You are on page 1of 15

ESWA 9990 No.

of Pages 15, Model 5G

4 May 2015

Expert Systems with Applications xxx (2015) xxxxxx


Contents lists available at ScienceDirect

Expert Systems with Applications

journal homepage:


3 An SEMarticial-neural-network analysis of the relationships between

4 SERVPERF, customer satisfaction and loyalty among low-cost and
5 full-service airline
8 Lai-Ying Leong a, Teck-Soon Hew b, Voon-Hsien Lee a,, Keng-Boon Ooi c
9 a
Faculty of Business and Finance, Universiti Tunku Abdul Rahman, Jalan Universiti, Bandar Barat, 31900 Kampar, Perak, Malaysia
10 b
Faculty of Business and Accountancy, University of Malaya, 50603 Lembah Pantai, Kuala Lumpur, Malaysia
11 c
Faculty of Business, Management and Accountancy, Linton University College, Persiaran UTL, Bandar Universiti Teknologi Legenda, Batu 12, 71700 Mantin, Negeri Sembilan, Malaysia

a r t i c l e i n f o a b s t r a c t
1 5
2 5
16 Article history: There is a dearth of studies pertaining to the inuence of SERVPERF on customer satisfaction and 26
17 Available online xxxx customer loyalty among low cost and full service airlines. Prior studies have measured service quality 27
using the GAP-5 model with SERVQUAL; however this study offers a new perspective by using the 28
18 Keywords: SERVPERF with an SEMarticial-neural-networks predictive analytic approach. This is different from 29
19 Customer satisfaction the previous studies as it contributes to application of expert systems and intelligent algorithms in the 30
20 Customer loyalty context of low cost and full service airline. The ndings revealed signicant inuences of SERVPERF 31
21 Airline industry
dimensions on customer satisfaction towards customer loyalty with 63.1% and 55.6% variance explained. 32
22 Articial neural networks
The implications from this research may help CEOs and managers of the air travel and tourism industry 33
24 players to make better decisions in their resource planning stage, at the same time improving customer 34
satisfaction and loyalty. 35
2015 Published by Elsevier Ltd. 36

40 1. Introduction freight ton kilometers (FTK), a 5.9% growth in 2013. In the 58
Malaysia airlines industry context, for the year 2010, it experi- 59
41 In the new era of the borderless world and modern society enced a 12.4% growth of passenger trafc followed by a growth 60
42 together with the emergence of low cost airlines, air travelling rate of 10.7% for 2011 and only 5% for 2012 (Maybank IB 61
43 has been a kind of necessity rather than a luxury activity. Due to Research, 2013). It surged to a 21-year record high of 18.4% in 62
44 the keen competition among global airliners, various marketing 2013 (Maybank IB Research, 2014). 63
45 strategies and approaches have been employed to increase passen- The airline industry in Malaysia is dominated by full service 64
46 ger and cargo trafc volume. Nevertheless, global airline business Malaysia Airlines (MAS) as the national airline with 100 destina- 65
47 has experienced a roller-coaster phenomenon during the last few tions worldwide and Air Asia as the rst low cost carrier serving 66
48 years. According to the CEO of Air Asia, the global airline industry 75 destinations in 21 nations via its Air Asia X, Indonesia Air Asia 67
49 has picked up its pace again in 2010 while the forecast for 2011 is and Thai Air Asia wings (Wong & Musa, 2011). Air Asia has received 68
50 encouraging though with element of caution (Air Asia, 2010 the Best Low-Cost Airline category at Business Traveller 69
51 Annual Report). The International Air Transport Association AsiaPacics 2012 annual Travel Awards ceremony in Hong 70
52 (IATA, 2014) has forecasted the industrys net prot level from Kong (The Star, 2012), the rst ASEAN Commercial Aviation 71
53 USD12.9 billion in 2013 to USD18.7 billion for 2014. IATA further Award 2013 in Langkawi (The Star, 2013a) as well as the Worlds 72
54 estimates a 5.3% in passenger growth (RPK) for 2013 and 5.8% for Best Low Cost Airlines award for 2009 and 2010 from Skytrax 73
55 2014 from 3.129 billions passengers to 3.304 billions in 2014. In and was voted as the top three Best Regional airlines in the low 74
56 terms of the international revenue passenger kilometers (RPK), cost carrier category by Skytrax World Airline Award in 2006. 75
57 IATA has estimated a 6.1% growth while for the international Lately, it has also won the award for Worlds and Asias Best 76
Low-Cost Airlines in the Paris 2013 World Airlines Award (The 77
Star, 2013b) and the World Travel Awards Asia and Australasia in 78
Corresponding author. Tel.: +60 5 4688888; fax: +60 5 4667407. Dubai (The Star, 2013c) whereby MAS was also named as Asias 79
E-mail addresses: (L.-Y. Leong), Leading Airline. The competition between the two airlines has 80
(T.-S. Hew), (V.-H. Lee),
been enormous especially with Air Asia promoting aggressively 81
(K.-B. Ooi).
0957-4174/ 2015 Published by Elsevier Ltd.

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

2 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

82 its tagline Now Everyone Can Fly. In fact, the airline sector has to ascertain the direct impact of the SERVPERF dimensions towards 94
83 been ascertained to be an intangible service industry by Clemes, full service and low cost airlines (LCC) customer satisfaction (CS) 95
84 Gan, Kao, and Choong (2008). A good service quality (SQ) will lead and customer loyalty (CL) using a Structural Equation 96
85 to increase in prots (Buzzell & Gale, 1989) while maintaining an Modeling (SEM)articial neural network predictive analytic 97
86 organizations competitive advantage (Park, Robertson, & Wu, approach. Predictive analytics are useful for generating new 98
87 2004). Therefore, the impact of SQ on airline customers theory, developing new measures, comparing competing 99
88 satisfaction and loyalty is worth studying as the ndings would theories, improving existing theories, assessing the relevance 100
89 be benecial to not only the two airlines but also other airlines of theories, and assessing the predictability of empirical 101
90 worldwide in increasing their passenger trafc as well as phenomena (Shmueli & Koppius, 2011, p. 554). Finally, there are 102
91 improving the protability. Majority of the previous studies hardly any pervious studies that examine whether differences 103
92 (Table 1) have focused on measuring airlines SQ using the exist in the SERVPERF provided between full service and low cost 104
93 SERVQUALs GAP-5 model. Perhaps this research is among the rst airline. 105

Table 1
Studies on service quality, customer satisfaction, loyalty and/or behavioral intention of airline industry.

Researcher(s) Country Sample size Unit of analysis Measurement Method of analysis Findings
model of service
Abdullah et al. Malaysia 391 Passenger at the Kuala SERVQUAL Factor analysis, Chi- SERVQUAL dimensions inuence
(2007) Lumpur International square test and CS, intention to re-patron and
Airport (KLIA) regression analysis likelihood to recommend to
Arifn et al. (2010) Malaysia 100 Passenger of low cost SERVQUAL Factor analysis and Caring and tangible was the
carriers terminal stepwise multiple most important dimension of
(LCCT) at KLIA regression analysis service quality followed by
reliability, responsiveness,
affordability and visual
attractiveness. Only caring and
tangible signicantly predict
customers satisfaction on
service quality
Atilgan, Akinci, and Turkey 235 Passenger at the SERVQUAL Paired t-tests and Gaps exist for the dimensions of
Aksoy (2008) Antalya International correspondence food and beverage, cabin,
Airport analysis aesthetics, convenience,
dependability, in-ight activities
and personnel
Chau and Kao (2009) Taiwan 263 Passenger at Taoyuan SERVQUAL Paired-sample t-tests The gap-5 sizes quality
International Airport and linear regression dimensions have signicant
and London Heathrow impact on CS and service value
International Airport
Clemes et al. (2008) New 428 International air SERVQUAL Multiple regression Assurance, comfort,
Zealand traveler analysis, ANOVA, t- convenience, timeliness,
tests helpfulness, meals, security and
safety are positively associated
to PSQ. PSQ is affected by
travelers gender, age,
occupation, marital status and
income. PSQ and price inuence
CS. CS has a strong inuence on
future BI
Cunningham et al. U.S. and 105 (U.S.) Student of an evening SERVPERF Regression analysis US: reliability, connections and
(2002) Korea MBA program using air in-ight comfort inuence CS.
transportation Intention to re-patronize is
affected by reliability and
145 (Korea) Korea: reliability, risk factors
and assurance affect CS.
Intention to re-patronize is
affected by overall risk and
Huang (2009) Taiwan 602 Passenger at Taoyuan SERVQUAL SEM Service quality has direct effects
International Airport on passengers behavioral
Ling and Lin (2005) Taiwan 262 (Taiwan) Traveler between SERVPEX SEM, ANOVA CS is inuenced by tangibles,
and China 142 (China) Taiwan and Mainland reliability, responsiveness and
China assurance but not empathy.
Customers preference is affected
by tangibles and reliability
Nadiri et al. (2008) Cyprus 583 North Cyprus national SERVQUAL SEM SERVQUAL dimensions
airline passenger positively inuence CS and
repurchase intention. CS

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 3

Table 1 (continued)

Researcher(s) Country Sample size Unit of analysis Measurement Method of analysis Findings
model of service
positively inuences word-of-
Park, Robertson, and Australia 501 Australian SERVQUAL SEM In-ight service and convenience
Wu (2005) international and accessibility have positive
passenger effect on airline image and
behavioral intention
Park, Robertson, and Australia 501 Australian SERVQUAL SEM Perceived price, perceived value,
Wu (2006) international passenger satisfaction and
passenger airline image have direct effect
on passengers future behavioral
Prayag (2007) Mauritius 140 Passenger onboard SERVQUAL Paired t-tests, factor Perceptions of service are
ight from Cape Town analysis and stepwise inuenced by service efciency
to Mauritius regression analysis and affect, service
personalization, reliability and
tangibles. CS and willingness to
recommend are dependent on
service efciency and affect
Saha and Theingi Thailand 1212 Passenger of low cost SERVPERF SEM Tangibles, ground staff, ight
(2009) carriers schedules and attendants
inuence CS. CS affects BI and
word-of-mouth. Word-of-mouth
affects BI.
Wong and Musa Malaysia 388 Passenger at KLIA and SERVQUAL Exploratory Factor Air Asia was perceived better in
(2011) LCCT-KLIA Analysis using price, publicity and word-of-
principle component mouth dimension while
analysis with Varimax Malaysia Airlines was perceived
rotation better in tangibles, core service,
reputation and employee

Note: PSQ = perceived service quality; CS = customer satisfaction; CL = customer loyalty; BI = behavioral intention (intention to re-purchase/re-patronize); SEM = Structural
Equation Modeling.

