This article was downloaded by: [Mount St Vincent University]On: 07 October 2014, At: 13:30Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Total Quality Management & BusinessExcellencePublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/ctqm20

The determinants of consumers’switching intentions after servicefailureWen-Bao Lin aa Graduate Institute of Technology Management , NationalKaohsiung Normal University , 802, No. 116, Heping 1st Road,Lingya District, Kaohsiung City , Taiwan, Republic of ChinaPublished online: 18 Jan 2012.

To cite this article: Wen-Bao Lin (2012) The determinants of consumers’ switching intentionsafter service failure, Total Quality Management & Business Excellence, 23:7-8, 837-854, DOI:10.1080/14783363.2011.637808

To link to this article: http://dx.doi.org/10.1080/14783363.2011.637808

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

The determinants of consumers’ switching intentions after servicefailure

Wen-Bao Lin∗

Graduate Institute of Technology Management, National Kaohsiung Normal University, 802, No.116, Heping 1st Road, Lingya District, Kaohsiung City, Taiwan, Republic of China

This study explores the impact of two service failures on the switching intentions ofcustomers under the interferences of their negative emotions and service assurancestrength when managing controllable and uncontrollable service failure types. Thisstudy discusses the impact of relational embeddedness and negative word-of-mouthmessages on the switching intentions of customers regarding relationship networkand information communication, by conducting a survey with convenience samplingof consumers who frequently participate in tourism activities through tourismwebsites and who have experienced service failures. The empirical results are asfollows: the switching intentions of consumers who are affected by negativeemotions become stronger when service failures are controllable factors that couldbe managed and prevented, and the switching intentions of consumers who areaffected by service assurance strength lessens when service failures are controllablefactors that could be managed and prevented. The switching intentions of consumershave a positive correlation with the level of relational embeddedness and thestrength of word of mouth in an online shopping environment. This study alsodiscusses extant related research or theories, and proposes suggestions for futureresearch.

Keywords: service failure; assurance strength; relational embeddedness; word ofmouth

Introduction

Entrepreneurs reduce their dependency on multichannels because of the distribution cost

when enterprises face significant competition and lower gross profits. However, suppliers

are directly and significantly affected in a direct marketing environment when a service

failure or negative word of mouth occurs. For example, Donaton (2003) indicated that

word of mouth is the major factor that determines the purchase decision of consumers,

whereas Bansal and Voyer (2000) indicated that relationship strength to the message recei-

ver is another crucial factor that affects the purchase decision of consumers. In a market

dominated by consumer products, service failure must be solved quickly and efficiently, or

loss of customers ensues, and legal disputes may arise. Therefore, identifying methods to

reduce the negative impact caused by service failure is imperative. In general, the influ-

ence of the negative impact caused by inappropriate service failure management is

higher than that caused by satisfactory word of mouth. For suppliers, it is preferable to

effectively address and solve the unsatisfactory emotions of clients than to be perfunctory.

The crucial issues addressed in previous literature associated with the attributes of service

ISSN 1478-3363 print/ISSN 1478-3371 online

# 2012 Taylor & Francis

http://dx.doi.org/10.1080/14783363.2011.637808

http://www.tandfonline.com

∗Email: e2667@ms23.hinet.net, t3285@nknucc.nknu.edu.tw

Total Quality Management

Vol. 23, No. 7, July 2012, 837–854

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

failure include the following: (1) classification and cause of service failure. For example,

Bitner, Booms, and Mohr (1994) attributed the service failure to employee or customer

factors, which were divided into 4 categories and 16 subcategories; also, according to

Smith, Bolton, and Wanger (1999), service failures could be divided into outcome and

process service failures; (2) cause of service failure, in particular, in service industries.

For example, Mueller, Palmer, Mack, and Mcmullan (2003) compared the consumers in

a hospitality business between the USA and Ireland because the two countries have differ-

ent eating cultures. Through interviews with 700 consumers, they found that the consu-

mers of these two countries reported different preferences regarding the measures for

service failure recovery. However, excessive compensation does not have a significant

influence on the consumers’ intention of a repeat purchase; (3) discussion on following

variables regarding service failure, including association of satisfaction and recognition

of fairness (Smith et al., 1999; Weun, Beatty, & Jones, 2004). Craighead, Karwan, and

Miller (2004) used customer loyalty, perceived quality, severity of service failure, and

service commitment to distinguish the various service failure contact types. They indi-

cated that if proper service recovery measures are used, service failure may not cause per-

manent impairment to these enterprises. Tax, Brown, and Chandrashekaran (1998) found

that fair distribution, fair process, and fair interaction have a positive and significant

impact on the secondary satisfaction, which is different from the attribute of service

failure in the past. In general, they can be classified into controllable and uncontrollable

factors. In non-physical environments, such as online stores, the consumption tendency

of consumers is highly subject to sufficiency/insufficiency of their knowledge of the pro-

ducts. Bearden, Hardesty, and Rose (2001) indicated that, although consumers may be

highly confident on their familiarity with product information, they may fail to manage

negative emotions, which may be the determinant in the switching intentions of consumers

for the subsequent influence of service failure. According to structural holes theory pro-

posed by Burt (1992), a key to successful business operation is to determine how to

increase a compact network structure. To demonstrate this theory, this research explores

the methods to enhance the interaction of both parties in an online environment and

explores whether retention of consumer consumption intention under service assurance

provided by suppliers is also worth discussing. Although several studies examine

service failure, previous researchers have seldom discussed a single scenario or incident

of service failure from a fact analysis perspective. In other words, this research focuses

on reviewing historical consumption experience under service failure scenarios. This

study conducts a large-scale questionnaire survey and analyses the result by using multi-

variate analysis and the non-linear fuzzy neural network model. This approach was rarely

adopted in previous studies and is one the features of this research. The rest of this paper is

organised as follows. The second section presents the literature review and hypothesis; the

third section introduces the methodology; the fourth section details the empirical analysis;

and lastly, the final section offers a discussion and suggestions for future studies.

