Journal of Retailing and Consumer Services 29 (2016) 104–113

Contents lists available at ScienceDirect

Journal of Retailing and Consumer Services

http://d0969-69

n CorrE-m

park.thaCJebara

journal homepage: www.elsevier.com/locate/jretconser

Internet service providers' service quality and its effect on customerloyalty of different usage patterns

Thu Nguyen Quach a, Paramaporn Thaichon b,n, Charles Jebarajakirthy a

a Swinburne University of Technology, Victoria, Australiab S P Jain School of Global Management, 5 Figtree Drive, Sydney Olympic Park, Sydney, New South Wales 2127, Australia

a r t i c l e i n f o

Article history:Received 22 May 2015Received in revised form18 November 2015Accepted 21 November 2015Available online 5 December 2015

Keywords:Internet usage patternAttitudinal loyaltyBehavioural intentionsServices qualityInternet service provider (ISP)High-tech services

x.doi.org/10.1016/j.jretconser.2015.11.01289/& 2015 Elsevier Ltd. All rights reserved.

esponding author.ail addresses: Squach@swin.edu.au (T.N. Quacichon@spjain.edu.au (P. Thaichon),jakirthy@swin.edu.au (C. Jebarajakirthy).

a b s t r a c t

This study attempts to investigate the dimensions of an ISP's service quality, and their effects on cus-tomer loyalty in high-tech services. Data was obtained from 1231 internet users. The analyses includesegmenting ISPs' customers on the basis of their usage pattern and evaluating their perceptions of In-ternet service quality dimensions. Through the use of structural equation modelling and bias correctbootstrapping techniques, the study confirms that service quality dimensions can influence both atti-tudinal and behavioural loyalty. These effects, however, are different across different groups of ISPcustomers. The contribution of the present paper stems from the modelling of mediation effects and theincorporation of Internet usage that can help better explain the impact of service quality dimensions oncustomers' loyalty in high-tech service settings.

& 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Service quality is an important differentiator in a competitivebusiness environment, and a driver of service-based businesses(Zhao and Benedetto, 2013). By enhancing service quality, busi-nesses can influence customers' retention and loyalty (Thaichonet al., 2012). However, very few studies have assessed how dif-ferent aspects of Internet service providers' (ISP) service qualitywould influence their customers' loyalty (Vlachos and Vrecho-poulos, 2008). ISPs may benefit from obtaining accurate informa-tion regarding their customers' assessments of their brand's de-livered service quality; such information may enable service brandmanagers to formulate appropriate marketing strategies in orderto achieve competitive advantage and long term profitability. Thispaper attempts to fill this important research gap by investigatingthe effects of ISPs' service quality on their customers' loyalty in thehigh-tech Internet services. In addition, this paper segmentingISPs' customers on the basis of their usage pattern and evaluatingtheir perceptions of Internet service quality dimensions.

With the rise of technology-enabled services, the attention ofthe services literature has shifted to measurement and oper-ationalisation issues in service quality (Carlson and O’Cass, 2011;Ganguli and Roy, 2010). The earliest service quality model was

h),

introduced by Parasuraman et al. (1985), and was referred asSERVQUAL, including (1) tangibles; (2) reliability; (3) responsive-ness; (4) assurance; and (5) empathy. In addition to SERVQUAL,E-S-QUAL has been developed by Parasuraman et al. (2005) as anattempt to fully capture service quality in the new informationage. However, telecommunications service quality cannot effec-tively be measured by SERVQUAL or E-S-QUAL (He and Li, 2010) asthese scales lack the ability of addressing specific issues relevant tothis particular context, especially high-tech ISPs. In particular,while SERVQUAL applies to general service, E-S-QUAL focuses onservice providers who operate via the internet platform (Vlachosand Vrechopoulos, 2008) and not those who provide the internetconnection and platform for online business-to-business andbusiness-to-customer activities.

On the other hand, segmentation can help better leverage aservice provider's resources and capabilities to fully take ad-vantage of existing opportunities (Weinstein, 1987). As the needsof consumers are not homogenous, it is essential to divide themarket into various segments (Mazzoni et al., 2007). Although theconcept of market segmentation has been discussed extensively inthe literature (Mazzoni et al., 2007; Wedel and Kamakura, 2003),there is scarce empirical evidence of how ISPs can effectivelysegment their target audience. In this study customers are seg-mented based on their usage pattern, which is one of the mostlogical basis of segmentation in similar types of services (Mazzoniet al., 2007; Wedel and Kamakura, 2003). More specifically per-ceptions of service quality dimensions and their relationships withloyalty of light, medium and heavy users will be evaluated. Based

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113 105

on the foregoing discussion, the objectives of this research studyare: firstly, to establish the relationships between specific servicequality dimensions of residential internet services and customers'behavioural and attitudinal loyalty. Secondly, to investigate thedifferences between light, medium and heavy users of the high-tech residential internet services. Lastly, to provide managerialimplications to high-tech residential internet service providers.

2. Literature review and development of hypotheses

2.1. ISPs' service quality dimensions

Previous research indicates that judgment of overall servicequality in the telecommunications industry comes from custo-mers' perceptions of a stable and strong network quality (Lai et al.,2009), ready-to-serve customer support team (Aydin and Özer,2005), informative website support (Thaichon et al., 2012) and ahigh level of security and privacy that is trusted by customers(Roca et al., 2009). Network quality is one of the core servicedrivers in the telecommunications industry (Lai et al., 2009).Network quality in the internet services industry involves thequality and strength of the network signals (Wang et al., 2004),number of errors, downloading and uploading speed (Vlachos andVrechopoulos, 2008). Any break in the internet connectivity maylead to low perceptions of service quality. Moreover, when cus-tomers face problems in high-tech internet services, they oftenseek help and support from technical and customer service staff.For this reason, customer service teams are under constant pres-sure to perform their work reliably, dependably, and according toset protocols in order to meet their productivity goals and deliverquality customer service (Rod and Ashill, 2013). A study in theTurkish telecommunications industry demonstrates that handlingcustomers' complaint efficiently contributes to overall servicequality (Aydin and Özer, 2005).

Information technology tools are utilised to increase efficiencyand effectiveness of information delivery (Ganguli and Roy, 2010).Relevant, timely, and reliable information helps customers to ob-tain information and enable effective decision making (Hsieh,2013). Moreover, information quality plays an important role inbuilding customers' overall positive attitude towards the company(Jaiswal et al., 2010). In fact, a service provider facilitating highlevels of information quality and website support is often per-ceived to have better service quality. There have been considerableconcerns regarding safety and ethical behaviour in e-commerce(Limbu et al., 2011). Customers are prone to attribute low risks inpurchasing from service providers who are reputable in relation totheir security practices (Roca et al., 2009). Security refers to theextent a customer perceives the entire transaction as being safe,which includes payment methods and transmitting confidentialinformation (Chang and Chen, 2009; Thaichon et al., 2014). Privacyis often a concern of customers of high-tech services, and thisdimension relates to the customer perception of the quality ofprocesses used for personal information transmission and storage(Özgüven, 2011). Several studies report that security and privacyare related to service quality (Wolfinbarge and Gilly, 2003).

2.2. Behavioural and attitudinal loyalty

The concept of customer loyalty has received considerable at-tention in the marketing literature. There are many approaches tomeasuring customer loyalty and several studies have attempted todefine the “true nature” of loyalty. Basically, several researchersexplain loyalty purely from the behavioural point of view (Jaiswaland Niraj, 2011) whilst some argue that an attitudinal perspectiveis more reflective of customer loyalty (Flint et al., 2011; Jacoby and

Chestnut, 1978). This research embraces an integrated theory,which suggests that customer loyalty is a combination of bothbehavioural and attitudinal loyalty (Dick and Basu, 1994; Oliver,1999). In this respect, Flint et al. (2011) consider customers' loyaltyas a concept with multiple aspects including repurchase intentionand corresponding preferences and attitudes towards the brand.While behavioural loyalty is defined as repeat purchase (Zeithamlet al., 1996), this study considers attitudinal loyalty as customers'inner thoughts of attachment, word-of-mouth, and re-commendations (Zeithaml et al., 1996).