106 Hence, three research questions were formulated: loyalty from the context of low cost and full service airline passen- 135
gers. SERVPERF consists of 5 dimensions namely tangibles, reliabil- 136
107 RQ1: What are the dimensions of service quality performance of ity, responsiveness, assurance and empathy as shown in Fig. 1. 137
108 the airline industry?
109 RQ2: Is there any signicant difference between full service and
110 low cost airlines in terms of construct validity of the SERVPERF 2.2. Service quality (SQ) 138

111 dimensions and the causal relationships in the structural

112 model? Service quality (SQ) supports customers satisfaction while 139

113 RQ3: What is the impact of these service dimensions towards motivating recommendations and promoting intention to return 140

114 customer satisfaction of the airline industry? (Nadiri & Hussain, 2005). In return, customers satisfaction lead 141

115 to rise in protability, return of investment and market share 142

116 This study aims to examine the impacts of the SERVPERF (Legoherel, 1998; Stevens, Knutson, & Patton, 1995). No wonder, 143

117 dimensions on customer satisfaction towards loyalty in airlines SQ has been one of the hot topics of debate in the service literature 144

118 industry based on the above-mentioned research questions. (Ooi, Lin, Tan, & Chong, 2011). SQ is generally dened as when the 145

119 Hence, from the contextualization of the airline industry, there is delivered service is able to meet customers expectations, require- 146

120 no doubt that a study on the impacts of the SERVPERF dimensions ments and satisfaction. It is also dened as customers general 147

121 upon customer satisfaction and loyalty is indeed very much rele-
122 vant and imperative. The remainder of the paper is arranged as
123 such:

124 Presentation of the review of literature.

125 Development of the hypotheses.
126 Elucidation on research methodology.
127 Elaboration of data analysis and results.
128 Discussion of implications, conclusion, limitations and future
129 studies.

131 2. Literature review

132 2.1. Context for denitions

133 The SERVPERF model is utilized in measuring its impacts on

134 customer satisfaction and its indirect effects on the customer Fig. 1. Research model.

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

4 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

148 impression of the comparative superiority or inferiority of a service argued that CS is a determinant of SQ (Bolton & Drew, 1991; 211
149 or organization (Bitner & Hubbert, 1994). Parasuraman, Zeithaml, Oliver, 1980) and even some asserted that there is no relationship 212
150 and Berry (1985) dened SQ as the discrepancy between con- between SQ and CS (Fornell, 1992; Teas, 1993). Therefore, a study 213
151 sumers perceptions of services provided by a particular company on these causal relationships is indeed warranted to further ascer- 214
152 and their expectations about companies offering such services. tain these claims. 215
153 Parasuraman, Zeithaml, and Berry (1988) proposed a 22-item
154 instrument composing ve key elements in a SERVQUAL model. 2.4. Customer loyalty (CL) 216

155 These dimensions include (Parasuraman et al., 1988, p. 23):

Customer loyalty (CL) encompasses re-purchase intention, will- 217

156 a. Tangibles the physical environment embodied by objects ingness to suggest to others a product or service and reluctance to 218

157 (i.e. interior design) and subjects (i.e. the look of employees). change to another competitor (Cronin & Taylor, 1992). It can be 219

158 b. Reliability the service providers capability to offer precise manifested in behaviors such as re-purchase intention (Cronin & 220

159 and trustworthy services. Taylor, 1992; Fornell, 1992; Parasuraman et al., 1991a) or word- 221

160 c. Responsiveness a rms readiness to help its customers by of-mouth (Cronin & Taylor, 1992; Selnes, 1993). In order to retain 222

161 providing swift and efcient service performance. CL and competitive advantage, it is essential that airline companies 223

162 d. Assurance miscellaneous features that give condence to assess and revise their strategies from time to time (Abdullah, 224

163 customers (i.e. the rms particular service knowledge, Manaf, & Noor, 2007). For example, Natalisa and Subroto (2003) 225

164 courteous and trustworthy behavior of employees). recommended that domestic airline companies should honor pro- 226

165 e. Empathy the service rms willingness to offer each mises in their promotional campaign. Chin (2002) stressed that CL 227

166 customer with individual service. can be raised through frequent ier programs; while Dick and Basu 228

167 (1994) asserted that dependability and assurance will uplift CL. 229

168 The SERVQUAL gap model asserts that the customers evalua- Although CL is imperative for the survival of service companies, 230

169 tion of the general SQ is ascertained by the level and direction of it needs to be studied further (Gremler & Brown, 1996). 231

170 the gap between what is expected and what is perceived of the
171 actual performance. One of the rst few articles published on air- 3. Hypothesis development 232

172 line SQ were in 1988 (Gourdin, 1988); while the direct application
173 of the SQ gap model emerged in 1991 (Fick & Ritchie, 1991; 3.1. Relationship between service quality and customer satisfaction 233

174 Gourdin & Kloppenborg, 1991). Since then, many researchers

175 (Table 1) have embarked on using this scale in airline industry. The correlation between SQ and CS has been widely debated. As 234

176 Besides that, SERVQUAL model is also widely applied in various mentioned above, some researchers claimed that SQ leads to CS 235

177 other sectors such as telecommunication (Oh, 1995), retail stores while others either disagreed or argued that there is no relation- 236

178 (Lee & Lee, 1997), nancial services (Newman, 2001), health care ship between these constructs. Conventionally, SQ and CS have 237

179 (Kilbourne, Duffy, Duffy, & Giarchi, 2004), hotels (Juwaheer, been deemed equated (Saha & Theingi, 2009). SQ is assessed by a 238

180 2004) and public sector (Wisniewski, 2001). denite performance of a service with regard to a specic service 239

181 However, the SERVQUAL model has been criticized by Carman attribute; while CS is evaluated by customers general experience 240

182 (1990, p. 44) who suggested that it is better to collect data in of a service (Oliver, 1993). To attain a higher level of CS, majority 241

183 terms of the perception/expectation difference directly rather than of the researchers recommended that a high degree of SQ ought 242

184 to ask about each separately. It is also important to take into to be delivered since it is considered as a determinant of CS 243

185 account the level of experience of the customer with the service. (Cronin, Michael, Brady, & Hult, 2000; Cronin & Taylor, 1992). 244

186 On the other hand, the SERVPERF model was introduced by Cronin Even though most researchers agree on the distinctiveness of the 245

187 and Taylor (1992) with the argument that SQ ought to be gauged as denitions for both constructs, the causal relationship is yet to 246

188 an attitude and supported the use of perception statements only in be ascertained (Saha & Theingi, 2009). 247

189 the SQ measurement. Their ndings revealed that SERVPERF

3.1.1. Tangibles 248
190 explained more variation in the measurement of SQ in all the four
Parasuraman, Berry, and Zeithaml (1991b, p. 41) dened tangi- 249
191 service industries i.e. dry cleaning, banks, fast food services, and
bles as the appearance of physical facilities, equipment, personnel 250
192 pest control. Hence, SERVPERF is employed in this study to assess
and communication materials. Aghdaie and Faghani (2012), 251
193 the airline SQ.
Nadiri, Hussain, Ekiz, and Erdogan (2008), Kim and Lee (2011) 252
and Lee, Lee, and Yoou (2000) discovered that tangibles may give 253
194 2.3. Customer satisfaction (CS)
signicant positive impact on customer satisfaction. In the context 254
of airline industry, the tangible dimensions refer to physical condi- 255
195 Customer satisfaction (CS) may be dened as an individuals
tion of the carrier, the facilities and equipment installed, the 256
196 perception either discontent or pleasure by comparing the per-
appearance of the cabin crews, the food and beverages as well as 257
197 ceived performance of a product with respect to ones expectations
reading entertainment materials provided on ight. If these tangi- 258
198 (Lin, Chen, & Chiu, 2010). Anderson, Fornell, and Lehmann (1994)
bles are inadequate, unavailable or in poor conditions, then most 259
199 dened CS as the overall evaluation based on the total purchase
likely customers will not feel satised with the service of airline 260
200 and consumption experience with a good or service over time.
carrier. Hence, the hypothesis is formulated: 261
201 Theoretically, SQ can also be described as the outcome of a
202 subjective evaluation between what is expected and what is per-
H1. Tangibles have signicant and positive inuence on customer 262
203 ceived for post-purchase accomplishments (Fecikova, 2004). SQ
satisfaction. 263
204 motivates re-purchase intention (Cronin & Taylor, 1992; Fornell,
205 1992; Parasuraman, Berry, & Zeithaml, 1991a) and positive
206 word-of-mouth (Rogerson, 1983) while retaining customers 3.1.2. Reliability 264
207 (Fornell, 1992) and securing CL (Selnes, 1993). Nevertheless, the Reliability is described as the ability to perform the promised 265
208 causal relationship between SQ and CS has been constantly service dependably and accurately (Parasuraman et al., 1991b, p. 266
209 debated with some researchers claiming that SQ leads to CS 41). Reliability has been found to have signicant inuence on cus- 267
210 (Cronin & Taylor, 1992; Parasuraman et al., 1988) whilst others tomer satisfaction (Aghdaie & Faghani, 2012; Lee et al., 2000). In 268

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 5

269 terms of airline industry, reliability dimension is referred to as the (Dean, 2007, p. 163). CL maybe manifested in the form of re-pur- 324
270 degree to which a carrier is able to provide precise and trustable chases intention, word-of-mouth (WOM) communication and rec- 325
271 services as promised to the customers. This includes ight sched- ommendations made of the company (Nadiri et al., 2008). WOM is 326
272 ule punctuality, accurate and up-to-date ight information, lug- dened as the informal communication between consumers 327
273 gage safety and delivery, personal safety and security, about the characteristics of a business or a product (Kau & Loh, 328
274 contingency plans, etc. If the level of reliability is high, customers 2006, p. 103). In the context of the airline industry, re-purchase 329
275 will denitely be satised with the airline service. Therefore, the intention refers to customers plans to y in their next trip. 330
276 hypothesis proposed is: Studies have shown that higher the CS, higher the re-purchase 331
and WOM communication (Davidow, 2003; Ekiz & Arasli, 2007). 332
277 H2. There is a signicant and positive inuence of reliability on Therefore, the following hypothesis is put forward: 333
278 customer satisfaction.
H6. Customer satisfaction has a signicant and positive inuence 334
on CL. 335
279 3.1.3. Responsiveness
280 Responsiveness may be described as the willingness to help
281 customers and provide prompt service (Parasuraman et al., 4. Research methodology 336
282 1991b, p. 41). Responsiveness is a signicant antecedent of cus-
283 tomer satisfaction (Aghdaie & Faghani, 2012; Kim & Lee, 2011; 4.1. Sample and collection procedures 337
284 Lee et al., 2000). In the airline context, responsiveness refers to
285 the promptness of action taken whenever there are requests or Initially, a pre-test was carried out with two professors to ascer- 338
286 complaints from the customers. Immediate action must be taken tain the face and content validity of the survey questionnaire. Then 339
287 in order to provide the service assistance and help needed. If the a pilot test was conducted with 30 passengers to evaluate the 340
288 response time is poor then the service is deemed to be unsatisfac- questionnaire in terms of wording, clarity, relevancy and time 341
289 tory and will leave a negative impact on the customers satisfac- spent. Finally, a cross-sectional survey questionnaire was adminis- 342
290 tion. Hence, the next hypothesis is as follow: tered to 350 passengers at the Kuala Lumpur International Airport 343
(KLIA) for duration of two weeks via convenience sampling tech- 344
291 H3. Responsiveness has a signicant and positive inuence on nique. Out of these, 50 questionnaires were unusable as they were 345
292 customer satisfaction. incomplete. Hence the response rate is 85.7% using the similar 346
computation method as Cheng, Chen, and Chang (2008) and 347
Chen and Chang (2005). Convenient sample was engaged as there 348
293 3.1.4. Assurance
is no sampling frame available and the passengers in KLIA are most 349
294 Assurance can be described as the knowledge and courtesy of
likely the customers of Air-Asia or MAS. Furthermore, it is also in 350
295 employees and their ability to convey trust and condence
accordance with the approach used by Cheng et al. (2008), Chen 351
296 (Parasuraman et al., 1991b, p. 41). Previous studies (Clemes et al.,
and Chang (2005) and Lindenmeier and Tscheulin (2008). Besides 352
297 2008; Lee et al., 2000; Yi & La, 2003) have found that assurance
that, according to the airport, the prole of the respondents is quite 353
298 poses a signicant positive impact on customer satisfaction. In
similar with the frequent yers of the airport in terms of gender, 354
299 the airline perspective, it is referred as the ability of a carrier to
age, marital status and occupation (Chen & Chang, 2005). 355
300 provide trust and condence to the customers. Moreover, the capa-
301 bility to exhibit credibility and courtesy will play a vital role in
302 building customers condence and trust. This leads to the forma- 4.2. Variable measurements 356
303 tion of the next hypothesis:
4.2.1. Independent variable: SERVPERF measures 357
304 H4. Assurance has a signicant and positive inuence on customer To ascertain construct reliability and validity of the indepen- 358
305 satisfaction. dent variable (SERVPERF dimension), 22 items were adapted from 359
past studies shown in Table 2. A 7-point Likert scale was employed 360
to measure the level of agreement for every item starting from 361
306 3.1.5. Empathy strongly disagree (1) to strongly agree (7). 362
307 Empathy is referred to as the caring and individualized atten-
308 tion provided to the customer (Parasuraman et al., 1991b, p. 41).
4.2.2. Dependent variables: customer satisfaction and customer loyalty 363
309 This dimension is referred to the carriers ability to provide individ-
Two dependent variables namely CS and CL were measured 364
310 ualized attention to the customers. Empathy has been found to
with a 7-point Likert scale starting with 1 for strongly disagree 365
311 have signicant effect on customer satisfaction (Aghdaie &
to 7 for strongly agree. There were 4 items for CS and 4 items for 366
312 Faghani, 2012; Clemes et al., 2008; Lee et al., 2000; Suki, 2014).
CL which were all adapted from previous studies to guarantee con- 367
313 A cabin crew who has a high degree of empathy will have the abil-
struct reliability and validity. 368
314 ity to understand the feelings, needs and wants of their customers.
315 Customers will denitely feel appreciated and thus establishing a
316 good interpersonal relationship. Personal caring attention will cre- 4.3. Data analysis 369