Literature review

Westbrook and Oliver (1991) believed that customer emotion is a mental reaction to the

use of a product or to consumer experiences. Dube and Menon (2000) indicated that

emotion is a type of organisation and shape of the integration process, which includes

mental reaction and reflection on incidents that occurred in the past, and enables motive

and behaviour. According to Gardner (1985), consumer emotion is easily affected by inter-

personal relations and, therefore, causes a significant impact on retail stores or service

838 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

industries, especially for the communication quality between the clients and service staff.

According to Caruana (2002), customer satisfaction is an emotional reaction after pur-

chase, with the differences subject to the assessment of the service provider in the mind

of the customer. Oliver (1993) claimed that positive and negative emotions directly

affect satisfaction. As indicated in the empirical study by Dawson, Bloch, and Ridgway

(1990), the emotion of customers affects their satisfaction during the shopping experience.

In summary, based on reviews of the literature, customer satisfaction after service recov-

ery is affected for customers who are affected by negative emotions. According to the

research on the restaurant service environment by Derbaix and Pham (1991), the

service quality of restaurant business workers or the environment and facilities within res-

taurants causes a positive/negative emotional reaction in customers, and affects their sat-

isfaction assessment. Swinyard (1993) found that customers who experience positive

emotions when making a purchase tend to offer more positive comments. According to

Babin and Darden (1995), customer emotion is subject to the shopping environment.

Therefore, a negative emotion is experienced if the gap between expectation of the custo-

mer on service quality and the service provided by enterprises is too large. This gap also

affects their satisfaction with service recovery.

Several researchers have divided the attributes of service failure into the following: (1)

internal attribute and external attribute (Bitner, Booms, & Tetreault, 1990); (2) (a) locus –

service failure is attributed to business hardware, for example, layout, movement, and

service policy of the shopping environment or software, such as the ‘hard-working’ atti-

tude of service staff and their familiarity with service flow; (b) stability, which is the fre-

quency of service failure, for example, incidental or frequent service failures (Weiner,

1985; Folkes, Koletsky, & Graham, 1987); (c) controllability, in that the cause of

service failure is controllable, such as reinforcement of personnel training, and uncontrol-

lable, for example, hardware system setup (Folkes, 1984). The service failure attributes

can also be classified from different viewpoints. (1) From the supplier service perspective,

they can be divided into the following three categories: (a) failure caused by the service

delivery system or delivery process, such as service delay and service unavailability;

(b) the nature of responding to the special needs or requests of customers, such as if the

attendant did not follow the seating plan as requested by the customer or if the cook did

not prepare the food as requested by the customer; and (c) failures caused by the behaviour

of employees, such as the manner in which employees treat the customer, or if employees

respond to the needs of customers with an inappropriate attitude or manner (Bitner et al.,

1990). (2) From the perspective of the service received by consumers, service failure can

be segmented into core service failure (such as error charging, stockout) and service

encounter failure (such as negligence, inability to respond in time, and unprofessional

management (Keaveney, 1995) as the failure caused by the interaction between the

clients and service providers. (3) Determining whether failure is involved in basic contents

as criteria, in which service failures can be divided into the following: (a) outcome: the

proprietor did not provide the required fundamental service or core service to the custo-

mers, such as product deficiency or a product in short supply; (b) process: customers

feel uncomfortable about the bad attitude of service personnel and late product supply.

Regardless of the outcome or process, service failure is attributed to controllable

factors, which are commonly associated with the familiarity of staff with their own

business and the level of training that they have received, and are also associated with

the failure rate of operation procedures (Mohr & Bitner, 1995; Smith et al., 1999). Due

to the increasing trend of online shopping, scholars have also conducted related research

in this field. For example, Meuter, Ostrom, Roundtree, and Bitner (2000) studied

Total Quality Management 839

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

technology-driven self-service and proposed four major factors that cause customer dissa-

tisfaction, including technology, process, design, and failure caused by customers. Hollo-

way and Beatty (2003) used a critical incident technique to discuss the service failure of

online shopping, and summarised 24 frequent service failures in seven categories, includ-

ing dispatch, web page design, customer service, payment, safety, and mixture. Although

the classification of service failure varies by viewpoint, this research assumes that in the

online service environment, service failure can be divided into the following: (1) control-

lable, indicating the attitude and service provided by service personnel, which is called soft

service failure; and (2) uncontrollable, indicating occasional service failure, which is com-

monly associated with the system, service and external environment, or the entire industry

structure, such as competitions that enterprises are insufficiently able to manage and

improve, which is called hard service failure.

In general, when service failures are controllable factors that occur frequently or

factors that enterprises could manage, the customers’ expectation of these enterprises

may vary by their own perceptions. Therefore, customers will have expectations of enter-

prises, which are due to the causes of service failure, and which further affect their behav-

iour (Floyd & Voloudakis, 1999; Raaij & Pruyn, 1998). If the negative emotions of

customers are taken into consideration, the intention of customers to repurchase might

be significantly affected when a service failure is attributed to an internal cause and is con-

trollable, and when customers experience negative emotions while making a purchase,

which leads to higher switching intentions of customers. Conversely, when service failures

are uncontrollable and occur infrequently, or when these failures are attributed to the

external environment, the repeat purchase intentions of consumers will be reduced

when these customers experience negative emotions. However, if a detailed explanation

and appropriate compensation measures are provided, the negative reaction of customers

will be reduced and they will have lower switching intentions. In brief, the switching inten-

tions of consumers will be affected by the type of service failure and their negative

emotions. If service failures are controllable, attributable to an internal cause, or occur fre-

quently, then customers will have a lower intention to repurchase than if the service fail-

ures are uncontrollable, attributable to an external cause, or occur infrequently.