2.3. Relationships between service quality dimensions, behaviouraland attitudinal loyalty

It is commonly acknowledged that service quality drives cus-tomer loyalty and company profitability (Prentice, 2014). Thisstudy has endeavoured to study the effects of each service qualitydimension on behavioural and attitudinal loyalty. Network qualityincluding connectivity quality, clarity of signals, and speed of in-ternet is deemed to be the fundamental quality characteristics inhigh-tech services which affect customer retention (Seo et al.,2008). Other scholars also confirm that network quality is one ofthe most important drivers of customer loyalty when dealing withprepaid cell phone (Miranda-Gumucio et al., 2012). In the contextof the US mobile phone services, Cassab (2009) demonstrated thatnetwork quality has the largest coefficient values in the regressionanalysis of experimental data and therefore, has stronger influenceon customers' loyalty intention. Similar results are reported in theUS wireless services by Seo et al. (2008), who state that the con-nectivity quality of wireless is positively and significant related tocustomers' repurchase intention. Based on the foregoing discus-sion, we propose the following hypotheses:

H1a : Network quality is positively associated with customers'attitudinal loyalty.

H1b : Network quality is positively associated with customers'behavioural loyalty.

It has been suggested that customer service has a positive andsignificant impact on customer loyalty in technology-based bank-ing services (Ganguli and Roy, 2011). Responsiveness of technicaland customer service staff has significant positive influence onbehavioural loyalty in the Greek mobile telephony service (San-touridis and Trivellas, 2010). Customer care is the major determi-nant of repeat purchase intention and customer loyalty, whiledeficiencies in the quality of customer service are the main rea-sons for customers switching (Miranda-Gumucio et al., 2012).Hence, the following has been hypothesised:

H2a : Customer service and technical support influence custo-mers' attitudinal loyalty.

H2b : Customer service and technical support influence custo-mers' behavioural loyalty.

A service provider facilitating high levels of information qualityand website support is able to maintain a long term relationshipwith customers (Canhoto and Clark, 2013). Likewise, researchersreport that website design is found to be a determinant of loyaltyof online customers in South Africa and Australia (Caruana andEwing, 2010). Other website characteristics such as ease of use andinformation are significant influencers of customer loyalty ine-commerce (Toufaily et al., 2013) and content websites (Jaiswalet al., 2010). Based on the above discussion, the following arehypothesised:

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113106

H3a : Information quality influence customers' attitudinal loyalty.

H3b : Information quality influence customers' behaviouralloyalty.

Previous research reports that security and privacy are relatedto customer loyalty, especially in the e-commerce contexts (Rat-nasingham, 1998). Recently, Limbu et al. (2011) reported a positivelink between customer privacy and website loyalty in the USA. Inonline financial services, privacy is shown to have a direct effect onintention to recommend (Finn et al., 2009). Researchers believethat customer privacy plays a significant part in determining loy-alty (Jaiswal et al., 2010). Jin and Kim (2010) state that the con-tribution of security and privacy towards customer loyalty isgreater in online multichannel retailers than in offline retail set-tings in Korea. Hence, the following have been hypothesised:

H4a : Security and privacy influence customers' attitudinal loyalty.

H4b : Security and privacy influence customers' behaviouralloyalty.

There exists a strong relationship between attitudinal loyaltyand customer repurchase intentions as suggested by Bandyo-padhyay and Martell (2007). Han and Hyun (2012) demonstratethat conative loyalty has a positive effect on action loyalty in hotelservices in the United States. In the Korean online marketplaces,attitudinal loyalty is positively related to purchase intentions(Hong and Cho, 2011). A research in the Chinese mobile phoneservice reported that attitudinal loyalty has a significant positiveeffect on behavioural loyalty (Zhang et al., 2010). Therefore, wepropose that an ISP's service quality dimensions have both a directand indirect effect on its customers' behavioural loyalty. The in-direct effect is channelled via its customers' attitudinal loyalty.This is articulated in the proposed conceptual model depicted inFig. 1, and the following is hypothesised:

H5 : Attitudinal loyalty is a mediator in the relationship betweenspecific ISPs' service quality dimensions and behavioural loyalty.

Apart from identifying the four ISP service quality dimensionsin high-tech residential internet services, it is of interest that,different customers have distinctive needs and require tailoredapproaches (Mazzoni et al., 2007). In this process, customers aregenerally segmented as heavy, medium and light users (Thaichonet al., 2014). Heavy users are those who spend more than 29 h onthe internet every week (Assael, 2005). The average time spent onInternet of a normal person is from 9 hours to 20 h per week(ACMA, 2012). Heavy users often spend more than 29 h on theInternet weekly, while light users only use the Internet for lessthan 9 h every week (Assael, 2005). A study by Electronic

Fig. 1. Conceptual framework.

Transactions Development Agency (ETDA, 2013) discloses that ingeneral Internet users in Thailand who are online for 11 h perweek make up 35.7 per cent; those who use Internet from 11 to20 h weekly constitute 25.8 per cent; 10.7 per cent spend from 21to 41 h on the Internet per week; and 27.8 per cent spend morethan 41 h every week. Based on the usage segmentation of pre-vious research, this study determines three main groups of Inter-net users, namely light (i.e. less than 9 hours per week), medium(i.e. 9–29 h per week), and heavy users (i.e. more than 29 h perweek) (Thaichon et al., 2014). Heavy internet users, who are thelikes of online game players or frequent internet surfers, wouldpossibly perceive network quality as the pre-dominant dimensionwhich influences their perception of an ISP, and might considernetwork quality as the priority to choose and remain with a ser-vice provider. Light users would possibly perceive customer ser-vices and technical support as their priority for recommending aninternet service or in actual behaviour of staying with an ISP, sincethey are usually unfamiliar with technical issues. Therefore, it ismost likely that each specific service quality dimension dis-tinctively impacts customer loyalty (Prentice, 2014), which refersto both attitudinal and behavioural perspectives, depending ondifferently segmented groups of customers. Hence, the followinghave been hypothesised:

H6 : ISPs' service quality dimensions differently influence attitu-dinal loyalty across the three segments of light, medium and heavyusers.

H7 : ISPs' service quality dimensions differently influence beha-vioural loyalty across the three segments of light, medium andheavy users.

3. Method

3.1. The study sample

To test the hypotheses, an online survey was designed andconducted in all regions of Thailand. Thailand is endowed with awide variety of natural resources, a substantial population and arelatively strong economy. The enhanced investment on educationhas resulted in knowledge improvement and higher educationalqualification of the Thai people. Thailand is fast becoming an in-formation society. This is part of the reason for the considerabledevelopment of the telecommunications industry in Thailand.Overall contribution to GDP from the Internet in Thailand is ex-pected to be 3.8 per cent p.a. in 2020 (Telenor, 2015). As the In-ternet is a capital good that enables increased production across alleconomic activities, ISP industry plays a very important part in theThai economy (Telenor, 2015).

Internet in Thailand has passed its infancy and is currently inthe growth stage, poised to take off (Telenor, 2015). Thailand isranked third in South East Asia by way of residential internetusage with an estimated 17 million internet users in 2009 (CIA,2013) and over 27 million internet users in 2014 (NECTEC, 2015),representing a penetration rate of approximately 40% (NECTEC,2015). The Internet penetration in Thailand has seen a rapid in-crease since 2014 and the average internet user growth rate is 30%per year (NECTEC, 2015). Therefore, this provides a suitable sce-nario to study the behaviour of Internet users. However limitedresearch on customers' buyer behaviour of home internet serviceshas been conducted in Thailand. On the other hand, the compe-tition in Thailand among residential internet service providers isintense (Kim, 2015). Currently there are three majors ISPs and 16smaller ones across the country (Thaichon and Quach, 2013). Inthis highly competitive market, the churn rate of internet users

Table 1Sample structure.