317 ate a comfortable environment for customers during the ight.

318 Thus, we would put forward the following hypothesis: 4.3.1. Prole of respondents 370
Table 3 shows that the sample respondent consists of 49.7% 371

319 H5. Empathy has a signicant positive inuence on customer males and 50.3% females. 36.3% aged below 26, 36.3% aged 372

320 satisfaction. between 26 and 35, 16.3% aged between 36 and 45 and 11.0% aged 373
above 45. 60.3% of them are single, 33.7% married, 3.3% divorced 374
and 2.7% are widow or widower. 14.3% of them work with the gov- 375
321 3.1.6. Customer satisfaction and customer loyalty ernment, 29.3% with the private companies, 17.0% self-employed, 376
322 CL is dened as the degree to which a customer recommends, 21.0% are students, 6.0% are housewives, 8.0% are professionals 377
323 and expresses a preference for future use of a particular company and 4.3% retired. Salary wise, 31.3% earn below RM1000, 17.3% 378

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

6 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

Table 2 earn between RM1001 and RM2500, 36% earn between RM2501 379
Sources of adapted constructs. and RM4000 and 15.3% earn more than RM4000. 380
Constructs Number Sources Table 4 shows that in terms of customer segmentation, 66.3% 381
of items belong to Air Asia, 31% belong to Malaysia Airlines (MAS) and 382
Tangibles (TA) 4 Kumar, Mani, Mahalingam, and Vanjikovan 2.7% others. 74.0% of them y less than 5 times annually, 21% y 383
(2010) and Parasuraman, Berry, and Zeithaml between 6 and 10 times annually and 5.0% y for more than 10 384
(1991c) times annually. 13.7% are on business trips, 20.7% visiting friends 385
Reliability (RL. 5 Kumar et al. (2010) and Parasuraman et al.
or relatives, 63.3% are tourists and 2.3% travel for other purposes. 386

Responsiveness 4 Kumar et al. (2010) and Parasuraman et al. 46.7% paid their tickets personally, 32.3% sponsored by family or 387
(RP) (1991c) friend and the rest either by secretary or other sources. 388
Assurance (AS) 4 Kumar et al. (2010) and Parasuraman et al.
4.4. Statistical analysis 389
Empathy (EM) 5 Kumar et al. (2010) and Parasuraman et al.
Customer 3 Olorunniwo and Hsu (2006) and Parasuraman To examine the causal relationship between the constructs, 390
satisfaction et al. (1991b) Structural Equation Modeling (SEM) method with Maximum 391
(CS) Likelihood Estimation (MLE) was performed and analyzed with 392
Customer 4 Ismail, Haron, Ibrahim, and Mohd Isa (2006)
AMOS 18. A two-stage procedure was applied to the measurement 393
loyalty (CL) and Parasuraman et al. (1991b)
(CFA) model and the structural model as recommended by Hair, 394
Black, Babin, and Anderson (2010), Leong, Hew, Tan, and Ooi 395
(2013), Leong, Ooi, Chong, and Lin (2013) and Wang, Tsai, and 396
Tsai (2014). The rst stage involves the Exploratory Factor 397
Table 3
Demographic prole of the respondents. Analysis (EFA) that identies the underlying constructs, which is 398
then followed by the second stage of conrmatory factor analysis 399
Variables Frequency Percentage
(CFA) to ascertain the causal relationship between these constructs 400
Gender Male 149 49.7 via path analysis (Kaplanidou & Vogt, 2006; Liu et al., 2014; 401
Female 151 50.3
McGehee, Lee, OBannon, & Perdue, 2010; Nunkoo, Ramkissoon, & 402
Age <26 109 36.3 Gursoy, 2013; Palmer, Koenig-Lewis, & Medi Jones, 2013; 403
2635 109 36.3
Ramkissoon, Graham Smith, & Weiler, 2013; So, King, Sparks, & 404
3645 49 16.3
>45 33 11.0
Wang, 2014; Swanson & Horridge, 2004; Zopiatis, Constanti, & 405
Theocharous, 2014). 406
Marital status Single 181 60.3
Married 101 33.7
Divorced 10 3.3 4.4.1. Testing the assumptions of multivariate analysis 407
Widow/widower 8 2.7 Several multivariate assumptions are required in an SEM anal- 408
Occupation Government 43 14.3 ysis. These include normality of dataset, multicollinearity, linear- 409
Private 88 29.3 ity, outliers, bias, construct reliability and validity, adequate 410
Self-employed 51 17.0 sample size, convergent and discriminant validity. As the sample 411
Student 63 21.0
size for this research is 300, it is adequate for the SEM analysis 412
Housewife 18 6.0
Professional 24 8.0 (Hair et al., 2010). Common latent factor with AMOS 18 revealed 413
Retired 13 4.3 a shared variance of 2.56% implies non-existence of common 414

Individual monthly income <RM1000 94 31.3 method variance whereas the Mahalanobis d2/independent vari- 415
RM1001RM2500 52 17.3 ables <3 indicates non-existence of multivariate outliers (Lee, 416
RM2501RM4000 108 36.0 Ooi, Tan, & Chong, 2011). The maximum absolute values of skew- 417
>RM4000 46 15.3
ness and kurtosis are 0.689 (<1) and 0.434 (<2) respectively 418
Note: RM = Ringgit Malaysia (Malaysian currency unit). (Kline, 2005; Leong, Hew, Ooi, & Lin, 2011; Tan, Siah, Ooi, Hew, & 419
Chong, 2015). The histograms, normal PP and QQ plots of regres- 420
sion standardized residual further veried the existence of normal- 421
ity while scatter plots veried linearity (Tan, Ooi, Chong, & Hew, 422
Table 4 2014). The largest Variance Ination Factor (VIF) is 3.547 (<10) 423
Distribution of airline usage, travel purpose and airline selection.
and the smallest tolerance is 2.282 (>0.1). All correlation coef- 424
Variables Frequency Percentage cients are below 0.8 indicating no occurrence of multicollinearity 425

Frequently used airline Air Asia 199 66.3 (Kline, 2005; Teo, Tan, Ooi, Hew, & Yew, 2015). 426
Malaysia Airlines 93 31.0
(MAS) 4.4.2. Construct unidimensionality and reliability 427
Others 8 2.7
To examine the unidimensionality of every construct, 428
Average use of airline <5 times a year 222 74.0 Goodness-of-Fit Index (GFI) for each construct is evaluated. 429
services 610 times a year 63 21.0
Table 5 shows that GFI values ranged from 0.967 to 1.000 which 430
>10 times a year 15 5.0
is greater than the 0.90 threshold (Prajogo & Hong, 2008) indicat- 431
Purpose of travel Business 41 13.7
ing a good unidimensionality of the constructs. The construct reli- 432
Visiting friends/ 62 20.7
relatives ability (alpha) ranges from 0.850 to 0.898 were greater than the 433
Tourist 190 63.3 recommended threshold of 0.70 (Nunnally & Bernstein, 1994). 434
Others 7 2.3 Composite reliability (CR) which uses the actual factor loadings 435
Who helps selecting Yourself 140 46.7 instead of the equal weight for all the constructs was also calcu- 436
airline Secretary 20 6.7 lated. The CR values ranged from 0.862 to 0.894 with all the factor 437
Travel agent 40 13.3 loadings signicant at p < 0.001. Besides that Average Variance 438
Family/friend 97 32.3
Extracted (AVE) was also calculated with value ranging from 439
Others 3 1.0
0.602 to 0.726. 440

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 7

Table 5
Construct unidimensionality, reliability and convergent validity.

First order latent constructs First order Second order

Indicator Standardized Standardized Unidimensionality Reliability (aa,
loading loading (GFI) CRb; AVEc)
Tangibles (TA)
The airline that I choose has clean and comfortable interior/seat TA1 0.841 0.772 0.999 a = 0.850
The airline that I choose provide sufcient in-ight facilities, e.g. entertainment, TA2 0.743 CR = 0.862
phone, etc.
The frontline employees of the airline that I choose are neat appearing TA3 0.734 AVE = 0.610
Materials associated with the service are visually appealing at the airline that I TA4 0.800
choose, e.g. newspaper, magazines, guide map, etc.
Reliability (RL)
When the airline that I choose promises to complete customers specic task by a RL1 0.773 0.855 0.997 a = 0.898
certain time, they do
When I have a problem, the airline that I choose shows sincere interest in solving RL2 0.779 CR = 0.894
The airline that I choose performs the service right at the rst time RL3 0.812 AVE = 0.628
The airline that I choose provides the service at the time it promises to do so RL4 0.796
The airline that I choose delivers error free records RL5 0.800
Responsiveness (RP)
Employees of the airline that I choose tell me exactly when the service will be RP1 0.663 0.855 0.990 a = 0.857
performed, e.g. meal time
Employees of the airline that I choose give prompt service RP2 0.833 CR = 0.862
Employees of the airline that I choose are always willing to help me RP3 0.870 AVE = 0.612
Employees of the airline that I choose are never too busy to respond to my RP4 0.747
Assurance (AS)
The behavior of employees in the airline that I choose instill condence to me AS1 0.850 0.891 0.983 a = 0.865
I feel safe in the airline that I choose AS2 0.766 CR = 0.862
Employees of the airline that I choose are consistently courteous with me AS3 0.760 AVE = 0.611
Employees of the airline that I choose have the knowledge to answer my AS4 0.745
Empathy (EM)
The airline that I choose gives me individual attention EM1 0.760 0.842 0.967 a = 0.866
The airline that I choose has provide online service that convenient to all its EM2 0.631 CR = 0.882
The airline that I choose has employees who serve me in a polite manner EM3 0.803 AVE = 0.602
The airline that I choose has shown me best interest at heart EM4 0.812
The employees of the airline that I choose understands my specic need EM5 0.855
Customer satisfaction (CS)
I am satised with my decision to use this airline CS1 0.876 1.000 a = 0.888
My choice to use this airline was a nice one CS2 0.852 CR = 0.888
I feel that my experience with this airline has been enjoyable CS4 0.828 AVE = 0.726
Customer loyalty (CL)
I say positive things about the airline to other people CL1 0.854 0.999 a = 0.891
I intended to continue being a client of the airline for a long time to come CL2 0.818 CR = 0.887
I will encourage friends and relatives to use the services offered by the airline CL3 0.832 AVE = 0.663
To me, the airline clearly is able to provide the best services CL4 0.750

Note: CS3 was discarded due to poor standardized loading.

a = Cronbachs alpha.
CR = (Rki)2/[(Rki)2 + Rdi)], (ki = standardized factor loadings, i = observed variables, di = error variance).
AVE = Rki2/n (i = 1 . . . n, k = standardized factor loadings, i = observed variables).