H1: The switching intentions of customers, which are affected by negative emotions, will behigher when service errors are controllable factors than when they are uncontrollable factors.

Service assurance considered one of the measures that help enhance customer loyalty

and influence the behaviour of consumers and the performance of business operation.

Service assurance also helps to reduce the confidence crisis of consumers that is caused

by service failure.

For example, Wirtz (1998) proposed a service assurance model to prove that the estab-

lishment of an assurance system can effectively achieve various operation quality and mar-

keting goals within enterprises and help to reduce the impact of risk perception on

consumer behaviour. Boshoff (2002) indicated that service assurance that is delivered

through TV commercials will help to reduce the potential anxiety and uncertainty of shop-

pers when they make a purchase. Hays and Hills (2006) further established a structure of

service assurance strength with the micro-level behavioural theory. They indicated that

service assurance strength has a positive impact on service quality, customer satisfaction,

and loyalty through the use of three intervening variables, including marketing communi-

cation, employee motivation and vision, and learning through service failure. Although

several previous studies have indicated that service assurance strength has a positive

impact on service quality and customer reaction, this study assumes that, when additional

840 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

service assurance is provided by enterprises, controllable service failure has a higher nega-

tive impact on the switching intentions of consumers than uncontrollable service failures,

due to the following reasons: (1) in a controllable service failure scenario, management

reinforcement and enhancement of education training within enterprises can help to

enhance the self-performance of employees to further reinforce and compensate for the

adverse effects caused by service failure on customers. Therefore, the switching intentions

of customers, which are due to service failure, may be reduced; (2) there may be significant

opportunities for enterprises to internally improve the ‘service commitment management’

and ‘management of internal marketing communications’.

H2: The switching intentions of customers, which are affected by service assurance strength,will be lower when service errors are controllable factors than when they are uncontrollablefactors.

The term embeddedness has various meanings or usages in different topics. For

example, in the research of organisational management and industry development,

embeddedness is regarded as an interaction between people. Granovetter (1985) indicated

that ‘relational embeddedness’, which focuses on the interaction process between people,

will also affect the quality of the interaction. Moorman, Zaltman, and Deshpande (1992)

indicated that an effective interpersonal network will build trust and reduce transaction

costs. The underlying aim of embeddedness is to establish a relationship based on commit-

ment and trust on the Internet. Gruen, Summers, and Acito (2000) argued that suppliers

can help to reinforce the buying incentives of customers, indicating that customers will

have a mental dependency on the network, which is a commitment where the customers

are willing to interact with the website, either emotionally or relationally. However, the

contact between the client and the service staff is indirect and infrequent. In a service

failure scenario, when service recovery and gifts are provided by the suppliers, a trust

relationship built between the suppliers and online consumers may be longer-lasting

than the relationship built between the suppliers and consumers of physical channels,

and such longer-lasting relationship may lead to lower switching intentions of consumers.

H3: The level of the switching intentions of consumers in an online environment has a nega-tive correlation with the level of relational embeddedness.

In the past, scholars tended to work on the features and the impact of electronic word of

mouth. For example, Hanson (2000) indicated that word-of-mouth behaviour on the Inter-

net will enable a faster and more extensive effect between people in the virtual world. If

customers are displeased, they use a communication platform on the Internet to rapidly

disseminate the information to other users. According to previous research, word of

mouth has played a crucial role in affecting consumer behaviour and attitude. Word of

mouth is also more effective than traditional sales by people and various marketing

tools (Bickart & Schindler, 2001). Consumers tend to pay more attention to negative

word of mouth than to positive word of mouth, and do so with more informed judgment

(Miserski, 1982; Herr, Frank, & Kim, 1991). Therefore, consumers will regard this product

as an inferior product. However, they may not regard a product as a high-quality product

when they have heard a positive message regarding this product; as a high-quality product

should have such characteristics, and consumers consider high quality and performance to

be standard (Herr et al., 1991). In general, the more convenient the transaction and the

lower the incidence of consumers receiving negative word of mouth, the more likely it

is that their switching intentions will be lower (Hanson, 2000).

H4: The switching intentions of consumers have a positive correlation with the strength ofword of mouth.

Total Quality Management 841

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

Methodology

Research structure

The research structure is established as shown in Figure 1. The type of service failure attri-

bution is an independent variable, and customers’ negative emotions and service assurance

strength are moderating variables that affect customers’ switching intentions. In addition,

relational embeddedness and strength of negative word of mouth are independent vari-

ables that affect customers’ switching intentions.

Definition of variables

Table 1 lists the established primary research variables of this study. These variables were

designed to be evaluated on a five-point Likert scale.

Sampling methods and sample characteristics

A pre-test of this designed questionnaire was conducted with the assistance of several

travel service companies to ensure the clarity of the questions. The pretest was adminis-

tered to 15 tourists who frequently participated in tourism activities through the tourism

websites.

After the questionnaire was finalised, the samples were collected by convenience

sampling and distributed through electronic mail to the survey participants. To be

precise about the target audience, we cooperated with www.oecfood.net to allow the

website members to fill in the questionnaire through a link on the homepage and

through an e-newsletter that was sent to the website members. We provided incentives

to enhance the willingness of respondents to answer the questionnaire. A notice of this

questionnaire survey was also posted on bbs (electronic bulletin board system) to attract

more participants and to collect a wider spectrum of opinions. A total of 617 question-

naires were collected. A total of 296 questionnaires remained after excluding those partici-

pants who did not frequently conduct online purchases (322 questionnaires). The

respondents with contradictory or illogical answers (12 questionnaires) and respondents

with no service failure experiences (40 questionnaires) were also excluded. A final total

of 243 valid responses were obtained, resulting in a valid response rate of 82.37%. Accord-

ing to the distribution of the research samples by characteristics, the samples generally fall

Figure 1. Research structure.

842 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

Table 1. Conceptual definitions of various aspects in this research.