Demographic profile Current study(%)

2013 (%) Chi squarestatistics

Gender χ2 (1)¼2.388,p¼0.877

Male 45.6 47.8Female 55.4 52.2

Age χ2 (4)¼8.405,p¼0.078

o19 7.5 6.620–29 29.1 30.830–39 34.3 32.740–49 17.2 19.350þ 11.9 10.6

Monthly household income χ2 (4)¼7.344,p¼0.119

Under 10,000 7.2 6.410,001–30,000 34.9 36.330,001–50,000 20.6 21.550,001–100,000 23.9 24.5Over 100,000 13.4 11.3

Education χ2 (3)¼6.963,p¼0.073

Secondary or belowqualification

6.1 5.5

2 Years college or associatedegree

8.3 6.6

Bachelor's degree 59.7 61.1Postgraduate degree orhigher

25.9 26.8

Number of Internet users in ahousehold

χ2 (4)¼5.327,p¼0.255

1 13.2 14.72 14.7 15.93 18.9 19.24 25.9 24.25 or more 27.3 26

Location χ2 (1)¼2.855,p¼0.091

Bangkok and surroundingsuburbs

48.3 45.9

Others 51.7 54.1

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113 107

was approximately 12% in 2009 (Thaichon et al., 2012). This sce-nario, therefore, poses huge challenges to ISPs especially in thearea of customers' repurchase intention.

3.2. Data collection

Data was collected from residential internet users in all regionsof Thailand in 2013. Customer databases of well-established ISPswho account for 95% of the Thai home Internet services (True,2013) in Thailand were utilised as the sampling frame. The se-lected comprehensive customer databases incorporates diversecustomer profiles, including those who have switched from otherISPs or those who wish to change to other service providers.Hence, the sampling frame is representative of the entire popu-lation of Thai home Internet service customers. Simple randomi-sation was chosen to achieve freedom from human bias and toavoid classification errors (Black, 1999). This means that each in-dividual in the population had an equal chance of being includedin the sample (Teddlie and Yu, 2007). As such, the participants forthe survey were randomly selected from the customer databasesby computer software. This sampling technique has several ad-vantages as it is simple and easy to apply, especially when acomprehensive list of ISP customers was obtained (Rossi et al.,2013). The corporations did not have any control in determiningwho participated in the survey, and did not benefit from influen-cing or creating bias during the data collection process. The surveyresponses were stored in the Opinio database. The companies didnot have access to this data. They were only interested in an in-dependent and academically rigorous process, the findings ofwhich would assist them in developing and designing long-termcustomer retention strategies.

It was calculated that the representative sample of Thailand'spopulation would be a number exceeding 700 (using a confidencelevel of 95 per cent and a sampling error of 72.5). Hence weemailed a total of 4000 surveys in two stages, i.e. 2000 surveyswere distributed in all geographical regions of Thailand and theother 2000 were similarly emailed to participants a week later.The online survey was made available via the Opinio platform. Theweb link was relayed for the online survey and emailed to ISPcustomers who were randomly selected from the databases. Re-sponses to the online survey were automatically returned to theresearchers through the Opinio platform. Opinio software enablesthe production and reporting of a survey and assures the anon-ymity, confidentiality and privacy of the respondents. In order toachieve accurate results, and in particular, to prevent multipleattempts by the same respondent, the default in Opinio was set asfollows: ‘not allow multiple submissions', and ‘prevent withcookies and IP-address check'. This means that the Opinio softwarerecognised every respondent's IP address (computer ID), and onlyone completed survey was accepted from a particular computer.The Opinio platform was kept live for a period of three months.

3.3. Respondent profiles

The final usable sample size was 1231. In terms of the profile ofthe respondents, as illustrated in the Table 1, the sample structureis very similar to the overall structure of Internet users as specifiedin previous research in Thailand (ETDA, 2013). For instance, 45.6per cent of the total respondents were males, while 55.4 per centwere females. This is consistent with the results of previous studyof Internet users in Thailand in 2013 with 47.8 per cent males and52.2 per cent females (ETDA, 2013). Similarly, Generation Y madeup more than 50 per cent of the respondents, consistent with thefindings of the earlier research (ETDA, 2013). In addition, re-spondents with income from 10,001 to 30,000 baths were thelargest group while Bachelor's degree dominated the categories of

education level. Most of respondents stayed in a household with3 or more Internet users. These findings were in line with the 2013study of ISP customers in Thailand (ETDA, 2013). Almost half of therespondents in this study came from Bangkok and its surroundingsuburbs. This can be explained by the fact that the penetration rateof Internet in Thailand is only approximately 40% (NECTEC, 2015)and Internet is mainly available in cities and urban areas. In orderto further compare the respondents' structure in the sample andthe overall population structure, Chi-square statistics were ob-tained for each of the socio-demographic variables. The resultspresented in Table 1 indicate that there is no significant differencebetween the sample and population distributions. In terms of In-ternet usage, in this study 21.2 per cent of the respondents werelight users, 33.1 per cent were medium users and 45.7 per centwere heavy users.

In terms of Internet usage groups, a majority of light users werebetween 29 and 38 years (i.e. 41.2 per cent) and 39 and 49 years(i.e. 35.7 per cent). Interestingly 68.9 per cent of pensioners fell inthe category of light users. This could be explained by the factelder people are usually not very conversant with technology,especially in an Asian context. In terms of gender, females weremore likely to be light users than males. In addition, approxi-mately 70 per cent of full time workers were medium users. Therewere no noticeably uncommon patterns in the age groups and thedistribution of gender among medium users. Heavy users weremore distinctive when compared to the other two segments.Young Internet users made up the largest percentage of this group;

Table 2Online activitiesn.

Online activity Heavyuser (%)

Medium user(%)

Lightuser (%)

Difference test

Emails 30.4 44.3 67.4 χ2 (2)¼5.823,p¼ .054

Search forinformation

54.2 15.1 40.2 χ2 (2)¼10.621,p¼ .006

Social media 38.1 33.5 20.1 χ2 (2)¼2.669,p¼ .259

n Respondents could give more than one answer.

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113108

more than 61.3 per cent were in the 18–28 age bracket and 49.7per cent were in the 29–38 age bracket. Consequently, more thanhalf of the students in this study were classified as heavy users. Inrelations to online activities, heavy users spent most time onsearching for information (i.e. 54.2 per cent), 38.1 per cent selectedsocial media and 33.9 per cent picked emails. In contrast, 44.3 percent of medium users chose emails as their primary activity fol-lowed by social media (i.e. 33.5 per cent) and information search(i.e. 15.1 per cent). Similarly, emails were the most popular amonglight users with 67.4 per cent; information search and social mediawere next with 40.2 per cent and 20.1 per cent respectively. Chisquare difference test was conducted and reported in Table 2. Theresults indicate a significant difference between these groups ofcustomers in terms of information search behaviour.

3.4. Measures

Respondents were required to rate their perceptions for everyitem using a likert scale which was anchored at 1 for stronglydisagree and 5 for strongly agree. These statements originally inEnglish were translated into Thai language by a professionaltranslator. Subsequently, the translated Thai version was backtranslated into English. Significant misunderstanding or confusioncaused by a cross-cultural transformation was detected throughthe back-translation process. Discrepancies in the translation werecarefully inspected and corrected to ensure that the items re-flected the original meaning, and did not contain any socialjudgments. To confirm the error-free translation, the translatedversions were then crosschecked by three other bilingual re-searchers to ensure face and content validity. The items of the

Table 3Instrument items and reliability indices.