Table 6
Discriminant validity: correlation coefcients and Chi-square difference test.

Construct correlations and square root of AVE

TA 0.781
RL 0.741 0.792
RP 0.669 0.737 0.782
(176.661) (236.434)
AS 0.692 0.766 0.751 0.781
(207.426) (268.652) (250.192)
EM 0.633 0.709 0.728 0.758 0.776
(145.676) (70.417) (232.758) (275.169)
CS 0.713 0.725 0.695 0.680 0.648 0.852
(208.677) (213.562) (201.788) (184.830) (157.684)

Note: n = 300; ***p < 0.001; **p < 0.01; square root of AVE (in bold) is shown on the diagonal; Off-diagonal elements are the correlation coefcients and values in the brackets
show the chi-square difference statistics Dv2 with df = 1; TA = tangibles; RL = reliability; RP = responsiveness; AS = assurance; EM = empathy; CS = customer satisfaction.

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

8 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

Table 7
Measures of the model t.

Goodness of t measures v2 test statistics/df p-Value GFI AGFI CFI NFI IFI TLI RMSEA SRMR
a a a a a a b b c
Recommended value 63.00 >0.05 P0.90 P0.90 P0.90 P0.90 P0.90 P0.90 60.08 60.1d
CFA model 0.680 0.606 0.997 0.987 1.000 0.998 1.000 1.000 0.000 0.0074
Structural model 0.697 0.675 0.995 0.982 1.000 0.997 1.000 1.000 0.000 0.0067

Sources: GFI = Goodness-of-Fit Index; AGFI = Adjusted Goodness-of-Fit Index; CFI = Comparative Fit Index; NFI = Normed Fit Index; IFI = Incremental Fit Index; TLI = Tucker-
Lewis Index; RMSEA = Root Mean Square Error of Approximation; SRMR = Standardized Root Mean Residual.
Bagozzi and Yi (1988) and Hair et al. (2010).
Arbuckle (2008), Byrne (2001) and Hair et al. (2010).
Browne and Cudeck (1993).
Hoang, Igel, and Laosirihongthong (2006).

441 4.4.3. Measurement model conrmatory factor analysis NFI = 0.998; IFI = 1.000; TLI = 1.000; RMSEA = 0.000; SRMR = 477
442 Construct validity measures the level to which a scale mea- 0.0074) exceeded the recommended thresholds (Hu & Bentler, 478
443 sures what it intends to measure (Hew & Leong, 2011, p. 21). 1999). Thus, it may be concluded that the measurement model is 479
444 During the CFA stage, convergent and discriminant validity were able to t very well with the dataset. 480
445 veried (Chen & Chen, 2014; Lee, 2013). Convergent validity
446 describes the ability of the construct to measure what it is sup-
447 posed to measure. Fornell and Larcker (1981) asserted that conver- 4.4.4. Comparison of group invariance between low cost and full 481
448 gent validity is attained if all the AVEs are greater than 0.50, all the service airlines 482
449 CRs are greater than 0.70 and all the factor loadings are signicant. The comparison between the low cost and full service airlines is 483
450 As shown in Table 5 all these criteria were satised, indicating the done based on results of the multiple group analysis (MGA). In 484
451 constructs in the battery have successfully passed the convergent order to perform MGA, chi-square test was used and a signicance 485
452 validity test. Discriminant validity, on the other hand, describes in the difference of the chi-square value between the congural 486
453 the level of the construct and its indicators to be different from model (unconstrained) and the constrained model (where all paths 487
454 other constructs and their respective indicators. Kline (1998) sug- were constrained to equal for both groups) implies existence of dif- 488
455 gested that to examine discriminant validity, the correlation coef- ferences between the two groups (Byrne, 2010; Feng, Prajogo, Tan, 489
456 cients should range between 0.10 and 0.90. Table 6 shows that & Sohal, 2006). To establish whether a signicant difference exists, 490
457 this criterion was fullled. Discriminant validity is obtained as the difference in chi-square value Dv2 and degree of freedom 491
458 the AVE of a construct is larger than the square root of its correla- Ddf is compared to the critical value of the chi-square test at 492
459 tion coefcients with other constructs (Leong, Hew, Ooi, & Lin, 0.05 level v20:05 . If it is larger than this critical value then there 493
460 2012; Teo, Cheah, Leong, Hew, & Shum, 2012) as shown in is signicant difference between the groups. Table 8 indicated no 494
461 Table 6. Furthermore, chi-square difference test (Zait & Bertea, signicant difference exists between the full service and low cost 495
462 2011) was also performed and the result showed that all constructs airlines for all the constructs of the SERVPERF model except for 496
463 were signicantly different. Hence, for the measurement model, the reliability dimension. This indicates that most of the constructs 497
464 both the reliability and validity of the constructs have been suc- of the SERVPERF model are robust across both full service and low 498
465 cessfully veried. cost airlines. Table 9 further veries the group invariance of the 499
466 Continuing from the rst order measurement model, a second measurement model of the SERVPERF dimension and thus there 500
467 order measurement model based on the mean values of the indica- are no differences in terms of construct validity across the two 501
468 tors of each construct was also developed. To assess the tness of groups. In terms of causal relationship difference in the structural 502
469 the measurement model with the dataset, ten goodness of t mea- model across the groups, although the chi-square difference test 503
470 sures were used namely the normed chi-square, p-value, (Table 9) indicated a signicant difference but upon closer inspec- 504
471 Goodness-of-Fit Index (GFI), Adjusted Goodness-of-Fit Index tion, it was found that these differences were not on the six 505
472 (AGFI), Comparative Fit Index (CFI), Normed Fit Index (NFI), hypothesized paths whereby all the critical ratios for differences 506
473 Incremental Fit Index (IFI), Tucker-Lewis Index (TLI), Root Mean (Table 10) for these paths have absolute values less than 1.960 at 507
474 Square Error of Approximation (RMSEA) and Standardized Root p < 0.05 (Byrne, 2010). Hence, a generalization can be made that 508
475 Mean Residual (SRMR). As illustrated in Table 7, all these measures no signicant differences exist in terms of the causal relationships 509
476 (v2/df = 0.680, p = 0.606; GFI = 0.997; AGFI = 0.987; CFI = 1.000; between the two groups. 510

Table 8
Comparison of measurement model of the SERVPERF dimensions between low cost and full service airline.

SERVPERF dimensions Original model without imposing equality Model with Difference v20:05 critical value Group variance
constrains equality between
constrains original and
imposed constrained
v2MASFS (df) v2Air AsiaLC (df) v2 df v2 df Dv2 Ddf

Tangibles 0.145 (1) 0.913 (1) 1.058 2 4.321 6 3.263 4 9.488 NS

Reliability 2.316 (4) 1.448 (4) 3.764 8 21.001 13 17.237 5 11.070 S
Responsiveness 3.331 (2) 5.343 (2) 8.674 4 9.757 7 1.083 3 7.815 NS
Assurance 0.624 (2) 2.299 (1) 2.923 3 7.441 6 4.518 3 7.815 NS
Empathy 4.496 (3) 2.972 (2) 7.468 5 15.708 11 8.240 6 12.592 NS

Note: NS not signicant; S signicant.

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 9

Table 9
Comparison of the structural model of SERVPERF dimensions between low cost and full service airline.

Original model without imposing equality Model with Difference v20:05 critical value Group variance
constrains equality between
constrains original and
imposed constrained
v2MASFS (df) v2Air AsiaLC (df) v2 df v2 df Dv2 Ddf

Measurement model 3.905 (4) 1.168 (4) 5.073 8 16.772 14 11.699 6 12.592 NS
Structural model 4.389 (7) 5.214 (7) 9.603 14 40.576 28 30.973 14 23.685 S

Note: FS full service carrier; LC low cost carrier; NS not signicant; S signicant.

Table 10
Critical ratios for differences between parameters (structural model).

Air Asia (LC)

MAS (FS) RL ? CS 1.273 0.801 0.298 1.378 0.300 6.138
AS ? CS 1.130 0.887 0.181 1.222 0.403 5.803
RP ? CS 0.576 2.871 1.681 0.579 2.425 4.876
TA ? CS 1.207 0.925 0.203 1.313 0.416 6.252
EM ? CS 1.931 0.394 0.841 2.106 0.187 7.707
CS ? CL 5.319 7.476 6.411 5.662 7.298 0.767

Note: FS full service carrier; LC low cost carrier; TA = tangibles; RL = reliability; RP = responsiveness; AS = assurance; EM = empathy; CS = customer satisfaction;
CL = customer loyalty.

Table 11
Results of path analysis.