Dimension Conceptual definitions and questions Reference and revision of related literatures

Type of servicefailure attribution

Service failure attribution refers to the judgment of the attribution ofservice failure occurrence, which is divided into the following twocategories: controllable and uncontrollable. The statements of thisaspect are specified as follows: in general, service failure is caused byan inferior product, according to your consumption experience; ingeneral, service failure is caused by the service policy within theenterprise, according to your consumption experience; in general,service failure is usually caused by service flow, according to yourconsumption experience; in general, service failure may be caused byirresistible factors, such as price policy; in general, service failure iscommonly caused by the online system, according to yourconsumption experience, in general, service failure is caused byfactors that suppliers could prevent in advance, for example,reinforcement of the education training of employees; in general,service failure may be associated with competition in the industryenvironment, according to your consumption experience

In consideration of the viewpoints of Binter et al. (1990) and Kelly andDavis (1994), service failure attribution refers to the judgement ofthe attribution of service failure occurrence, which is divided intocontrollable (attitude and service provided during the servicedelivery process, which belongs to soft service failure) anduncontrollable (occasional service failure, which is commonlyassociated with the system, service and external environment, or theentire industry structure, such as competitions that enterprises areinsufficiently able to manage and improve, which belongs to hardservice failure)

Customers’ negativeemotions

The following five negative emotions of customers, developed byscholars, were designed and compiled as questions for thequestionnaire: furious, upset, angry, annoyed, and displeased

According to the viewpoint of Schoefer and Ennew (2002)

Service assurancestrength

The contents of the questionnaire were designed with regard to standardsof service assurance, the value provided by service assurance, theconsumers’ expectation of service recovery, and the level of easycompensation. The questions of this aspect were listed as follows: inthe belief of unlimited service provided to the customers; in the beliefthat suppliers possess a concept of ‘the customer is always right’; inthe belief that the services provided to the customer by suppliers aremore valuable than those of the competitors; customers will be awareof the service standard and service quality provided by the suppliersprior to consumption; suppliers will be able to efficiently respond tocustomer complaints; when there is a problem with the service, thefrontline employees will be able to solve the problem; we believe thatsuppliers will be able to compensate the customers when there is a

Questions are revised according to the point of view of Hays and Hill(2006)

(Continued)

To

tal

Qu

ality

Ma

na

gem

ent

84

3

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

Table 1. Continued.

Dimension Conceptual definitions and questions Reference and revision of related literatures

service problem; once the customers are dissatisfied with the serviceprovided, the suppliers will offer a discount as a compensation to thecustomer; customers will be compensated by suppliers in various ways

Relationalembeddedness

The sellers are able to increase contact with consumers to increase thequality of the interaction through the use of several methods. Thequestions of this aspect are as follows: you believe that sellers will payattention to the feelings of consumers; you will feel a certain level oftrust in the products and services of sellers; you assume that the sellerwill stay in touch with you for later transactional information; you arecomfortable with the transactional conditions of the seller due tofrequent interactions through the internet; you assume that the servicequality is subject to efficient communication channels with sellers

Granovetter (1985) emphasised on the interaction quality,communication, trust, and other measurements as the basis ofquestionnaire design.

Strength of negativeword of mouth

The strength of negative word of mouth indicates the level of negativitythat is perceived by the receivers of such messages from other users onthe internet. The questions of this aspect include: online negative wordof mouth is persuasive to me; I am quite serious about the negativeword of mouth on the Internet; negative word of mouth on the Internethas a negative meaning to me; I am positive about the negative word ofmouth on the internet; I am quite impressed by online negative word ofmouth

Questions are revised according to the viewpoint of Herr et al. (1991)

Customers’ switchingintentions

The switching intentions of customers refer to the occurrence ofcustomers switching to other stores or other brands due todissatisfaction with the service failure process. The questions of thisaspects include: the slow service of this company makes me feel likeswitching to other companies; the harsh requirements on the specialoffer makes me feel like switching to other companies; the poorattitude of service personnel makes me feel like switching to othercompanies; the slow response to customer feedback makes me feel likeswitching to other companies, the incomplete network system of thatcompany makes me feel like switching to other companies

Questions to be designed by an author who is relevant to this aspect

Note: These variables are designed to be evaluated on a five-point Likert scale.

84

4W

.-B.

Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

into the following two categories: 30–40-year-old age groups and college or higher edu-

cation background, sourced mainly from tourism websites.

Independent tests on these two samples were conducted to ensure that the valid sample

was large enough to represent the entire sampling population. The results are shown in

Table 2. In addition to a significant difference regarding the number of people travelling over-

seas, the number of tourism commodities purchased, and occupation, these two samples

reported insignificant difference in other aspects, such as time for data search, data source,

gender, age, education level, which indicates that these two samples reported the same

core characteristics and that the returned questionnaires are significant to the population.

Method of data analysis

Factor analysis

To reduce the attributes in each dimension, we conducted factor analyses with the princi-

pal component analysis on the type of service failure attribution, the negative emotions of

customers, service assurance strength, relational embeddedness, strength of negative

word-of-mouth message, and switching intentions of consumers and used varimax to

derive the principal factors, with an eigenvalue greater than 1 and a factor loading

greater than 0.5, with a difference of over 0.3 between factors (Hair, Anderson,

Tatham, & Black, 1998).

Reliability and validity analysis

According to Nunnally (1978), a reliability coefficient of above 0.7 represents a high level

of reliability. Cuieford (1965) suggested that a Cronbach’s a-value that exceeds 0.7 rep-

resents a high reliability and a Cronbach’s a-value below 0.35 should be rejected. The

minimum reliability of 0.7 on the various aspects of this research indicates a certain

level of reliability (as shown in Table 3). Regarding the validity aspect, the content validity

may be observed from the citations of related literature, factor loading, and other items,

while the establishment of validity follows the item-total correlation method by Kerlinger

(1986), that is, assuming that the total score is effective, the size of the item-total coeffi-

cient is the measurement indicator for construct validity. The item-total coefficients of

Table 2. Independent test of tourism characteristics and population chi-square test on the samplesfrom different sources (samples returned and unreturned).