Items

NQ I do not experience any Internet disconnection from this ISPThe Internet downloading and uploading speed meet my expectatiThe Internet speed does not reduce regardless peak or off-peak ho

CS Customer service personnel are knowledgeableCustomer service personnel are willing to respond to my enquiriesMy technical problems are solved promptly

IW This ISP provides sufficient informationThis ISP provides up-to-date informationThis ISP provides relevant information

SP I feel that my personal information is protected at this ISPI feel that my financial information is protected at this ISPI feel that the transactions with this ISP are secured

AL I consider myself to be a loyal patron of this ISPI would say positive things about this ISP to other peopleI would recommend this ISP to someone who seeks my advice

BL I would consider this ISP as my first choice to buy servicesI would do more business with this ISP in the next few yearsI would do less business with this ISP in the next few years (-)

Notes: FL¼ factor loadings, α¼Cronbach's alpha, CR¼Construct reliability, AVE¼Averasupport; IW¼ information and website support; PS¼privacy and security; AL¼attitudin

survey are shown in Table 3, which is depicted in the Section 4.Vlachos and Vrechopoulos (2008) connection quality scale wasused to measure network quality. The scale examines connectionquality for mobile phone services which is very similar to thenature of an ISP's network quality. The customer service scale wastaken from Wolfinbarge and Gilly (2003) which addresses bothcustomer service and technical support in the ISP's offerings. Fourdifferent scales relating to information and website support fromChae et al. (2002), Lin (2007), Kim and Niehm (2009) and Vlachosand Vrechopoulos (2008) were considered. After a thoughtfulanalysis, the Kim and Niehm's (2009) information quality scalewas selected as this scale has stronger factor loadings (.80–.83),and Cronbach's alpha rate (α¼ .96). Vlachos and Vrehopoulos's(2008) privacy scale was selected. The scale's measurement itemsinvestigate whether customers feel safe parting with informationduring transactions, and also seek their opinions on security fea-tures of an ISP. The loyalty scale from Kim and Niehm (2009) waschosen to measure attitudinal loyalty since it has strong factorloadings (.71–.95) and Cronbach's alpha (α¼ .93). Additionally, thisscale covers all the aspects of attitudinal loyalty and is used todetermine if customers consider themselves to be loyal patrons ofa particular ISP. Zeithaml et al. (1996) behavioural loyalty scale andcomplaining behaviour scale were selected to measure anotheraspect of loyalty in the ISP market. They have reasonable factorloadings (.74–.94) for behavioural loyalty scale; and (.76–.99) forcomplaining behaviour scale. They aim to investigate whether thecustomer wishes to stay with a particular ISP in the next few yearsor to switch.

4. Analysis and results

4.1. Factor analysis and validity testing

The multi-scale nature of the data comprising ordinal scalesrequires the use of polychoric correlation matrices of softwareprogrammes (Hair et al., 1998). Hence, AMOS version 20 (Analysisof Moment Structures) was employed to analyse the data. Theitems used to assess the ISP's service quality were factor analysedto confirm dimensionalities. Confirmatory factor analysis (CFA)was performed to examine whether theoretical relationship be-tween items and their hypothesised factors were supported by thedata (Cunningham, 2010). Table 3 depicts 6 constructs in the

FL α CR AVE

.714 .828 .833 .625ons .845urs .809

.872 .887 .889 .727

.886

.798

.832 .876 .864 .679

.822

.818

.734 .855 .847 .651

.765

.911

.822 .917 .913 .779

.898

.924

.897 .797 .796 .576

.800

.532

ge variance extracted, NQ¼network quality; CS¼customer service and technicalal loyalty; BL¼behavioural loyalty.

Table 4Correlations between variables.

NQ CS IW PS AL BL

NQ .625 .368 .461 .340 .378 .383CS .607 .727 .557 .364 .324 .323IW .679 .746 .679 .579 .490 .458PS .583 .603 .761 .651 .432 .419AL .615 .569 .700 .657 .779 .560BL .619 .568 .677 .647 .748 .576

Notes: The diagonal elements are the AVEs (italicised and bolded). The lower-lefttriangle elements (italicised) are correlations and the upper-right triangle elementsare the squared correlations between constructs. All correlations are significant atthe.01 level (2-tailed). NQ¼network quality; CS¼customer service and technicalsupport; IW¼ information and website support; PS¼privacy and security;AL¼attitudinal loyalty; BL¼behavioural loyalty.

Table 6Regression results for the mediation of the effect of attitudinal loyalty on beha-vioural loyalty.

Structural path Standardised coefficients Standardised indirect effectsDirect modelNQ-BL .242nnn –

CS-BL .063 –

IW-BL .268nnn –

PS-BL .267nnn –

R2 BL1¼ .524Mediation modelNQ-AL .222nnn –

CS-AL .036 –

IW-AL .326nnn –

PS-AL .258nnn –

AL-BL .938nnn –

NQ-BL .043 .208nnn

CS-BL .030 .033IW-BL .050 .306nnn

PS-BL .025 .242nnn

R2BL2¼ .913

Notes: NQ¼Network quality; CS¼Customer service and technical support;IW¼ Information and website support; PS¼Privacy and security; AL¼Attitudinalloyalty; BL¼Behavioural loyaltyR2BL1 is the proportion of variance in BL explained by the direct modelR2 BL2 is the proportion of variance in BL explained by the mediation model

nnn pr .001.

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113 109

conceptual model, associated indicators, and their standardisedcoefficients, reliability indices and average variance extracted.Factor loadings of all indicators were positive and statisticallysignificant. Reliability estimates for each of the construct, i.e.Cronbach's alpha, and composite reliabilities exceeded thethreshold .70 (Nunnally, 1978). Table 4 presents the correlationsand squared correlations between constructs. The correlationsbetween constructs were significant, ranging from .583 to .761which were well below the .90 cut-off (Tabachinick and Fidell,2001). Therefore, there was no redundancy or violation of multi-collinearity. In addition, the average variance extracted (AVE) foreach factor was over .50, indicative of adequate convergence(Fornell and Larcker, 1981). The construct AVE estimates werelarger than the corresponding squared inter-construct correlationestimates (SIC), thereby supporting discriminant validity.

4.2. Hypotheses testing

Structural equation modelling (SEM) was conducted to ex-amine the research model. SEM is suitable as it allows testing ofstructural models, specifically those containing latent constructs(Anderson and Gerbing, 1988). SEM enables the estimation ofmultiple and crossed relationships between dependent and in-dependent variables, and is able to denote constructs unobservedin these relationships as well as dealing with measurement errorin the estimation process (Beerli et al. (2004). The results arepresented in Table 5. Whereas Chi-square statistic known to behighly sensitive to sample size was significant (χ2 (120)¼563.180,p¼ .000), the other fit indices (CMIN/DF¼4.693, GFI.970,

Table 5Results for the relationships between service quality dimensions, attitudinal andbehavioural loyalty, coefficients.