Hypothesis Path Estimate Std. Critical p- Remarks

Error ratio Value
H1 TA ? CS 0.304 0.058 5.277 0.000 Supported
H2 RL ? CS 0.271 0.067 4.066 0.000 Supported
H3 RP ? CS 0.247 0.064 3.854 0.000 Supported
H4 AS ? CS 0.051 0.071 0.718 0.473 Not
H5 EM ? CS 0.099 0.060 1.648 0.099 Not
H6 CS ? CL 0.924 0.042 21.815 0.000 Supported
Note: p < 0.001; TA = tangibles; RL = reliability; RP = responsiveness;
AS = assurance; EM = empathy; CS = customer satisfaction; CL = customer loyalty.

explained by the model (R2). The results implied that 63.1% of 523
the variance in CS is explained by the relevant SERVPERF dimen- 524
sion while CS is able to explain 55.6% of the variance in CL. As 525
demonstrated in Table 11, the ndings further reveal that tangibles 526
(b = 0.304, p < 0.001), reliability (b = 0.271, p < 0.001) and respon- 527
siveness (b = 0.247, p < 0.001) have signicant and positive impact 528

Fig. 2. SEM analysis result. on CS while CS (b = 0.924, p < 0.001) positively impacted on CL. 529
Therefore, H1, H2, H3 and H6 were supported. 530
However, assurance (b = 0.051, p = 0.473) and empathy 531
511 4.4.5. Structural model
(b = 0.099, p = 0.099) did not show signicant impact on CS. 532
512 Similar to the measurement model, ten goodness-of-t mea-
Hence, H4 and H5 were not supported. Besides that, based on the 533
513 sures were used to gauge the goodness-of-t of the structural
two-tailed signicance of the bias-corrected percentile method 534
514 model. As reported in Table 7, all the measures (v2/df = 0.697,
using bootstrapping (Table 12) there were indirect effects of tangi- 535
515 p = 0.675; GFI = 0.995; AGFI = 0.982; CFI = 1.000; NFI = 0.997;
bles (0.369), reliability (0.315) and responsiveness (0.276) on CL. 536
516 IFI = 1.000; TLI = 1.000; RMSEA = 0.000; SRMR = 0.0067) exceeded
This indicates that CS mediates the causal relationship between 537
517 the recommended thresholds. Hence, the structural model was
tangibles, reliability, responsiveness and CL. However, there were 538
518 able to t very well with the collected data.
no signicant indirect effects of empathy and assurance on CL. 539

519 4.5. Hypothesis testing

4.6. Articial neural network analysis 540
520 In the path analysis, signicance of a path is determined based
521 on its p-value. Fig. 2 shows the standardized path coefcient (b) Although SEM has been commonly applied in verifying hypoth- 541
522 with p-value in bracket together with the corresponding variance esized causal relationships, it has been rarely combined with other 542

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

10 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

Table 12 possibilities of over-simplifying the complexities in decision mak- 587

Standardized indirect effects with bootstrap signicance. ing processes (Sim, Tan, Wong, Ooi, & Hew, 2014), the use of the 588
EM AS RP RL TA SEMANN approach in this study would complement each other. 589

CL 0.115 0.055 0.276

0.369 Hence, in this study, the signicant determinants from the SEM 590
(0.194) (0.604) (0.005) (0.016) (0.004) analysis were employed as the input variables for the ANN 591
analysis. 592
Note: Value in bracket indicates two-tailed signicance using bias-corrected per-
centile method with bootstrap; *p < 0.05; **p < 0.01; TA = tangibles; RL = reliability;
For this study, we have engaged multi-layer perceptron (MLP) 593
RP = responsiveness; AS = assurance; EM = empathy; CS = customer satisfaction; with Feedforward-Back Propagation (FFBP) algorithm where the 594
CL = customer loyalty. input signals are fed in forward direction while the error signals 595
propagate in backward direction. The MLP consists of three layers 596
i.e. input, hidden and output layers. 597
In order to avoid over-tting, a ten-fold cross validation with 598
543 expert systems or articial intelligence algorithms (Chong, 2013a;
data partition of 90:10 for training and testing was conducted 599
544 Hsu, Shih, Huang, Lin, & Lin, 2009; Wong, Law, & Yau, 2011). Since
using SPSS 21. The number of hidden neurons was generated auto- 600
545 SEM is only able to examine linear relationships, it may sometimes
matically and the sigmoid function was used for both output and 601
546 oversimplify the complexities involved in the human decision
hidden layers (Tan, Ooi, Leong, & Lin, 2014). The relevance of the 602
547 making processes. To properly address this issue, ANN is incorpo-
predictor variables was validated based on the number of non-zero 603
548 rated in identifying the non-compensatory and non-linear relation-
synaptic weight connected to the hidden layer (Table 13). 604
549 ships in the research model as it is capable to learn complex linear
Root Mean Square of Error (RMSE) was used in assessing the 605
550 and non-linear relationships between SERVPERF dimensions and
accuracy of the NN model while the normalized importance for 606
551 customer satisfaction. Previous linear and compensatory model
each predictor was calculated in the sensitivity analysis 607
552 has the likelihood to assume that the shortfall in one of the predic-
(Table 14). The mean RMSE values range from 0.0742 to 0.0819, 608
553 tors can be compensated by improving other predictors. For exam-
which indicate a high predictive accuracy. Based on the sensitivity 609
554 ple, the shortfall in responsiveness towards customer satisfaction
analysis, responsiveness was reported to be the most essential 610
555 may be compensated by the improved reliability, empathy or other
determinant of customer satisfaction followed by tangibles and 611
556 SERVPERF dimensions. However, compensatory model may not be
reliability. 612
557 applicable to the SERVPERF dimensions as these dimensions are
The use of the two-stage predictive-analytic (Scott & Walczak, 613
558 entirely different and have different impacts towards customer
2009; Shmueli & Koppius, 2010) SEMneural network analysis 614
559 satisfaction. Through the identication of non-linear relationship,
may provide a more holistic understanding and thus may provide 615
560 ANN will be able to examine the non-compensatory effects of
signicant methodological contribution from the statistical point 616
561 SERVPERF on customer satisfaction. Besides that, ANN is capable
of view. This is because the non-compensatory neural network 617
562 of producing more accurate predictions compared to the tradi-
analysis is able to complement the weaknesses of the compen- 618
563 tional regression techniques such as SEM, logistic, multiple and
satory and linear SEM analysis. The study will open up another 619
564 discriminant regression (Chiang, Zhang, & Zhou, 2006; Chong,
new perspective in understanding the impact of SERVPERF on cus- 620
565 2013a; Leong, Hew, et al., 2013; Morris, Greer, Hughes, & Clark,
tomer satisfaction between low cost and full service airlines based 621
566 2004).
on the results of the multiple group analysis. This is one of the sig- 622
567 Generally, articial neural network (ANN) refers to a device that
nicant additional contributions to the existing literature in the 623
568 is made to model the manner in which human brain conducts a
sense that expert system or articial intelligence was applied in 624
569 specic function or job through the use of electronic components
a new context of service quality on customer satisfaction of low 625
570 or computer software simulation. ANN is described as a massively
cost and full service airlines. This is different from the contexts 626
571 parallel distributed processor made up of simple processing units,
of previous studies on m-commerce adoption (Chong, 2013a), 627
572 which has a natural propensity for storing experiential knowledge
inter-organizational system standard adoption (Chan & Chong, 628
573 and making it available for use (Haykin, 1999, p. 24). ANN is
2012), open IOS adoption (Chong & Bai, 2014), NFC-enabled mobile 629
574 made up of a large number of simple processing elements known
credit card acceptance (Leong, Hew, et al., 2013), mobile music 630
575 as nodes or neurons (Palmer, Jos Montao, & Ses, 2006, p.
acceptance (Sim et al., 2014) and mobile learning acceptance 631
576 782). Compared with traditional linear models such as MRA,
(Tan et al., 2014). 632
577 SEM, Logistic Regression and Multiple Discriminant Analysis,
578 ANN is an articial intelligence (AI) tool that has outperformed
579 these models as it has the ability to detect both linear and nonlin- 5. Discussion 633
580 ear relationships with high predictive accuracy (Leong, Hew, et al.,
581 2013). Besides, ANN requires no multivariate assumptions such as The differences in the order of the service quality dimensions on 634
582 normality, linearity or homoscedasticity to be fullled (Lee, Leong, customer satisfaction between the SEM and the ANN analyses are 635
583 Hew, & Ooi, 2013). expected as the SEM analysis is capable of analyzing only compen- 636
584 Even with these advantages, NN is not so suitable for causal satory linear relationships whereas the ANN is more robust and is 637
585 relationship hypothesis testing due to its black-box operating capable of analyzing both compensatory linear relationships and 638
586 nature and since linear models (e.g. SEM, MRA) have the non compensatory and non linear relationships. Due to these 639

Table 13
Relevance of predictor variables based on at least one non-zero synaptic weight with hidden neurons.

Input Neural networks

variable NN1 NN2 NN3 NN4 NN5 NN6 NN7 NN8 NN9 NN10

Note: The symbol U indicates that the variable is relevant; TA = Tangible, RL = Reliability, RP = Responsiveness.

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 11

Table 14
RMSE values and Sensitivity Analysis.

Neural network Input neurons: TA, RL, Input neurons: CS Neural network Relative importance
Output nodes: CS Output nodes: CL TA RL RP
Training Testing Training Testing
1 0.0786 0.0655 0.0781 0.0712 1 0.310 0.328 0.362
2 0.0783 0.0738 0.0803 0.0672 2 0.401 0.367 0.232
3 0.0764 0.072 0.0776 0.0763 3 0.344 0.333 0.323
4 0.0851 0.0719 0.0867 0.0548 4 0.228 0.380 0.392
5 0.0814 0.0678 0.0837 0.0788 5 0.416 0.233 0.350
6 0.0791 0.0734 0.084 0.0770 6 0.357 0.176 0.457
7 0.0839 0.0775 0.0882 0.0772 7 0.220 0.378 0.402
8 0.0854 0.1000 0.0806 0.1025 8 0.289 0.392 0.319
9 0.0829 0.0655 0.0851 0.0941 9 0.311 0.353 0.335
10 0.0854 0.0743 0.0750 0.0963 10 0.379 0.224 0.397
Mean RMSE 0.0816 0.0742 0.0819 0.0795 Mean relative importance 0.326 0.316 0.357
Standard deviation 0.0034 0.0099 0.0043 0.0144 Normalized importance (%) 91.2 88.7 100.0

Note: TA = tangible, RL = reliability, RP = responsiveness, CS = customer satisfaction, CL = customer loyalty.

640 differences, the order of regression weights from the SEM may not condition. Since reliability also has the second strongest indirect 682
641 be the same as the order of normalized importance from the ANN. effect on CL, airline airlines should always provide the best reliable 683
642 These differences are similar to those of Chan and Chong (2012), products and services in order to build the condence of their cus- 684
643 Chong (2013a,b), Chong and Bai (2014) and Sim et al. (2014). tomers. When customers have their condence towards the car- 685
644 Based on the fact that ANN has out-performed other traditional rier, they will consider to re-patron with the same carrier in 686
645 regression techniques such as SEM, logistic, multiple and discrim- future. This will lead to resistance to switch to other competitors 687
646 inant regression (Chiang et al., 2006; Chong, 2013a; Morris et al., and hence retaining CL. 688
647 2004), the normalized importance of the predictors obtained from
648 the ANN analysis were used instead of the SEM regression weights. 5.3. Responsiveness 689