ItemTesting method(chi-square test) p-Value

Sample source vs. type of commodity purchased 6.12 0.341Sample source vs. time for data search 5.38 0.398Sample source vs. data source 4.16 0.513Sample source vs. number of tourism commodity

purchased16.28 0.016∗

Sample source vs. average number of local travel 9.26 0.085Sample source vs. average number of abroad travel 11.26 0.039∗

Sample source vs. gender 1.64 0.426Sample source vs. education level 3.15 0.604Sample source vs. occupation 21.38 0.002∗∗

∗p , 0.05.∗∗p , 0.01.

Total Quality Management 845

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

various factors in this research are above 0.5, indicating that there is a certain level of con-

struct validity.

Empirical results

This research introduced the factors derived from service failure attribution as the basis for

clustering. Table 3 presents the results derived from Ward’s method on cluster analysis. In

the determination of the most appropriate cluster number, according to the principle of

screen test, cluster has to be given up to obtain the optimal cluster when the amount

increases drastically. As shown in Table 4, when the types of service failure attribution

are divided into two clusters, in which RSQ and ERSQ demonstrate the highest growth

amount, the percentage of explanatory variables within the cluster will be reduced signifi-

cantly. It is more appropriate to divide service failure attributions into two clusters.

To discuss the effect of cluster analysis, discriminant analysis was conducted to obtain

a classification accuracy ratio to determine the effect of cluster analysis. The accurate dis-

criminant ratio of clusters was tested via cross tabulations between real clustering and

theoretical clustering. As shown in Table 5, the classification accuracy ratios of these

two clusters are 77.95% and 69.83%, with the overall classification accuracy ratio deter-

mined as 74.07% [hit ratio ¼ (99 + 81)/243].

A hierarchical regression analysis was applied to verify H1, with steps taken to place

the independent variables (various types of service failure attributions) and extraneous

variables (negative emotions) into the regression equation by order. The dependent vari-

ables represent the switching intentions of customers, followed by the placing of the cross-

interaction between the independent variables and extraneous variables into the regression

Table 3. Credibility list of various measurement variables in this research.

Potential variables Number of measurement variables Cronbach’s a coefficient

Type of service failure attribution 8 0.7962Customers’ negative emotions 5 0.8634Service assurance strength 10 0.8273Relational embeddedness 5 0.8067Strength of negative word of mouth 5 0.8358Customers’ switching intentions 6 0.8425

Table 4. Cluster analysis results by Ward’s method for service failure attributions.

Number of cluster Number of sample SPRSQ RSQ +RSQ ERSQ +ERSQ

8 27 0.01543 0.759 0.025 0.9213 0.00877 25 0.01768 0.734 0.029 0.9126 0.01146 21 0.02047 0.705 0.031 0.9012 0.02435 30 0.02455 0.681 0.019 0.8825 0.01874 36 0.02793 0.662 0.078 0.8539 0.05843 45 0.03667 0.584 0.091 0.7955 0.14922 59 0.07103 0.493 0.6463

Note: Definition of symbols: SPRSQ: when these two clusters are connected, there is a decrease in the percentageof explanatory variables within the cluster; RSQ: R2; +RSQ: number of RSQ increased by each extra number inthe cluster; ERSQ: ERSQ derived under the assumption of equal distribution; +ERSQ: number of ERSQincreased by each extra number in the cluster.

846 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

equation. If the correlation demonstrates a significant standard regression coefficient, then

that extraneous variables will have an interference effect; otherwise, it has no interference

effect. As shown in Table 6, the cross-interaction between negative emotion and control-

lable service failure attribution has a significant impact on the switching intentions of cus-

tomers (F ¼ 15.38, p , 0.01). Therefore, H1 is supported. We also applied a hierarchical

regression analysis to verify H2. The dependent variables also served as the switching

intentions of customers, and the research placed the independent variables (different

types of service failure attributions) and extraneous variables (negative emotions) into

the regression equation by order, followed by the placement of the cross-interaction

between independent variables and extraneous variables into the regression equation.

As shown in Table 7, the correlation between the service assurance strength and the

uncontrollable service failure attributions has a negative (with 20.42 coefficient)

impact (F ¼ 9.54, p , 0.001) on the switching intentions of customers. Thus, H2 is

also supported.

To verify H3 and H4, a fuzzy neural network technique was used to fuzzify the col-

lected figures and to transform them into the fuzzy volume through membership and

fuzzy set. Thus, we transformed the internal relationships (precise math model) among

the input and output of the descriptive system of the original input and output figures

into a type of corresponding fuzzy relationship presented by the conditional sentence

‘if’ (fuzzy set of input language variance) and ‘then’ (fuzzy set of output language

Table 5. Discriminant analysis of service failure attributions clusters.

Real analysis

Theoretical analysis

Cluster 1 Cluster 1 Total

Cluster 1 99 28 12777.95% 22.05% 100%

Cluster 1 35 81 11630.17% 69.83% 100%

and in total 134 109 24355.14% 44.86% 100%

Table 6. Relationship between various service failure attributions and the switching intentions ofconsumers when a negative emotion extraneous variable is added.

Independent variable and extraneousvariable

Dependent variable: Consumers’ switchingintentions

Controllable service failure type (a1) 0.59∗∗

Uncontrollable service failure type (a2) 0.43∗

Negative emotion (b) 0.18Cross-interaction

a1b 0.31∗∗

a2b 0.11R2 0.61DR2 23.5%F value 15.38∗∗

∗p , 0.05.∗∗p , 0.01.