Estimate S.E. C.R. P

AL o— NQ .234 .030 7.885 nnn

AL o— CS .045 .038 1.193 .233AL o— IW .400 .053 7.566 nnn

AL o— SP .297 .037 8.048 nnn

BL o— NQ .045 .023 1.958 .050BL o— CS .038 .028 1.358 .174BL o— IW -.062 .041 -1.515 .130BL o— SP .029 .028 1.036 .300BL o— AL .941 .024 39.805 nnn

Notes: 1. NQ¼network quality; CS¼customer service and technical support;IW¼ information and website support; PS¼privacy and security; AL¼attitudinalloyalty; BL¼behavioural loyalty; CFI¼comparative fit index; GFI goodness of fitindex; AGFI¼adjusted goodness of fit index; RMSEA¼root mean square error ofapproximation; TLI¼The Tucker-Lewis coefficient;2. Model fit indices: χ2 (120)¼563.180, p¼ .000, χ2/df¼4.693, GFI¼ .970, AGFI¼ .957,TLI¼ .978, CFI¼ .983, RMSEA¼ .042; 3. R2 (BL)¼ .913; R2 (AL)¼ .53.

nnn p Values are statistical significant at .001 levels;

AGFI¼ .957, TLI¼ .978, CFI¼ .983, RMSEA¼ .042) indicate that themodel was a good fit to the data. The ISP's service quality ex-plained 53.0% of variance in attitudinal loyalty (R2¼53.0%), andthe whole model explained 91.3% of variance in behavioural loy-alty (R2¼91.3%). The direct effects of network quality, informationand website support, privacy and security were found to be sig-nificant on attitudinal loyalty. However, none of service qualitydimensions directly influenced behavioural loyalty. Attitudinalloyalty was the only direct determinant of behavioural loyalty.

In order to test the mediation relationships and indirect effectsof service quality dimensions on behavioural loyalty, two com-peting models, namely direct effect model and mediation model(i.e. the research model) were examined in line with the approachof Singh et al. (1994). The mediation relationship was tested on thebasis of standardised residual co-variances and modification indexvalues. Additionally, the bias corrected bootstrap was used in orderto aid the confirmation of the mediation effect as recommendedby Preacher and Kelley (2011). The results are provided in Table 6.

All service quality dimensions, except for customer service andtechnical support, significantly influenced behavioural loyalty inthe direct model. The mediation model test demonstrates that theexplained variance in behavioural loyalty increased from 52.4% to91.3% by incorporating attitudinal loyalty as a mediator. Themediation effect of attitudinal loyalty was confirmed in the med-iation model which demonstrated the following: (1) higher var-iances, (2) a significant effect of the ISP's service quality dimen-sions (except for customer service and website support) on atti-tudinal loyalty, (3) insignificant relationship between servicequality dimensions and behavioural loyalty, and (4) significantlyinfluence of attitudinal loyalty on behavioural loyalty. In otherwords, attitudinal loyalty fully mediated the relationships betweennetwork quality, information and website support, privacy andsecurity, and behavioural loyalty. Drawing upon the results ofobtained by the bias-corrected bootstrap with 10,000 resamples,significant indirect effects on behavioural loyalty were found forall dimensions, except for customer service and website support.James and Brett (1984) assert that complete mediation happenswhen the effect of the independent variable(s) on the dependentvariable completely disappears once the mediator is added as a

Table 7Results for the relationships between service quality dimensions, attitudinal and behavioural loyalty, coefficients among light, medium and heavy users groups.

Path Light users Medium users Heavy users

Direct effects Indirect effects Total effects Direct effects Indirect effects Total effects Direct effects Indirect effects Total effects

AL o— NQ .213nnn .166n .279nnn

AL o— CS .044 .034 .054AL o— IW .458nnn .398nn .364nnn

AL o— SP .308nnn .386nnn .257nnn

BL o— NQ .033 .201nn .234nn .169nn .149n .318nn .019 .268nn .287nn

BL o— CS � .001 .042 .04 .119n .031 .149n .018 .052 .069BL o— IW � .023 .433nn .41nn � .323nn .356nn .034 .003 .349nn .352nn

BL o— SP .001 .291nn .291nn .084 .345nn .429nn .021 .246nn .267nn

BL o— AL .945nnn .945nnn .894nnn .894nnn .959nnn .959nnn

Goodness of fit indices χ2(120)¼261.439, CMIN/DF¼2.179, GFI¼ .953,AGFI¼ .933, TLI¼ .976, CFI¼ .981, RMSEA¼ .045,

χ2(120)¼23.768, CMIN/DF¼1.923,GFI¼ .951, AGFI¼ .930, TLI¼ .976, CFI¼ .982,RMSEA¼ .044,

χ2(120)¼362.058, CMIN/DF¼3.017,GFI¼ .960, AGFI¼ .943, TLI¼ .975, CFI¼ .980,RMSEA¼ .045,

90% CI¼ .037:.052 90% CI¼ .036:.053 90% CI¼ .040:.050

Chi Square difference test Δχ2(24)¼39.765, p¼ .023

Notes: NQ¼network quality; CS¼customer service and technical support; IW¼ information and website support; PS¼privacy and security; AL¼attitudinal loyalty;BL¼behavioural loyalty; CFI¼comparative fit index; GFI goodness of fit index; AGFI¼adjusted goodness of fit index; RMSEA¼root mean square error of approximation;TLI¼The Tucker-Lewis coefficient;

n pr .05nn pr .01nnn pr .001.

Table 8Chi square difference tests.

Path Δχ2 df p

AL o— NQ 2.189 2 .335AL o— CS .047 2 .977AL o— IW .635 2 .728AL o— SP 1.762 2 .414BL o— NQ 5.516 2 .063BL o— CS 2.724 2 .256BL o— IW 7.715 2 .021BL o— SP 2.189 2 .579

Notes: CS¼Customer Service; NQ¼Network Quality; IW¼ Information and Websitesupport; SP¼Security and Privacy; AL¼attitudinal loyalty; BL¼behavioural loyalty.

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113110

predictor of the dependent variable. On this basis, it can be con-cluded that the relationship between network quality, informationand website support, privacy and security, and behavioural loyaltywas fully mediated by attitudinal loyalty. Hence, H1a, H1b, H3a, H3b,

H4a, H4b, and H5 found support.In order to test H6 and H7 structural equation modelling with

multigroup invariance testing was conducted. Results are shown inTable 7. Firstly, the possibility that a fully constrained model wasinvariant across groups was tested. This means specification of amodel in which all factor loadings, all factor variances and allfactor covariances were constrained equal across light, medium,and heavy users (Byrne, 2004). As indicated in Table 7, the dif-ference of overall Chi square test (Δχ2(24)¼39.765, p¼ .023) wasstatistically significant, indicating that some equality constraintsdid not hold across the light, medium and heavy users.

To further verify the differences in each structural path be-tween the three groups of users, it is recommended to separatelyexamine structural models for the three segments. The models ofInternet usage showed reasonable fit to the data (Table 7). Apreliminary examination of the structural models of the threeusage groups illustrates that customer service and technical sup-port's effects on both behavioural and attitudinal loyalty remainedinsignificant among the three internet user groups. Networkquality, privacy and security, and information and website support

were significant predictors of attitudinal loyalty among the threegroups. Noticeably, information and website support's total effectson behavioural loyalty was only found to be significant amonglight users and heavy users. In fact, though positively affectingattitudinal loyalty, information and website support had a nega-tive direct impact on behavioural loyalty of medium users, whichresulted in insignificant total effect. This signifies some potentialdifferences in the relationships between four service quality di-mensions and behavioural loyalty across the light, medium andheavy groups.

In order to confirm differences in the structural paths of in-terest among three groups of Internet usage, a test for invarianceof factor loadings was conducted. In this process, the un-constrained model was run, and eight paths (from four qualitydimensions to attitudinal and behavioural loyalty) were fixed to beinvariant in all groups to arrive at a constrained model (Cun-ningham, 2010). In the eight models, only one model testing thepath from information and website support towards behaviouralloyalty resulted in significant difference in the chi-square test(Table 8). This result confirms that the effects of information andwebsite support on behavioural loyalty was not the same forpeople from different internet usage groups (Δχ2 (2)¼7.715,p¼ .021). Therefore, H6 was rejected and H7 was partiallysupported.

5. Discussion

This study advances the literature related to service quality as akey determinant of customer loyalty among different groups ofcustomers in high tech services context. The contribution to extantresearch can be considered relatively robust as the research modelexplained a considerable proportion of the variance in the criter-ion variables.