649 5.1. Tangibles Responsiveness was found to have the most important predic- 690
tor of CS which is consistent to Ling and Lin (2005), Abdullah 691
650 The results of the ANN analysis revealed that tangibles are the et al. (2007), Nadiri et al. (2008) and Arifn et al. (2010). 692
651 second important predictor of CS. This is in agreement with Therefore, airlines need to consider provision of swift and prompt 693
652 Abdullah et al. (2007), Arifn, Salleh, Aziz, and Asbudin (2010), services in order to meet customers demands. Instant assistance 694
653 Ling and Lin (2005), Nadiri et al. (2008) and Saha and Theingi and help required whenever there are requests or complaints from 695
654 (2009). It has the strongest direct impact on CS besides having the customers. In order to address this issue, constant and period- 696
655 the strongest indirect effect on CL. Therefore, airlines need to ical trainings are required for all cabin crews and ground staffs 697
656 ensure that the physical conditions of their aircrafts, facilities regarding the response time and responsive actions in various sit- 698
657 and equipment as well as the reading and entertainment materials uations. Besides that, airlines should always provide a channel in 699
658 provide in-ight and the appearance of their cabin crews are the form of customer service evaluation form, suggestion boxes, 700
659 always tip-top. It is only with the utmost environment, customers customer feedback form, email, etc as a way of measuring the 701
660 will be satised with their journey. Since tangibles also have a responsiveness of their workers. Assessment may also be done by 702
661 strong signicant indirect effect on CL, airlines need also put ample the superiors to evaluate their workers response time and actions. 703
662 priority and attention in terms of the tangibles dimension so that Responsiveness also indirectly poses a signicant effect on CL and 704
663 existing customers can be retained, thus ensuring long term CL thus airlines need to put due attention to ensure that customers 705
664 so that they will be able to sustain the competitive advantages. are loyal to their company. This will reduce turn-over rate of cus- 706
tomers to their competitors while sustaining the competitive edge. 707
665 5.2. Reliability
5.4. Assurance 708
666 In the study, reliability was found to have a signicant direct
667 impact on CS although it is the third important predictor after Assurance has insignicant direct impact on CS. It contradicts to 709
668 responsiveness and tangibles. This is in accordance with the studies carried out by Abdullah et al. (2007), Cunningham et al. 710
669 Abdullah et al. (2007), Arifn et al. (2010), Cunningham, Young, (2002), Ling and Lin (2005) and Nadiri et al. (2008). Such a nding 711
670 and Lee (2002), Ling and Lin (2005) and Nadiri et al. (2008). can be justiable by the high level of knowledge and courtesy 712
671 Hence, the airlines need to ensure that their products and services among the cabin crew of the airlines. For example, the excellent 713
672 are always trustable and reliable. This can be done by providing services of MAS and Air Asia have been proven by the various 714
673 punctuality in ight schedules, ensuring safety of luggage and their awards presented to both MAS and Air Asia. In fact, MAS won 715
674 deliveries, guaranteeing the safety and security of their passengers, the Staff Excellence Service for Asia and Worlds Best 716
675 providing the accurate and up-to-date ight information and quick Economical Class awards at the World Airline Awards in 717
676 actions on contingency plans whenever there are emergencies or Hamburg (World Airline awards, 2010) and Asias Leading 718
677 mishaps. Constant and periodical monitoring and supervision is Airline award at the World Travel Awards Asia and Australasia 719
678 required on all the tangibles of the carrier as a way to maintain in Dubai (The Star, 2013c). Air Asia won the Worlds Best Low- 720
679 the highest level of physical condition and appearance. The superi- Cost Airline award for the third consecutive year in 2011 (World 721
680 ors need to immerse themselves through walkabouts from time to Airline awards, 2011). Therefore, customers do not perceive assur- 722
681 time to personally check that all the tangibles are in good ance as a factor in determining their satisfaction towards a carrier. 723

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

12 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

724 Based on these track records, airlines need to maintain a high level 6.2. Managerial implications 780
725 of assurance with their customers so that their customers are
726 always satised and loyal to them. Besides, no signicant indirect In terms of managerial implications, the empirical ndings 781
727 effect was detected of assurance on CL. Hence, customers do not obtained from this study can contribute to CEOs and managers of 782
728 perceive assurance as a factor in determining their loyalty towards the airline industry by providing them the vital information about 783
729 the carrier. Their decisions were indirectly inuenced by tangibles, the dimensions of the SERVPERF model that affects SQ as well as 784
730 reliability and responsiveness of the SERVPERF dimensions as elu- the importance of sustaining SQ in order to raise CL. Since tangibles 785
731 cidated above. were found to be an important antecedent of CS, an extra effort is 786
required to ensure that the physical conditions of their aircrafts, 787
732 5.5. Empathy equipment and facilities as well as the appearances of their staffs 788
are in the top forms. Utilizing the Total Quality Management 789
733 Empathy is reported to have an insignicant impact on CS and it (TQM) system together with the implementation of the ISO 9000 790
734 is contrary to Abdullah et al. (2007), Nadiri et al. (2008) and Arifn standards may be some of the options. 791
735 et al. (2010) but in agreement to Ling and Lin (2005). The insignif- Providing reliable products and services are among the other 792
736 icant impact may be due to the customers perception that they are important factors to be well addressed. Airline companies are sub- 793
737 well understood and they have been receiving adequate individual jected to the evaluation on reliability by their customers. Due to 794
738 attention and care from the cabin crew and hence do not consider the high competitiveness of the airline industry, any slack or short- 795
739 empathy as a factor in determining their satisfaction towards the coming in this area will surely lead to unwanted results. Therefore, 796
740 carrier. Another reason is that due to keen competition in airline internal assessment and auditing should be carried out in order to 797
741 industry, most airlines have given considerable focus to provide measure the level of reliability. External measures such as survey 798
742 individual and personal attention and care to their customers as form, face-to-face short interview, online or telephone survey, 799
743 a strategy of customer retention. Customer First is the favorable email feedback, etc are among the options to be chosen. By enter- 800
744 tagline among most of the airlines. Moreover, no signicant indi- taining these pieces of feedbacks, airlines need to take the neces- 801
745 rect effect was reported of empathy on CL whereby customers do sary actions to remedy any shortcomings so that the degree of 802
746 not consider it to be vital in determining their loyalty to a carrier. reliability is tip-topped. 803
To stay competitive, a carrier may also provide fast and respon- 804
747 5.6. Customer satisfaction sive services to their customers. Responsiveness is imperative to 805
ensure CS. Hence, evaluation and assessment must be conducted 806
748 The research revealed that CS has a very strong direct impact on periodically to ascertain the high level of responsiveness among 807
749 CL. This is in line with Clemes et al. (2008), Nadiri et al. (2008) and its workers. Simulations and mock tests may be done to check 808
750 Saha and Theingi (2009). Hence, airline companies need to ensure the level of responsiveness such as response time, response 809
751 that each and every one of their passengers is satised with the actions, ability to act depending on various unexpected situations, 810
752 products and services provided in order to guarantee future re-pa- etc. Besides that, feedback form, survey questionnaire, interview, 811
753 tronage and uphold customer loyalty. In order to achieve this etc may also be used to gather information regarding the degree 812
754 objective, they should provide a good channel or platform for cus- of responsiveness among the workers. 813
755 tomers to express their gratitude and satisfaction regarding the Last but not the least, airline airlines need to ensure that CS is 814
756 products and services received. For instance, customer feedback always sustained as it is the important antecedent to CL. It is only 815
757 form, survey questionnaire, short face-to-face interview or satis- with satisfaction, customers will become loyal to them. In this per- 816
758 faction ballot box are among the recommendations to be consid- spective, they must change their strategies every now and then to 817
759 ered. Besides, online surveys, telephone interviews and emails suit the requirements and wants of the customers in order to keep 818
760 interview may also be other options to gather information pertain- them satised. To achieve this aim, key performance indicators 819
761 ing to customers satisfaction. Then based on the feedback informa- (KPI) may be introduced to all workers so that the performance 820
762 tion, the carrier may further improvise and upgrade their products of each worker can be systematically assessed. Sometimes, it is 821
763 and services so that the degree and gratication among their cus- inevitable to use the Hire and Fire policy to ensure that only 822
764 tomers may be further elevated and airlines aim to get customer the best workers stay and only the best products and services are 823
765 enlightenment. offered. 824

766 6. Implications 7. Conclusion, limitations and future research direction 825

767 6.1. Theoretical implications This research empirically investigate SERVPERFs inuence on 826
both CS and CL among the full service and low cost airlines in 827
768 Theoretically, this research has been able to move forward the the Malaysian airline industry, with a further examination on 828
769 literature pertaining to the impact of SQ and CS on CL of the airline whether a signicant difference lies between full service and low 829
770 industry. The research is among the rst to examine these impacts cost airlines in terms of construct validity of the SERVPERF dimen- 830
771 using the SERVPERF model with a multi-stage SEMneural net- sions. In response to rst research question (RQ1), the dimensions 831
772 work approach in both full service and low cost airlines. Besides, of service quality performance of the airline industry chosen for 832
773 the mediation effects were also studied based on the indirect this research are tangible, reliability, responsiveness, assurance 833
774 effects of the SEM analysis. In addition, the model is among the and empathy. As for the second research question (RQ2), no signif- 834
775 rst to be tested for group invariance between full service and icant difference was reported to exist between full service and low 835
776 low cost airlines and was found to be robust across both groups. cost airlines for all SERVPERF constructs except for the reliability 836
777 Thus the conceptual model of this study may be used as a source component. Furthermore, no signicant differences exist in terms 837
778 of reference for other researchers who are interested in studying of causal relationship difference in the structural model across 838
779 the impact of SQ on both CS and CL. the two groups. As for the impacts of SERVPERF dimensions 839