Total Quality Management 847

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

variance), which fulfilled the fuzzy model of the system. In addition, the semantic variance

was divided into various levels, such as low, medium, and high, or a more detailed var-

iance to achieve a more precise effect. ‘Fuzzification’ � ‘Fuzzy Inference’ � ‘Fuzzy

Decision’ are the basic elements comprising the fuzzy system. The network structure of

the fuzzy system in connection referred to a type of fuzzy neural network. This type of

fuzzy neural network on the end of input and output was completely equal to the fuzzy

system. The internal weighting or point parameter could be modified by learning. In

addition, a certain degree of learning algorithm could automatically result in the proper

shape and fuzzy rule of the membership function. After modifying these membership func-

tions and fuzzy rules, the non-linear model of this system is obtained.

A fuzzy neural network framework of two inputs and one output was used for descrip-

tion. The pattern with more inputs and outputs may be expanded upon in this model. The

framework is shown in Figure 2.

Table 7. Relationship between various service failure attributions and the switching intentions ofconsumers when a service assurance strength extraneous variable is considered.

Independent variable and extraneousvariable

Dependent variable: Consumers’ switchingintentions

Controllable service failure type (a1) 0.78∗∗∗

Uncontrollable service failure type (a2) 0.35∗

Service failure strength (b) 0.24Cross-interaction

a1b 20.42∗∗

a2b 0.15R2 0.58DR2 29.1%F value 9.54∗∗

∗p , 0.05.∗∗p , 0.01.∗∗∗p , 0.001.

Figure 2. Fuzzy neural network framework.

848 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

Layer 1: input layer

Input units: I(1)1 = X1, i = 1, 2,

Output units: O(1)ij = I(1)

i , i = 1, 2; j = 1, 2, . . . , n.

Layer 2: fuzzy interface (linguistic term layer)

In this layer, the researcher used the first layer of input to develop the membership of

the related membership functions upon Gaussian function

Input units: I(2)ij = −

(O(1)ij − aij)

2

b2ij

, i = 1, 2; j = 1, 2, . . . , n,

Output units: O(2)ij = mAij

= exp (I(2)ij ), i = 1, 2; j = 1, 2, . . . , n,

where aij and bij are the centre and the width parameters of the Gaussian function,

respectively.

Layer 3: fuzzy inference (rule layer)

In this layer, the researcher inferred the suitability level of each rule in the rule base

Input units: I(3)(j−l)n+l = O(2)

ij O(2)2l , j = 1, 2, L, n; l = 1, 2, . . . , n,

Output units: O(3)i = mi = I(3)

i , i = 1, 2, . . . , m ( = n2).

Layer 4: defuzzification interface and output (Output layer)

Input units: I(4) =∑m

p=1

O(3)p Wp,

Output units: O(4) = m∗ = I(4)

∑mp=1 O(3)

p

.

The typical rule can be inferred from the above formulas, as below:

if X1 ismA11then W1 = K1

K1 = Constan t (zero − order Sugeno fuzzy model)

or

K1 = p × X1 + q × X2 + r (first − order Sugeno fuzzy model, p, q, r are constants).

In terms of the learning algorithm of membership function, we used the steepest

descent method of the backpropagation model, and the learning algorithm of rule base

(K) was based on least squares estimation.

Total Quality Management 849

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

The steps of the research are as follows.

(i) Reorganise the figures in the returned questionnaires, while there were 184 pieces

of figures.

(ii) Define the assumed input/output variable and the number of the corresponding

membership function of each variable.

(iii) After the training of a fuzzy neural model, calculate one epoch after training 243

pieces of data; manage parameter renewal during this period to reach the opti-

mised membership function shape and rule base.

A fuzzy neural model was obtained after the training, and we tested the influence of the

input variables on the output variables. As we focused on the influence of certain input

variables on the output variables, except for this input variable, other input variables

were fixed (using the average of these 243 pieces of data to obtain the least influence

from these variables).

The reasons to operate the fuzzy neural network model

The main purpose of applying the non-linear fuzzy neural network model is to reduce the

variables for the input of the neural network initial value. The researcher also used a non-

linear method to understand the interaction among the variables. White (1989) also indicated

that a neural network revealed the capacity to recognise data patterns and relationships, which

could be applied to the scope of multivariate data analysis. The main reason that this research

applied a fuzzy neural network model is that this model was the most broadly used and com-

pletely developed model. It is suitable not only for prediction and classification, but also for

uncertain behavioral systems. This method has the following advantages: it could completely

approximate to any non-linear function; the samples in this research represented a type of

high level of non-linear function; and all qualitative messages can be evenly distributed to

the neural in the network. Thus, they have powerful characteristics. The model adopts a par-

allel distribution to quickly manage several calculations, which is suitable for the non-linear

system of complicated behavioral science in business management.

We discussed whether a significant correlation exists between relational embedded-

ness and the switching intentions of consumers, as assumed in H3. In regard to the empiri-

cal results of the fuzzy neural network model, each input variable has two membership

functions – low and high, while the rule base introduces the zero-order Sugeno fuzzy

model. The average testing error rate is 0.1985 after an approximate learning cycle of

119. The testing result is shown in Figure 3, where ‘†’ represents the testing statistics

Figure 3. Various variables and membership function of ‘relational embeddedness’ and ‘consumers’switching intentions’.

850 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

of the input variables, while ‘V’ represents the output statistics inferred from the fuzzy

neural network model. The distribution point of ‘relational embeddedness’ and ‘consu-

mers’ switching intentions’ presents as a cross-status distribution and indicates a negative

correlation. Hence, H3 is supported.

To verify H4, each input variable was set to have two membership functions – low and

high, while the rule base also introduced the zero-order Sugeno fuzzy model. The average

error rate is 0.1987 after a learning cycle of 138. The testing results are shown in Figure 4.

From the distribution point of the input variables and output variables of the fuzzy neural

model, we observed that the distribution point of ‘strength of negative word of mouth’ and

‘consumers’ switching intentions’ report a consistent distribution status, and indicates a

positive correlation. Therefore, H4 is validated.