5.1. ISPs' service quality dimensions, attitudinal and behaviouralloyalty

The results show that all service quality dimensions except for

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113 111

customer service and technical support were positively related toattitudinal and behavioural loyalty. Information and website sup-port was the predominant predictor of attitudinal loyalty, andbehavioural loyalty of ISP customers. The role of privacy and se-curity was also confirmed. These results support previous findingson the relationship between loyalty and information and websitesupport (Toufaily et al., 2013), and privacy and security (Limbuet al., 2011). Interestingly, customer service and technical supportwas neither a significant predictor of attitudinal nor behaviouralloyalty, contrasting with the finding of Santouridis and Trivellas(2010), which claimed that this dimension had the most sig-nificant impact on loyalty. This can be explained by the fact thatcustomers in Asian culture typically often seek help from theirfriends or family for technical issues and tend to avoid commu-nicating directly with the service providers due to ego concernsand value orientations (Neuliep, 2012). Another possible ex-planation could be that this dimension manifests its effect viaother variables not included in this study, for example, satisfaction(Caruana and Ewing, 2010).

Network quality significantly contributed to attitudinal andbehavioural loyalty mirroring prior research by Miranda-Gumucioet al. (2012). This finding confirms that the basic need of an ISPcustomer is still the core service performance, i.e. internetdownloading/uploading speed, signal stability and consistency.However, unlike findings of previous research, this dimension wasnot the most significant predictor of attitudinal and behaviouralloyalty in this study. This phenomenon can be justified as networkquality has become more stable and reliable among the ISPs,especially in big cities such as Bangkok, which no longer makes it adifferentiating tool.

5.2. ISPs' service quality dimensions, loyalty among light, mediumand heavy users

In considering the usage profile of respondent, informationquality had different effects on behavioural loyalty across differentgroups of internet users. Information and website support mani-fested its positive and significant impact on behavioural loyalty oflight and heavy users through attitudinal loyalty. The currentstudy was conducted in high-tech services in which customers aremore likely to communicate with the company through websites.When customers feel that they have access to correct and ade-quate information, they are most likely to remain with thecompany.

In contrast, information and website support directly had asignificant negative effect on behavioural loyalty among mediumusers. Medium users are often more conversant with the internetthan light users, which makes information support less necessary.In addition, they do not exhibit a need for considerable amount ofinformation like heavy users or light users. In fact, Table 2 illus-trates that medium users spent more time on emails and socialmedia than on information search, in contrast to the other twogroups of users. Also, whereas many pensioners were light usersand a majority of heavy users were students, most of mediumusers were full time employees who are usually “money rich timepoor”. Therefore, too much information might result in informa-tion overload for medium users and become dysfunctional (Jacoby1984). In this case, it lessens the desire of customers to stay withtheir current service provider. These reasons together might ex-plain why medium users do not attach much importance to in-formation support in their quest towards behavioural loyalty.

5.3. The mediation role of attitudinal loyalty in the relationship be-tween ISPs' service quality dimensions and behavioural loyalty

Results from the testing of the mediation effects confirmed that

the relationship between the ISP's service quality dimensions, i.e.network quality, information and website support, privacy andsecurity, and behavioural loyalty were fully mediated by attitu-dinal loyalty. The model which incorporated attitudinal loyalty as amediator explained 92.9% of the variance in behavioural loyalty ofthe ISP's customers in this study. This finding reveals a salient andactive role of attitudes of customers as a significant determinantfor their repeat purchase behaviour. Fostering attitudinal loyaltyincreases the likelihood of the ISP becoming a supplier of choice,winning new customers through positive word of mouth and re-commendations, and improving customer retention. For this rea-son, it is essential for service providers to investigate the factorsthat influence customer attitudes towards the firm.

6. Implications

This study has both theoretical and practical implications.Overall service quality is widely considered as one of key factorsdetermining customer attitudinal and behavioural loyalty. How-ever the results in this study highlight that in fact different ISP'sservice quality dimensions generate different effects on the twoperspectives of loyalty. It is recommended that research on servicequality should not just focus primarily on network quality but onother service quality dimensions, as prominence has been given tothis dimension in extant literature (He and Li, 2010; Kim and Yoon,2004). This study revealed that service quality dimensions exertdifferent effects on customer loyalty among the various groups ofcustomers segmented based on their usage pattern, confirmingthe importance of customer segmentation. Moreover, this study isoriginal in that it is the first of its kind which attempts to in-vestigate the dimensions of an ISP's service quality, and their ef-fects on loyalty in high-tech services.

This study has evaluated ISPs' service quality dimensions whichare identified as (1) network quality, (2) customer service andtechnical support, (3) information quality and website informationsupport, (4) security and privacy. By enhancing service quality,firms can influence customers' behavioural and attitudinal loyalty,which are critical for an ISP's success and long term sustainability.As a result, ISPs will be able to reduce the current issues relating tocustomer switch and churn rate in the residential internet servicesmarket.

Managers of ISPs should also be aware that, although theconcept of service quality is multidimensional, not all dimensionscontribute equally to loyalty. Different service quality dimensionsexert different effects on customer loyalty. Information and web-site support was found to be the predominant predictor of bothtypes of loyalty instead of the presumed network quality, sig-nifying a potential change of behaviour and perception of ISPcustomers. This underscores a need of improving company web-sites and information support to the service subscribers, such aswebsite accessibility and user-friendly interface, information ac-curacy, and timeliness. Although customer service was not foundto be a direct antecedent of loyalty in this study, companies stillneed to maintain a satisfactory level of their customer service andtechnical support performance and keep up with the industrystandards. In addition ISPs can promote a diversity of supportforms, such as online and offline support. Particularly onlinesupport is convenient for customers experiencing simple issues orthose perceiving direct face-to-face communication as being toointimate and personal. Furthermore, the finding confirms thatinternet customers are heterogeneous. In order to effectively andefficiently retain customers, an ISP needs to understand its cus-tomers' characteristics. It is critical to look into aspects that arerelevant to the attitudinal and behavioural loyalty of specific cus-tomer groups in order to leverage its key resources and maximise

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113112

profitability. By highlighting features of customers' interest, ser-vice providers can better enhance customer loyalty.

7. Limitations and future research directions

There are several limitations of this study. Firstly it is possiblethat customers may have no choice but to stay with the serviceprovider as the switching costs are high (Shirin and Puth, 2011).This explains why seemingly loyal customers, who have largevolumes and high frequency of purchases, can quickly switch toother alternatives as switching costs decrease (Dick ad Basu, 1994;Shirin and Puth, 2011). Switching costs can be considered as thecosts (including monetary, psychological and emotional costs)involved in moving from a service provider to another (Porter,1980). Switching costs help companies to deal with inevitablevariations in service quality (Jones et al., 2000). In other words,despite a decrease in service quality, customers might stay withthe company because the perceived costs of changing to a newservice provider exceed the potential benefits of switching (Lamet al., 2004). De Ruyter, Wetzels and Bloemer (1998) confirm themoderating effects of switching costs on the relationship betweenservice quality and loyalty. As such, incorporating switching costinto future research could open up more opportunities to under-stand the relationship between service quality and loyalty.

In addition, the choice of context (home internet services inThailand) where the proposed conceptual model in this study wasempirically tested may restrict the generalisability of the findings.In the interest of generalisation, replication in other countries andsectors to further evaluate the model would be a step towardsaddressing this limitation. Finally, this study examines the effectsof the cognitive evaluation, i.e. service quality dimensions, oncustomer loyalty. An investigation on customers' affective eva-luations, for instance trust, satisfaction, and commitment, asantecedents of loyalty and their interrelationships with servicequality dimensions might be a fruitful area of research and alsovaluable in overcoming the limitation of this study.

References

ACMA, 2012. Internet use and social networking by young people AustralianCommunications and Media Authority. Available at: ⟨http://www.acma.gov.au/webwr/_assets/main/lib310665/no1_internet_use_social_networking_young_people.pdf⟩ (accessed 01.04.14).