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 13

840 towards customer satisfaction of the airline industry, tangibles, Bolton, R. N., & Drew, J. H. (1991). A multistage model of customers assessments of 907
service quality and value. Journal of Consumer Research, 17(4), 375384. 908
841 reliability and responsiveness were found to have a direct, signi- 909
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model t. Newbury
842 cant and positive impact on CS. At the same time these indirectly Park, CA: Sage Publications. 910
843 impact on CL as well. While assurance and empathy pose no direct Buzzell, R. D., & Gale, B. T. (1989). The PIMS principles. New York, NY: Free Press. 911
Byrne, B. M. (2001). Structural equations modeling with AMOS: Basic concepts, 912
844 and signicant impact on CS and no indirect signicant impact on 913
applications, and programming. New Jersey: Lawrence Erlbaum Associates, Inc.,
845 CL, thus answering the third research question (RQ3). Publisher. 914
846 One of theoretical contributions of this research is the use of Byrne, B. M. (2010). Structural equation modeling with AMOS: Basic concepts, 915
applications, and programming (2nd ed.). New York: Routledge, Taylor and 916
847 expert systems and articial intelligence systems (i.e. ANN) in
Francis Group. 917
848 the new context of low cost and full service airlines compared to Carman, J. M. (1990). Consumer perceptions of service quality: An assessment of the 918
849 other existing studies on m-commerce adoption (Chong, 2013a), SERVQUAL dimensions. Journal of Retailing, 66(2), 2745. 919
850 inter-organizational system standard adoption (Chan & Chong, Chan, F. T., & Chong, A. Y. (2012). A SEMneural network approach for 920
understanding determinants of inter-organizational system standard adoption 921
851 2012), open IOS adoption (Chong & Bai, 2014), NFC-enabled mobile and performances. Decision Support Systems, 54(1), 621630. 922
852 credit card acceptance (Leong, Hew, et al., 2013), mobile music Chau, V. S., & Kao, Y. Y. (2009). Bridge over troubled water or long and winding 923
853 acceptance (Sim et al., 2014) and mobile learning acceptance road? Gap-5 airline service quality performance measures. Managing Service 924
Quality, 19(1), 106134. 925
854 (Tan et al., 2014). Furthermore, the use of SERVPERF instead of Chen, F. Y., & Chang, Y. H. (2005). Examining airline service quality from a process 926
855 the well-studied SERVQUAL dimensions using multiple group anal- perspective. Journal of Air Transport Management, 11(2), 7987. 927
856 ysis may provide better understanding from a new perspective. Chen, C. F., & Chen, S. C. (2014). Investigating the effects of job demands and job 928
resources on cabin crew safety behaviors. Tourism Management, 41(April), 929
857 With the predictive analytic (Scott & Walczak, 2009; Shmueli & 930
858 Koppius, 2011) SEMANN approach, the study may also provide Cheng, J. H., Chen, F. Y., & Chang, Y. H. (2008). Airline relationship quality: An 931
859 methodological contribution in terms of statistical analysis examination of Taiwanese passengers. Tourism Management, 29(3), 932
487499. 933
860 techniques. 934
Chiang, W. K., Zhang, D., & Zhou, L. (2006). Predicting and explaining patronage
861 Several limitations were found in this research. Firstly, this behavior toward web and traditional stores using neural networks: A 935
862 research was conducted in a developing ASEAN country and hence comparative analysis with logistic regression. Decision Support Systems, 41(2), 936
514531. 937
863 the results may not be able to generalize to other geographical
Chin, A. (2002). Impact of frequent yer programs on the demand for air travel. 938
864 regions. Future research can overcome such limitation by carrying Journal of Air Transportation, 7(2), 5386. 939
865 out a cross-country comparative study. Secondly, it used a cross- Chong, A. Y. L. (2013a). A two-staged SEMneural network approach for 940
866 sectional survey approach and thus the effect of time was not understanding and predicting the determinants of m-commerce adoption. 941
Expert Systems with Applications, 40(4), 12401247. 942
867 examined. It is suggested that in future, a longitudinal survey Chong, A. Y. L. (2013b). Predicting m-commerce adoption determinants: A neural 943
868 may be conducted. The research used the SERVPERF model and this network approach. Expert Systems with Applications, 40(2), 523530. 944
869 has limited representation, therefore, in future other SQ models Chong, A. Y. L., & Bai, R. (2014). Predicting open IOS adoption in SMEs: An integrated 945
SEMneural network approach. Expert Systems with Applications, 41(1), 946
870 such as SERVPEX (Robledo, 2001), weighted SERVPERF and 221229. 947
871 SERVQUAL (Jain & Gupta, 2004) may be deployed to examine their Clemes, M. D., Gan, C., Kao, T. H., & Choong, M. (2008). An empirical analysis of 948
872 impacts on CS and CL. It would be an interesting nding if a customer satisfaction in international air travel. Innovative Marketing, 4(2), 949
4962. 950
873 research could be carried out on the unprecedented aviation event 951
Cronin, J., Michael, K., Brady, M. K., & Hult, G. T. (2000). Assessing the effects of
874 such as the mysterious lost of the Malaysian Airline Boeing 777 quality, value, and customer satisfaction on consumer behavioral intentions in 952
875 Flight MH370 (The Star, 2014) on customer loyalty. Finally, in service environments. Journal of Retailing, 76(2), 193218. 953
Cronin, J., & Taylor, S. A. (1992). Measuring service quality: A re-examination and 954
876 future, a study on the impact of customers character traits on CS 955
extension. Journal of Marketing, 56(July), 5567.
877 and CL may also be conducted. Cunningham, L. F., Young, C. E., & Lee, M. (2002). Cross-cultural perspectives of 956
service quality and risk in air transportation. Journal of Air Transportation, 7(1), 957
326. 958
878 Acknowledgements Davidow, M. (2003). Have you heard the word? The effect of word of mouth on 959
perceived justice, satisfaction and repurchase intentions following complaint 960
handling. Journal of Consumer Satisfaction, Dissatisfaction and Complaining 961
879 The authors would like to express their heartfelt gratitude to a 962
Behavior, 16(1), 6780.
880 group of undergraduate students who have put in their time and Dean, A. M. (2007). The impact of the customer orientation of call center employees 963
881 effort in assisting the collection of the required data for this study. on customers affective commitment and loyalty. Journal of Service Research, 964
10(2), 161173. 965
Dick, A. S., & Basu, K. (1994). Customer loyalty toward an integrated conceptual 966
882 References framework. Journal of Academy of Marketing Science, 22(2), 99113. 967
Ekiz, H. E., & Arasli, H. (2007). Measuring the impacts of organizational responses: 968
Case of Northern Cyprus hotels. Managing Global Transitions: International 969
883 Abdullah, K., Manaf, N. H. A., & Noor, K. M. (2007). Measuring the service quality of
Research Journal, 5(3), 271287. 970
884 airline services in Malaysia. Journal of Economics and Management, 15(1), 129.
Fecikova, I. (2004). Research and concepts: An index method for measurement of 971
885 Aghdaie, S. A., & Faghani, F. (2012). Mobile banking service quality and customer
customer satisfaction. TQM Magazine, 16(1), 5766. 972
886 satisfaction (application of SERVQUAL model). International Journal of
Feng, J., Prajogo, D. I., Tan, K. C., & Sohal, A. S. (2006). The impact of TQM practices on 973
887 Management and Business Research, 2(4), 351361.
performance. A comparative study between Australian and Singaporean 974
888 Air Asia 2010 Annual Report. Available from: <
organizations. European Journal of Innovation Management, 9(3), 269278. 975
889 resources/my/common/pdf/AirAsia/IR/AirAsia_AR10.pdf> (Accessed
Fick, G. R., & Ritchie, J. R. B. (1991). Measuring service quality in the travel and 976
890 21.11.2013).
tourism industry. Journal of Travel Research Fall, 29. 977
891 Anderson, E. W., Fornell, C., & Lehmann, D. R. (1994). Customer satisfaction, market
Fornell, C. (1992). A national customer satisfaction barometer: The Swedish 978
892 share, and protability: Findings from Sweden. Journal of Marketing, 58(July),
experience. Journal of Marketing, 56, 621. 979
893 5366.
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with 980
894 Arbuckle, J. L. (2008). Amos 17.0.2 (Version 17) [Software and Manual]. FL: Amos
unobservable variables and measurement error. Journal of Marketing Research, 981
895 Development Corporation Crawfordville.
18(1), 3950. 982
896 Arifn, A. A. M., Salleh, A. H. M., Aziz, N. A., & Asbudin, A. A. (2010). Service quality
Gourdin, K. N. (1988). Bringing quality back to commercial air travel. Transportation 983
897 and satisfaction for low cost carriers. International Review of Business Research
Journal, 2329. 984
898 Papers, 6(1), 4756.
Gourdin, K. N., & Kloppenborg, T. J. (1991). Identifying service gaps in commercial 985
899 Atilgan, E., Akinci, S., & Aksoy, S. (2008). Measuring and mapping customers
air travel: The rst step toward quality improvement. Transportation Journal, 986
900 expectations and perceptions for airlines: The Sunexpress case with the GAPS
31(1), 2230. 987
901 model. Journal of Global Strategic Management, 3, 6778.
Gremler, D. D., & Brown, S. (1996). Service loyalty: Its nature, importance and 988
902 Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation model. Journal
implications. In B. Edvardsson (Ed.), Advancing service quality: A global 989
903 of Academy of Marketing Science, 16(1), 7494.
perspective. International Service Quality Association. 990
904 Bitner, M. J., & Hubbert, A. R. (1994). Service quality: New directions in theory and
Hair, J. F., Jr., Black, W. C., Babin, B. J., & Anderson, R. E. (2010). Multivariate data 991
905 practice. In R. T. Rust & R. Oliver (Eds.), Encounter satisfaction versus overall
analysis: A global perspective (7th ed.). NJ: Prentice Hall. 992
906 satisfaction versus quality (pp. 77). Thousand Oaks, California: Sage Publications.