Discussions and suggestions

The empirical results of the research support all of the hypotheses. From these results, a

number of conclusions are derived. (1) As H1 and H2 are valid, suppliers should aim to

efficiently reduce controllable service failures; otherwise, their efforts to improve the

relationship with consumers may be wasted. This conclusion is consistent with the con-

clusions of Raaij and Pruyn (1998), that is, customers may have expected mental

impressions of the enterprises through service failure attributions, which may further

affect the reactions of customers. (2) In the physical shopping environment, consumers

commonly attribute the service failure to insufficient management by the suppliers,

such as control/education training of service staff (Siew, Swee, & Low, 1997).

However, the research results reaffirmed the influence of the ‘buyer–seller relationship

connection’ and ‘negative word-of-mouth communication’ on the enhancement of exist-

ing customer retention in a ‘non-physical environment’. (3) Although several variables

will affect the switching intentions of consumers, such as peer pressure, brand equity,

and brand loyalty, the ability of the suppliers to take action at the appropriate time may

affect the switching intentions of consumers. For example, Anton, Camarero, and

Carrero (2007) proposed that poorer service quality and lower service commitment will

reduce customer satisfaction. However, it has an indirect impact on the switching inten-

tions of consumers, and the incidence of price inequity provokes the anger of consumers

and has significant effects on the switching intentions of consumers. However, the empiri-

cal results of this research indicated that customers may be able to accept uncontrollable

service failure caused by the suppliers; however, the suppliers must be proactive in

improving those controllable service failures; otherwise ,the result will be insufficient,

even if a customer friendly system is introduced. (5) Although non-physical channels

Figure 4. Various variables and membership function of ‘strength of negative word of word’ and‘consumers’ switching intentions’.

Total Quality Management 851

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

are superior to physical channels, the customer behaviour in non-physical channels is dif-

ficult to manage. A positive relationship with customers that is established online will not

only help to enhance the strength of determinants of relationship, but also conforms to the

findings of Burt (1992), who emphasised that the connection of the relationship structure

will help to increase the control of message and initiative. (6) Regarding H4, the result is

consistent with the elaboration likelihood mood theory, in that the manner in which con-

sumers in this research manage the information is more inclined to central routes. It is

suggested that suppliers should be relational and persuasive about what they deliver to

consumers to gain the recognition of these consumers.

The features of this research are as follows: (1) the empirical results of the linear stat-

istic model and the non-linear fuzzy neural network model were combined, which was

rarely done in previous studies; (2) the scenario perspective was introduced to discuss

the determinants of the switching intention of consumers. This approach distinguishes our

research from those that focused on compensation measures (Harris, Grewal, Mohr, &

Bernhardt, 2006) or determinants of the cognition of fairness of consumers (Sparts &

McColl-Kennedy, 2001; Patterson, Cowley, & Prasongsukarn, 2006).

The suggestions for further research are as follows: (1) researchers can further enable

precise measurements of the possible differences and trends that determine the switching

intentions of consumers through a vertical section in terms of viewpoints in different

stages; (2) they can introduce additional variables that affect the switching intentions of

consumers, such as the service recovery policy of suppliers, quality of service delivery

and after-sales services, and other related variables to complete the service recovery

model of suppliers; and (3) they can also compare various distribution channels, such as

online consumers in various aspects, to derive various management implications.

References

Anton, C., Camarero, C., & Carrero, M. (2007). The moderating effect of satisfaction on consumers’switching intention. Psychology and Marketing, 24(6), 511–538.

Babin, B. & Darden, W. (1995). Consumer self-regulation in a retail environment. Journal ofRetailing, 71(1), 47–71.

Bansal, H.S., & Voyer, P.A. (2000). Word-of-mouth processes within a services purchase decisioncontext. Journal of Service Research, 3(2), 166–177.

Bearden, W.O., Hardesty, D.M., & Rose, R.L. (2001). Consumer self-confidence: Refinements inconceptualization and measurement. Journal of Consumer Research, 28, 121–134.

Bickart, B., & Schindler, R.M. (2001). Internet forums as influential sources of consumer infor-mation. Journal of Interactive Marketing, 15(3), 1–40.

Bitner, M.J., Booms, B.H., & Mohr, L.A. (1994). Critical service encounter: The employee’s view-point. Journal of Marketing, 58(10), 95–106.

Bitner, M.J., Boom, B.M., & Tetreault, M.S. (1990). The service encounter: Diagnosing favorableand unfavorable incidents. Journal of Marketing, 54, 69–82.

Boshoff, C. (2002). Service advertising: An exploratory study of risk perceptions. Journal of ServiceResearch, 4(4), 290–298.

Burt, R.S. (1992). Structural holes: The social structure of competition. Cambridge, MA: HarvardUniversity Press.

Caruana, A. (2002). Service loyalty: The effects of service quality and the mediating role of custo-mer satisfaction. European Journal of Marketing, 36, 811–828.

Craighead, C.W., Karwan, K.R. & Miller, J.L. (2004). The effects of severity of Failure and custo-mer loyalty on service recovery strategies. Production & Operations Management, 13(4),307–321.

Cuieford, J.P. (1965). Fundamental statistics in psychology and education. New York, NY: McGrawHill.

Dawson, S., Bloch, P.H., & Ridgway, N.M. (1990). Shopping motives, emotional states, and retailoutcomes. Journal of Retailing, 66(4), 408–427.

852 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

Derbaix, C., & Pham, M.T. (1991). Affective reactions to consumption situations: A pilot investi-gation. Journal of Economic Psychology, 12(2), 325–355.

Donaton, S. (2003). Marketing’s new fascination: Figuring out word-of-mouth. Advertising Age,74(46), 18.

Dube, L., & Menon, K. (2000). Multiple roles of consumption emotions in post-purchase satisfactionwith extended service transactions. International Journal of Service Industry Management,11(3), 287–304.