Anderson, J., Gerbing, W., 1988. Structural equation modelling in practice: a reviewand recommended two stage approach. Psychol. Bull. 27 (1), 5–24.

Assael, H., 2005. A demographic and psychographic profile of heavy internet usersand users by type of internet usage. J. Advert. Res. 45 (1), 93–123.

Aydin, S., Özer, G., 2005. The analysis of antecedents of customer loyalty in theTurkish mobile telecommunication market. Eur. J. Mark. 39 (7/8), 910–925.

Bandyopadhyay, S., Martell, M., 2007. Does attitudinal loyalty influence behaviouralloyalty? A theoretical and empirical study. J. Retail. Consum. Serv. 14 (1), 35–44.

Beerli, A., Martin, J.D., Quintana, A., 2004. A model of customer loyalty in the retailbanking market. Eur. J. Mark. 38 (1/2), 253–275.

Black, T.R., 1999. Doing quantitative research in the social sciences: An integratedapproach to research design, measurement, and statistics. SAGE Publications,Thousand Oaks, CA.

Byrne, B.M., 2004. Testing for multigroup invariance using AMOS graphics: a roadless traveled. Struct. Equ. Model. 11 (2), 272–300.

Canhoto, A.I., Clark, M., 2013. Customer service 140 characters at a time: the users’perspective. J. Mark. Manag. 29 (5/6), 522–544.

Carlson, J., O’Cass, A., 2011. Developing a framework for understanding e-servicequality, its antecedents, consequences, and mediators. Manag. Serv. Qual.: Int. J.21 (3), 264–286.

Caruana, A., Ewing, M.T., 2010. How corporate reputation, quality, and value in-fluence online loyalty. J. Bus. Res. 63, 1103–1110.

Cassab, H., 2009. Investigating the dynamics of service attributes in multi-channelenvironments. J. Retail. Consum. Serv. 16, 25–30.

Chae, M., Kim, J., Kim, H., Ryu, H., 2002. Information quality for mobile internetservices: a theoretical model with empirical validation. Electron. Mark. 12 (1),38–46.

Chang, H., Chen, S., 2009. Consumer perception of interface quality, security, and

loyalty in electronic commerce. Inf. Manag. 46 (7), 411–417.CIA, 2013. The World Factbook. Retrieved from: ⟨https://www.cia.gov/library/pub

lications/the-world-factbook/geos/th.html⟩ (accessed 01.08.13).Cunningham, E., 2010. A practical guide to structural equation modelling using

Amos, 1 ed. Statsline, Melbourne.De Ruyter, K., Wetzels, M., Bloemer, J., 1998. On the relationship between perceived

service quality, service loyalty and switching costs. Int. J. Serv. Ind. Manag. 9 (5),436–453.

Dick, S., Basu, K., 1994. Customer loyalty: toward an integrated conceptual frame-work. J. Acad. Mark. Sci. 22 (2), 99–113.

ETDA, 2013. Thailand Internet User Profile 2013 Electronic Transactions Develop-ment Agency. Available at: ⟨http://www.etda.or.th/etda_website/files/system/InternetUserProfile_2013_14.11.13.pdf⟩ (accessed 15.05.2014).

Finn, A., Wang, L., Frank, T., 2009. Attribute perceptions, customer satisfaction andintention to recommend e-services. J. Interact. Mark. 23, 209–220.

Flint, J., Blocker, C., Boutin, P., 2011. Customer value anticipation, customer sa-tisfaction and loyalty: an empirical examination. Ind. Mark. Manag. 40,219–230.

Fornell, C., Larcker, D.F., 1981. Evaluating structural equation models with un-observable variables and measurement error. J. Market. Res., 39–50.

Ganguli, S., Roy, S.K., 2010. Service quality dimensions of hybrid services. Manag.Serv. Qual. 20 (5), 404–424.

Ganguli, S., Roy, S.K., 2011. Generic technology-based service quality dimensions inbanking. Impact on customer satisfaction and loyalty. Int. J. Bank. Mark. 29 (2),168–189.

Hair, J., Anderson, R., Tatham, R., Black, C., 1998. Multivariate Data Analysis, 5 ed.Upper Saddle River, NJ, Prentice Hall.

Han, H., Hyun, S.S., 2012. An extension of the four-stage loyalty model: the criticalrole of positive switching barriers. J. Travel Tour Mark. 29 (1), 40–56.

He, H., Li, Y., 2010. Key service drivers for high-tech service brand equity: themediating role of overall service quality and perceived value. J. Mark. Manag. 27(1/2), 77–99.

Hong, P., Cho, H., 2011. The impact of consumer trust on attitudinal loyalty andpurchase intentions in B2C e-marketplaces: Intermediary trust vs. seller trust.Int. J. Inf. Manag. 31 (5), 469–479.

Hsieh, P., 2013. Perceived opportunism (PO) in e-return service encounters. Manag.Serv. Qual. 23 (2), 96–110.

Jacoby, J., 1984. Perspectives on information overload. J. Consum. Res. 10 (4),432–435.

Jacoby, J., Chestnut, R.W., 1978. Brand Loyalty: Measurement and Management.John Wiley & Sons, New York.

Jaiswal, A.K., Niraj, R., 2011. Examining mediating role of attitudinal loyalty andnonlinear effects in satisfaction-behavioral intentions relationship. J. Serv.Mark. 25 (3), 165–175.

Jaiswal, A.K., Niraj, R., Venugopal, P., 2010. Context-general and context-specificdeterminants of online satisfaction and loyalty for commerce and content sites.J. Interact. Mark. 24, 222–238.

James, L.R., Brett, J.M., 1984. Mediators, moderators, and tests for mediation. J. Appl.Psychol. 69, 307–321. http://dx.doi.org/10.1037/ 0021-9010.69.2.307.

Jin, B., Kim, J., 2010. Multichannel versus pure e-tailers in Korea: evaluation ofonline store attributes and their impacts on e-loyalty. Int. Rev. Retail., Distrib.Consum. Res. 20 (2), 217–236.

Jones, Michael A., Mothersbaugh, David L., Beatty, Sharon E., 2000. Switching bar-riers and repurchase intentions in services. J. Retail. 76, 259–274 (Summer).

Kim, G., 2015. South Asia ISPs Have To Move Faster As Competition Builds, Spec-trum Futures. ⟨http://spectrumfutures.org/south-asia-isps-have-to-move-faster-as-competition-builds/⟩ (accessed 18.08.15).

Kim, H., Yoon, C., 2004. Determinants of subscriber churn and customer loyalty inthe Korean mobile telephony market. Telecommun. Policy 28 (9), 751–765.

Kim, H., Niehm, L.S., 2009. The impact of website quality on information quality,value, and loyalty intentions in apparel retailing. J. Interact. Market. 23 (3),221–233.

Lai, F., Griffin, M., Babin, B., 2009. How quality, value, image, and satisfaction createloyalty at a Chinese telecom. J. Bus. Res. 62, 980–986.

Lam, S.Y., Shankar, V., Erramilli, M.K., Murthy, B., 2004. Customer value, satisfaction,loyalty, and switching costs: an illustration from a business-to-business servicecontext. J. Acad. Mark. Sci. 32 (3), 293–311.

Limbu, Y.B., Wolf, M., Lunsford, D.L., 2011. Consumers’ perceptions of online ethicsand its effects on satisfaction and loyalty. J. Res. Interact. Mark. 5 (1), 71–89.

Lin, J., 2007. Online stickiness: its antecedents and effect on purchasing intention.Behav. Inf. Technol. 26 (6), 507–516.

Mazzoni, C., Castaldi, L., Addeo, F., 2007. Consumer behavior in the Italian mobiletelecommunication market. Telecommun. Policy 31 (10/11), 632–647.

