brasel, S.A. (2004), ‘When good brands do bad,’ Journal of Consumer Research, vol. 31, June 2004.

Ahluwalia, R, Burnkrant, R.E. And Unnava, H.R. (2000), ‘Consumer response to negative publicity: The moderating role of commitment,’ Journal of Marketing Research, vol. 37, 203-214

Ahluwalia, R. (2002), ‘How prevalent is the negativity effect in consumer environments?,’ Journal of Consumer Research, vol. 29, 270-279

Anon. 2011, 19 August 2011. Flight delays: you’re grounded. The Economist (Gulliver Blog). Available from:

http://www.economist.com/blogs/gulliver/2011/08/flight-delays [accessed 19 September 2011]

Beasley, Mark T. (2008), ‘Seemingly Unrelated Regression (SUR) Models as a solution to Path Analytic Models with Correlated Errors,’ Multiple Linear Regression Viewpoints, vol. 34 (I).

Bergkvist, L and Rossiter, J.R. (2007), ‘The predictive validity of multiple-item versus single-item measures of the same constructs,’ Journal of Marketing Research, Vol. XLIV, 175-184.

Bitner, M.J, Booms, B.H. and Tetrault, M.S. (1990), ‘The service encounter: Diagnosing favorable and unfavorable incidents,’ Journal of Marketing, 54, 71-84

Brady, M.K., Cronin Jr., J.J., Fox, G.L. and Roehm, M.L. (2008), ‘Strategies to offset performance failures: the role of brand equity,’ Journal of Retailing, 84 (2), 151-164

Dong, B., Evans, K.R. And Zou, S. (2008), ‘The effects of customer participation in co-created service recovery,’ Journal of the Academy of Marketing Science, 36, 123-137

Van Doorn, J. and Verhoef, P.C. (2008), Critical Incidents and the impact of satisfaction on customer share,‘ Journal of Marketing, 72, 123-142

Escalas, J.E. (1996), ‘Narrative processing: building connections between brands and the self.’ Unpublished doctorial dissertation, Duke University, Durham, NC.

Folkes, V.S. (1988), ‘Recent attribution research in consumer behavior: A review and new directions,’ Journal of Consumer Research, 14, 548-565.

Fournier, S. (1998), ‘Consumers and their brands: Developing relationship theory in consumer research,’ Journal of Consumer Research, 24, 343–373.

(39)

Harvey, J.H. and Weary, G. (1984), ‘Current issues in attribution theory and research,’ Annual Review of Psychology, vol. 35, 427-459

Hoffman, K., Douglas, S.W. and H.M. Rotalsky (1995), ‘Tracking service failures and employee recovery efforts,’ Journal of services marketing, 9 (2), 49-61

Hoyer, W.D. and MacInnis, D.J. (2008), Consumer Behaviour, South-Western, 4th revised edition.

Hui, Michael K., Mrugank V. Thakor, and Ravi Gill (1998), “The Effect of Delay Type and Service Stage on Consumers’ Re- actions to Waiting,” Journal of Consumer Research, 24 (4), 469–79.

Jones, M.A., Reynolds, K.E., Mothersbaugh, D.L. and Beatty, S.E. (2007), ‘The positive and negative effects of switching costs on relational outcomes,’ Journal of Service Research, vol. 9, page 335-355

Kahneman, D. and Tversky, D. (1979), ‘Prospect Theory: An Analysis of Decision Making Under Risk,’ Econometrica, 47, 263–91.

Keaveney, S.M. (1995), ‘Customer switching behavior in service industries: An exploratory study,’ Journal of Marketing, 59, pg. 71

Keller, K.L. (1993), ‘Conceptualizing, measuring, and managing customer-based brand equity,’ Journal of Marketing, 57, 1–22.

Keller, K.L. (2008), ‘Strategic Brand Management: Building, Measuring, and Managing Brand Equity,’ New Jersey - Pearson Education.

Kunda, Z. (1990), ‘The Case for Motivated Reasoning,’ Psychological Bulletin, vol. 108, 480-498.

Leone, R.P, Rao, V.R., Keller, K.L, Luo, A.M., McAlister, L and Srivastava, R. (2006), ‘Linking brand equity to customer equity,’ Journal of Service Research, 14, 125-138.

Lewis, R. and Spanier, G (1979), ‘Theorizing about the quality and stability of marriage,’ Contemporary Theories about the Family, Vol 2., New York: Free Press, 268-294

Mattila, A.S. (2004), ‘The impact of service failures on customer loyalty: The moderating role of affective commitment,’ International Journal of Service Industry Management, vol. 15, 134-149.

Mattila, A.S (2009), ‘Do women like options more than men? An examination in the context of service recovery, Journal of Services Marketing, 24, 499-508

An empirical study into why brand equity helps when service fails