Folkes, V.S. (1984). Consumer reactions to product failure: An attributive approach. Journal ofConsumer Research, 10(3), 398–409.

Folkes, V.S., Koletsky, S., & Graham, J.L. (1987). A field study of casual inferences and consumerreaction: The view from the airport. Journal of Consumer Research, 13, 534–539.

Gardner, B. (1985). Market-oriented agricultural policy: Its relevance for 1985. Agribusiness, 1(3),219–225.

Granovetter, M. (1985). Economic action and social structure: The problem of embeddedness.American Journal of Sociology, 91(3), 481–510.

Gruen, T.W., Summers, J.O., & Acito, F. (2000). Relationship marketing activities, commitment andmembership behaviors in professional associations. Journal of Marketing, 64(3), 34–49.

Hair, J.F., Anderson, E.R., Tatham, L.R., & Black, W.C. (1998). Multivariate data analysis.New York, NY: Macmillan.

Hanson, W.A. (2000). Principles of internet marketing. Logan, UT: South-Western CollegePublishing.

Harris, K.E., Grewal, D., Mohr, L.A., & Bernhardt, K.L. (2006). Consumer responses to servicerecovery strategies: The moderating role of online versus offline environment. Journal ofBusiness Research, 59(4), 425–431.

Hays, J.M., & Hill, A.V. (2006). Service guarantee strength: The key to service quality. Journal ofOperations Management, 24(7), 53–764.

Herr, P., Frank, K., & Kim, J. (1991). Effects of word-of-mouth and product-attribute information onpersuasion – an accessibility-diagnostic perspective. Journal of Consumer Research, 17(4),454–462.

Holloway, B.B., & Beatty, S.E. (2003). Service failure in online retailing: A recovery opportunity.Journal of Service Research, 6(1), 92–105.

Keaveney, S.M. (1995). Customer switching behavior in service industries: An exploratory study.Journal of Marketing, 59(2), 71–82.

Kelly, S.W., & Davis, M.A. (1994). Antecedents to customer expectations for service recovery.Journal of Academy of Marketing Science, 22(1), 52–61.

Kerlinger, F.N. (1986). Foundations of behavioral research (3rd ed.). HRW Inc.Meuter, M.L., Ostrom, A.L., Roundtree, R.I., & Bitner, M.J. (2000). Self-service technologies:

Understanding customer satisfaction with technology-based service encounters. Journal ofMarketing, 64(3), 50–64.

Miserski, R.W. (1982). An attribution explanation of the disproportionate influence of unfavorableinformation. Journal of Consumer Research, 9, 301–310.

Mohr, L.A., & Bitner, M.J. (1995). The role of employee effort in satisfaction with service trans-actions. Journal of Business Research, 32(3), 239–252.

Moorman, C., Zaltman, G., & Deshpande, R. (1992). Relationships between provider and users ofmarket research: The dynamics of trust within and between organizations. Journal ofMarketing Research, 29, 314–328.

Mueller, R.D., Palmer, A., Mack, R., & McMullan, R. (2003). Service in the restaurant industry: AnAmerican and Irish comparison of service failures and recovery strategies. HospitalityManagement, 22, 395–418.

Nunnally, J. (1978). Psychometric theory. New York, NY: McGraw Hill.Oliver, R.L. (1993). Cognitive, affective, and attribute bases of the satisfaction response. Journal of

Consumer Research, 20, 418–430.Patterson, P.G., Cowley, E., & Prasongsukarn, K. (2006). Service failure recovery: The moderating

impact of individual-level cultural value orientation on perceptions of justice. InternationalJournal of Research in Marketing, 23(3), 263–277.

Raaij, W.V., & Pruyn, T.H. (1998). Customer control and evaluation of service validity andreliability. Psychology and Marketing, 15(3), 811–832.

Total Quality Management 853

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4

Schoefer, K., & Ennew, C. (2002). Emotional responses to service complaint experiences: The roleof perceived justice. Christel DeHaan Tourism and Travel Research Institute, 1–15.

Siew, M.L., Swee, H.A., & Low, L.H.L. (1997). Effects of physical environment and locus onservice evaluation. Journal of Retailing and Customer Services, 4(4), 231–237.

Smith, A.K., Bolton, R.N., & Wanger, J. (1999). A model of customer satisfaction with serviceencounters involving failure and recovery. Journal of Marketing Research, 36, 356–372.

Sparts, B.A., & McColl-Kennedy, J.R. (2001). Justice strategy options for increased satisfaction in aservices recovery setting. Journal of Business Research, 54(3), 209–218.

Swinyard, W.R. (1993). The effects of mood, involvement and quality of store experience on shop-ping intentions. Journal of Consumer Research, 20(2), 271–280.

Tax, S., Brown, S., & Chandrashekaran, M. (1998). Customer evaluations of service complaintexperiences: implications for relationship marketing. Journal of Marketing, 62(2), 60–76.

Weiner, B. (1985). Spontaneous’ casual thinking. Psychological Bulletin, 99, 186–200.Westbrook, R.A., & Oliver, R.L. (1991). The dimensionality of consumption emotion patterns and

consumer satisfaction. Journal of Consumer Research, 18(1), 84–91.Weun, S., Beatty, S.E., & Jones, M. (2004). The impact of service failure severity on service recov-

ery evaluations and post-recovery relationships. The Journal of Services Marketing, 18(2),133–146.

White, H. (1989). Some asymptotic results for learning in single hidden layer feed-forward networkmodels. Journal of the American Statistical Association, 84, 1003–1013.

Wirtz, J. (1998). Development of a service guarantee model. Asia Pacific Journal of Management,15(3), 51–75.

854 W.-B. Lin

Dow

nloa

ded

by [

Mou

nt S

t Vin

cent

Uni

vers

ity]

at 1

3:30

07

Oct

ober

201

4