Miranda-Gumucio, L., Gil-Pechuan, I., Palacios-Marques, D., 2012. An ExploratoryStudy of the Determinants of Switching and Loyalty in Prepaid Cell PhoneUsers. An Application of Concept Mapping. Springer-Verlag, Berlin Heidelberghttp://dx.doi.org/10.1007/s11628-012-0177-3.

NECTEC, 2015. Internet User and Statistics in Thailand, National Electronics andComputer Technology Centre. Available at: ⟨http://internet.nectec.or.th/webstats/home.iir?Sec¼home⟩ (accessed 01.08.15).

Neuliep, J.W., 2012. Intercultural Communication: A Contextual Approach. SAGEPublications, Inc., St. Norbert College.

Nunnally, J., 1978. Psychometric Theory. Mcgraw Hill, New York.Oliver, L., 1999. Whence consumer loyalty. J. Mark. 63 (1), 33–44.Özgüven, N., 2011. Analysis of the relationship between perceived security and

customer trust and loyalty in online shopping. Chin. Bus. Rev. 10 (11), 990–997.

T.N. Quach et al. / Journal of Retailing and Consumer Services 29 (2016) 104–113 113

Parasuraman, A., Zeithaml, V.A., Berry, L.L., 1985. A conceptual model of servicequality and its implications for future research? J. Mark. Manag. Vol. 49 (No. 1),41–50.

Parasuraman, A., Zeithaml, V.A., Malhotra, A., 2005. ES-QUAL a multiple-item scalefor assessing electronic service quality. J. Serv. Res. 7 (3), 213–233.

Porter, Michael E., 1980. Competitive Strategy: Technologies for Analyzing In-dustries and Competitors. Free Press, New York.

Preacher, K.J., Kelley, K., 2011. Effect size measures for mediation models: quanti-tative strategies for communicating indirect effects. Psychol. Methods 16 (2),93–115.

Prentice, C., 2014. Who stays, who walks, and why in high-intensity service con-texts. J. Bus. Res. 67 (4), 608–614. http://dx.doi.org/10.1016/j.jbusres.2013.02.044.

Ratnasingham, P., 1998. The importance of trust in electronic commerce. InternetRes. 8 (4), 313–321.

Roca, J., García, J., Vega, J., 2009. The importance of perceived trust, security andprivacy in online trading systems. Inf. Manag. Comput. Secur. 17 (2), 96–113.

Rod, M., Ashill, N.J., 2013. The impact of call centre stressors on inbound and out-bound call-centre agent burnout. Manag. Serv. Qual. 23 (3), 245–264.

Rossi, P.H., Wright, J.D., Anderson, A.B. (Eds.), 2013. Handbook of Survey Research.Academic Press, New York.

Santouridis, I., Trivellas, P., 2010. Investigating the impact of service quality andcustomer satisfaction on customer loyalty in mobile telephony in Greece. TotalQual. Manag. J. 22 (3), 330–343.

Seo, D., Ranganathan, C., Babad, Y., 2008. Two-level model of customer retention inthe US mobile telecommunications service market. Telecommun. Policy 32,182–196.

Shirin, A., Puth, G., 2011. Customer satisfaction, brand trust and variety seeking asdeterminants of brand loyalty. Afr. J. Bus. Manag. 5 (30), 11899–11915.

Singh, J., Goolsby, J.R., Rhoads, G.K., 1994. Behavioral and psychological con-sequences of boundary spanning burnout for customer service representatives.J. Mark. Res. 31 (4), 558–569.

Tabachinick, B.G., Fidell, L.S., 2001. Using multivariate statistics. Allyn and Bacon,Boston.

Teddlie, C., Yu, F., 2007. Mixed methods sampling a typology with examples. J. MixMethods Res. 1 (1), 77–100.

Telenor, 2015. Towards a Connected world Telenor Group. Available at: ⟨http://www.telenor.com/wp-content/uploads/2012/03/Towards-a-Connected-World-1MB.pdf⟩ (accessed 01.04.15).

Thaichon, P., Lobo, A., Prentice, C., Quach, T.N., 2014. The development of servicequality dimensions for internet service providers: Retaining customers of dif-ferent usage patterns. J. Retail. Consum. Serv. 21 (6), 1047–1058.

Thaichon, P., Quach, T.N., 2013. Integrated marketing communications and theirinfluences on brand loyalty: a Thai perspective. International Journal of Studiesin Thai Business. Soc. Cult. 2 (1), 55–79.

Thaichon, P., Lobo, A., Mitsis, A., 2012. Investigating the antecedents to loyalty ofInternet service providers in Thailand: Developing a conceptual model. Paperpresented at the Australian and New Zealand Marketing Academy Conference,Adelaide, Australia.

Toufaily, E., Ricard, L., Perrien, J., 2013. Customer loyalty to a commercial website:Descriptive meta-analysis of the empirical literature and proposal of an in-tegrative model. J. Bus. Res. 66, 1436–1447.

True, 2013. True Corporation Plc At a Glance. Available at:: ⟨http://true.lis

tedcompany.com/misc/PRESN/20130831-TRUE-glanceAugust2013-02.pdf⟩ (ac-cessed 01.04.14).

Vlachos, A., Vrechopoulos, P., 2008. Determinants of behavioral intentions in themobile internet services market. J. Serv. Mark. 22 (4), 280–291.

Wang, Y., Lo, H., Yang, Y., 2004. An integrated framework for service quality, cus-tomer value, satisfaction: evidence from China’s telecommunication industry.Inf. Syst. Front. 6 (4), 325–340.

Wedel, M., Kamakura, W., 2003. Market Segmentation. Conceptual and Methodo-logical Foundations. Kluwer Academic Publishers, Norwell, MA.

Weinstein, A., 1987. Market Segmentation. Probus Publishing, Chicago, IL.Wolfinbarge, M., Gilly, C., 2003. eTailQ: dimensionalizing, measuring and predicting

etail quality. J. Retail. 79 (3), 183–198.Zeithaml, V.A., Berry, L.L., Parasuraman, A., 1996. The behavioral consequences of

service quality. J. Mark. 60 (2), 31–46.Zhang, R., Li, X., Zhang, Y., 2010. Service quality, customer satisfaction and customer

loyalty of mobile communication industry in China. J. Glob. Acad. Mark. Sci. 20(3), 269–277.

Zhao, Y.L., Benedetto, C.A.D., 2013. Designing service quality to survive: empiricalevidence from Chinese new ventures. J. Bus. Res. 66, 1098–1107.

Thu Nguyen Quach holds a Master's Degree in Marketing and is currently workingtowards a PhD programme in Marketing in the Faculty of Business & Enterprise-Swinburne University of Technology, Melbourne, Australia. Her research interestsare in the area of services marketing, marketing research, consumer behaviour andrelationship marketing. Thu's research has been published in the Journal of Retailingand Consumer Services, Journal of Business and Industrial Marketing, and ServicesMarketing Quarterly, among others.

Paramaporn Thaichon is an Assistant Professor of Marketing at the S P Jain Schoolof Global Management, Sydney, Australia. His research has been published inleading marketing journals throughout Europe, North America and Australasia in-cluding but not limited to the Journal of Retailing and Consumer Services, Journal ofBusiness and Industrial Marketing, Asia Pacific Journal of Marketing and Logistics, andJournal of Relationship Marketing.

Dr Charles Jebarajakirthy is based in the Faculty of Business and Law at theSwinburne University of Technology, Melbourne, Australia. His research interestsare in the areas of market orientation, consumer behaviour, social marketing andcorporate social responsibility. Charles's research has been published in the Journalof Retailing and Consumer Services, Asia Pacific Journal of Marketing and Logistics,International Journal of Consumer Studies and Journal of Young Consumers, amongothers.