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

14 L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx

993 Haykin, S. (1999). Neural networks: A comprehensive foundation (2nd ed.). Upper Liu, J., Qu, H., Huang, D., Chen, G., Yue, X., Zhao, X., et al. (2014). The role of social 1078
994 Saddle River, NJ: Prentice Hall. capital in encouraging residents pro-environmental behaviors in community- 1079
995 Hew, T. S., & Leong, L. Y. (2011). An empirical analysis of Malaysian pre-university based ecotourism. Tourism Management, 41, 190201. 1080
996 students ICT competency gender differences. International Journal of Network Maybank IB Research (2013). Aviation: Cautiously optimistic. Available from: 1081
997 and Mobile Technologies, 2(1), 1529. < 1082
998 Hoang, D. T., Igel, B., & Laosirihongthong, T. (2006). The impact of total quality Outlook_SU_20130405_MIB2_3815.pdf> (Accessed 01.07.2013). 1083
999 management on innovation: Findings from a developing country. International Maybank IB Research (2014). 2013: A year to forget. Available from: <http:// 1084
1000 Journal of Quality & Reliability Management, 23(9), 10921117. 1085
1001 Hsu, C. I., Shih, M. L., Huang, B. W., Lin, B. Y., & Lin, C. N. (2009). Predicting tourism Summary_MKE_5899.pdf> (Accessed 01.05.2014). 1086
1002 loyalty using an integrated Bayesian network mechanism. Expert Systems with McGehee, N. G., Lee, S., OBannon, T. L., & Perdue, R. R. (2010). Tourism-related social 1087
1003 Applications, 36(9), 1176011763. capital and its relationship with other forms of capital: An exploratory study. 1088
1004 Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for t indexes in covariance structure Journal of Travel Research, 49(4), 486500. 1089
1005 analysis: Conventional criteria versus new alternatives. Structural Equation Morris, S. A., Greer, T. H., Hughes, C., & Clark, W. J. (2004). Prediction of CASE 1090
1006 Modeling: A Multidisciplinary Journal, 6(1), 155. adoption: A neural network approach. Industrial Management & Data Systems, 1091
1007 Huang, Y. K. (2009). The effect of airline service quality on passengers behavioural 104(2), 129135. 1092
1008 intentions using SERVQUAL scores: A Taiwan case study. Journal of the Eastern Nadiri, H., & Hussain, K. (2005). Diagnosing the zone of tolerance for hotel services. 1093
1009 Asia Society for Transportation Studies, 8, 23012314. Managing Service Quality, 15(3), 259277. 1094
1010 IATA (international air transport association) (2014). Financial forecast March Nadiri, H., Hussain, K., Ekiz, E. H., & Erdogan, S. (2008). An investigation on the 1095
1011 2014. Available from: < factors inuencing passengers loyalty in the north Cyprus National Airline. The 1096
1012 IATA-economic-brieng-nancial-forecast-March-2014.pdf> (Accessed TQM Journal, 20(3), 265280. 1097
1013 06.04.2014). Natalisa, D., & Subroto, B. (2003). Effects of management commitment on service 1098
1014 Ismail, I., Haron, H., Ibrahim, D. N., & Mohd Isa, S. (2006). Service quality, client quality to increase customer satisfaction of domestic airlines in Indonesia. 1099
1015 satisfaction and loyalty towards audit rms: Perceptions of Malaysian public Singapore Management Review, 25(1), 85104. 1100
1016 listed companies. Managerial Auditing Journal, 21(7), 738756. Newman, K. (2001). Interrogating SERVQUAL: A critical assessment of service 1101
1017 Jain, S. K., & Gupta, G. (2004). Measuring service quality: SERVQUAL vs. SERVPERF quality measurement in a high street retail bank. The International Journal of 1102
1018 scales. VIKALPA, 29(2), 2537. Bank Marketing, 19(3), 126139. 1103
1019 Juwaheer, T. D. (2004). Exploring international tourists; perceptions of hotel Nunkoo, R., Ramkissoon, H., & Gursoy, D. (2013). Use of structural equation 1104
1020 operations by using a modied SERVQUAL approach a case study of Mauritius. modeling in tourism research past, present, and future. Journal of Travel 1105
1021 Managing Service Quality, 14(5), 350364. Research, 52(6), 759771. 1106
1022 Kaplanidou, K., & Vogt, C. (2006). A structural analysis of destination travel Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). NY: McGraw- 1107
1023 intentions as a function of web site features. Journal of Travel Research, 45(2), Hill. 1108
1024 204216. Oh, S. (1995). Customer evaluations of telecommunication service quality. Faculty 1109
1025 Kau, A. K., & Loh, E. W. Y. (2006). The effects of service recovery on consumer working paper (in Korean), Yonsei University, Korea. 1110
1026 satisfaction: A comparison between complainants and non-complainants. Oliver, R. L. (1980). A cognitive model of the antecedents and consequences of 1111
1027 Journal of Services Marketing, 20(2), 101111. satisfaction decisions. Journal of Marketing Research, 17(4), 460469. 1112
1028 Kilbourne, W., Duffy, J. A., Duffy, M., & Giarchi, G. G. (2004). The applicability of Oliver, R. L. (1993). A conceptual model of service quality and service satisfaction: 1113
1029 SERVQUAL in cross-national measurements of health-care quality. Journal of Compatible goals, different concepts. In T. A. Swartz, D. E. Bowen, & S. W. Brown 1114
1030 Services Marketing, 18(7), 524533. (Eds.), Advances in services marketing and management: Research and practice 1115
1031 Kim, Y. K., & Lee, H. R. (2011). Customer satisfaction using low cost carriers. Tourism (2nd ed., pp. 6585). Greenwich, CT: JAI Press Inc. 1116
1032 Management, 32(2), 235243. Olorunniwo, F., & Hsu, M. K. (2006). A typology analysis of service quality, customer 1117
1033 Kline, R. B. (1998). Structural equation modeling. New York: The Guilford Press. satisfaction and behavioral intentions in mass services. Managing Service 1118
1034 Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). Quality, 16(2), 106123. 1119
1035 NY: The Guilford Press. Ooi, K. B., Lin, B., Tan, B. I., & Chong, A. Y. L. (2011). Are TQM practices supporting 1120
1036 Kumar, S. A., Mani, B. T., Mahalingam, S., & Vanjikovan, M. (2010). Inuence of customer satisfaction and service quality? Journal of Services Marketing, 25(6), 1121
1037 service quality on attitudinal loyalty in private retail banking: An empirical 410419. 1122
1038 study. Journal of Management Research, IX(4), 2138. Palmer, A., Jos Montao, J., & Ses, A. (2006). Designing an articial neural network 1123
1039 Lee, T. H. (2013). Inuence analysis of community resident support for sustainable for forecasting tourism time series. Tourism Management, 27(5), 781790. 1124
1040 tourism development. Tourism Management, 34, 3746. Palmer, A., Koenig-Lewis, N., & Medi Jones, L. E. (2013). The effects of residents 1125
1041 Lee, M., & Lee, I. (1997). Developing service marketing strategies for different types social identity and involvement on their advocacy of incoming tourism. Tourism 1126
1042 of retail stores. Journal of Distribution Channels (in Korean), 2(1), 914. Management, 38, 142151. 1127
1043 Lee, H., Lee, Y., & Yoou, D. (2000). The determinants of perceived service quality and Parasuraman, A., Berry, L. L., & Zeithaml, V. A. (1991a). Perceived service quality as a 1128
1044 its relationship with satisfaction. Journal of Service Marketing, 14(3), customer based performance measure: An empirical examination of 1129
1045 217231. organizational barriers using an extended service quality model. Human 1130
1046 Lee, V. H., Leong, L. Y., Hew, T. S., & Ooi, K. B. (2013). Knowledge management: A key Resource Management, 3564. 1131
1047 determinant in advancing technological innovation? Journal of Knowledge Parasuraman, A., Berry, L. L., & Zeithaml, V. A. (1991b). Understanding customer 1132
1048 Management, 17(6), 848872. expectations of service. Sloan Management Review, 32(3), 3949. 1133
1049 Lee, V. H., Ooi, K. B., Tan, B. I., & Chong, A. Y. L. (2011). A structural analysis of the Parasuraman, A., Berry, L. L., & Zeithaml, V. A. (1991c). Renement and 1134
1050 relationship between TQM practices and product innovation. Asia Journal of reassessment of the SERVQUAL scale. Journal of Retailing, 67(4), 420450. 1135
1051 Technology Innovation, 18(1), 7396. Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1985). A conceptual model of service 1136
1052 Legoherel, P. (1998). Quality of tourist services: The inuence of each participating quality and its implications for future research. The Journal of Marketing, 4150. 1137
1053 component on the consumers overall satisfaction regarding tourist services Parasuraman, A., Zeithaml, V. A., & Berry, L. L. (1988). SERVQUAL: A multiple-item 1138
1054 during a holiday. In Proceedings of 3rd international conference on tourism and scale for measuring consumer perceptions of service quality. Journal of Retailing, 1139
1055 hotel industry in Indo-China and Southeast Asia: Development, marketing and 64(1), 1240. 1140
1056 sustainability (pp. 4754). Thailand. Park, J. W., Robertson, R., & Wu, C. L. (2004). The effect of airline service quality on 1141
1057 Leong, L. Y., Hew, T. S., Ooi, K. B., & Lin, B. (2011). Inuence of gender and English passengers behavioral intentions: A Korean case study. Journal of Air Transport, 1142
1058 prociency on Facebook mobile adoption. International Journal of Mobile 10(6), 435439. 1143
1059 Communications, 9(5), 495521. Park, J. W., Robertson, R., & Wu, C. L. (2005). Investigating the effects of airline 1144
1060 Leong, L. Y., Hew, T. S., Ooi, K. B., & Lin, B. (2012). The determinants of customer service quality on airline image and passengers future behavioural intention: 1145
1061 loyalty in Malaysian mobile telecommunication services: A structural analysis. Findings from Australian international air passengers. The Journal of Tourism 1146
1062 International Journal of Services, Economics and Management, 4(3), 209235. Studies, 16(1), 211. 1147
1063 Leong, L. Y., Hew, T. S., Tan, G. W. H., & Ooi, K. B. (2013). Predicting the determinants Park, J. W., Robertson, R., & Wu, C. L. (2006). Modelling the impact of airline service 1148
1064 of the NFC-enabled mobile credit card acceptance. A neural networks approach. quality and marketing variables on passengers future behavioural intentions. 1149
1065 Expert Systems with Applications, 40(14), 56045620. Transportation Planning and Technology, 29(5), 359381. 1150
1066 Leong, L. Y., Ooi, K. B., Chong, A. Y. L., & Lin, B. (2013). Modeling the stimulators of Prajogo, D., & Hong, S. W. (2008). The effect of TQM on performance in R&D 1151
1067 the behavioral intention to use mobile entertainment: Does gender really environment: A perspective from South Korean rms. Technovation, 28(12), 1152
1068 matter? Computers in Human Behavior, 29(5), 21092121. 855863. 1153
1069 Lin, R. J., Chen, R. H., & Chiu, K. K. S. (2010). Customer relationship management and Prayag, G. (2007). Assessing international tourists perceptions of service quality at 1154
1070 innovation capability: An empirical study. Industrial Management & Data Air Mauritius. International Journal of Quality & Reliability Management, 24(5), 1155
1071 Systems, 101(1), 111133. 492514. 1156
1072 Lindenmeier, J., & Tscheulin, D. K. (2008). The effects of inventory control and Ramkissoon, H., Graham Smith, L. D., & Weiler, B. (2013). Testing the dimensionality 1157
1073 denied boarding on customer satisfaction: The case of capacity-based airline of place attachment and its relationships with place satisfaction and pro- 1158
1074 revenue management. Tourism Management, 29(1), 3243. environmental behaviours: A structural equation modelling approach. Tourism 1159
1075 Ling, F. I., & Lin, K. (2005). Difference in service quality of cross-strait airlines and its Management, 36, 552566. 1160
1076 effect on passengers preferences. Journal of the East Asia Society for Robledo, M. A. (2001). Measuring and managing service quality: Integrating 1161
1077 Transportation Studies, 6, 798813. customer expectations. Managing Service Quality, 11(1), 2231. 1162

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),
ESWA 9990 No. of Pages 15, Model 5G
4 May 2015

L.-Y. Leong et al. / Expert Systems with Applications xxx (2015) xxxxxx 15

1163 Rogerson, W. P. (1983). Reputation and product quality. Bell Journal of Economics, The Star (2012). Air Asia bags business travellers low-cost Airline award. Available 1207
1164 14(Autumn), 500510. from: < 1208
1165 Saha, G. C., & Theingi (2009). Service quality, satisfaction and behavioural 12110334&sec=nation> (Accessed 26.10.2013). 1209
1166 intentions: A study of low-cost airline carriers in Thailand. Managing Service The Star (2013a). Air Asia clinches rst Asean level award. Available from: <http:// 1210
1167 Quality, 19(3), 350371. 1211
1168 Scott, J. E., & Walczak, S. (2009). Cognitive engagement with a multimedia ERP business> (Accessed 02.04.2013). 1212
1169 training tool: Assessing computer self-efcacy and technology acceptance. The Star (2013b). Air Asia bags prestigious awards again. Available from: <http:// 1213
1170 Information & Management, 46(4), 221232. 1214
1171 Selnes, F. (1993). An examination of the effect of product performance on brand awards-again.aspx> (Accessed 30.06.2013). 1215
1172 reputation, satisfaction, and loyalty. European Journal of Marketing, 27(9), 1935. The Star (2013c). Malaysia Airlines and Air Asia take awards at travel Oscars. 1216
1173 Shmueli, G., & Koppius, O. R. (2011). Predictive analytics in information systems Available from: < 1217
1174 research. MIS Quarterly, 35(3), 553572. Malaysian-airlines-are-Asias-best-carriers.aspx> (Accessed 10.10.2013). 1218
1175 Sim, J. J., Tan, G. W. H., Wong, J. C. J., Ooi, K. B., & Hew, T. S. (2014). Understanding The Star (2014). Missing MH370: Timeline of events (Day 9). Available from: 1219
1176 and predicting the motivators of mobile music acceptance a multi-stage < 1220
1177 MRAarticial neural network approach. Telematics and Informatics, 31(4), Timeline-of-Events-Day-9/> (Accessed 17.03.2014). 1221
1178 569584. Wang, C. J., Tsai, H. T., & Tsai, M. T. (2014). Linking transformational leadership and 1222
1179 So, K. K. F., King, C., Sparks, B. A., & Wang, Y. (2014). The role of customer employee creativity in the hospitality industry: The inuences of creative role 1223
1180 engagement in building consumer loyalty to tourism brands. Journal of Travel identity, creative self-efcacy, and job complexity. Tourism Management, 40, 1224
1181 Research. 0047287514541008. 7989. 1225
1182 Stevens, P., Knutson, B., & Patton, M. (1995). DINESERV: A tool for measuring service Wisniewski, M. (2001). Assessing customer satisfaction with local authority 1226
1183 quality in restaurants. The Cornell Hotel and Restaurant Administration Quarterly, services using SERVQUAL. Total Quality Management, 12(7 & 8), 9951002. 1227
1184 5, 5660. Wong, T. C., Law, K. M. Y., & Yau, H. K. (2011). Analyzing supply chain operation 1228
1185 Suki, N. M. (2014). Passenger satisfaction with airline service quality in Malaysia: A models with the PC-algorithm and the neural network. Expert Systems with 1229
1186 structural equation modeling approach. Research in Transportation Business & Applications, 38(6), 75267534. 1230
1187 Management, 10, 2632. Wong, K. M., & Musa, G. (2011). Branding satisfaction in the airline industry a 1231
1188 Swanson, K. K., & Horridge, P. E. (2004). A structural model for souvenir comparative study of Malaysia Airlines and Air Asia. African Journal of Business 1232
1189 consumption, travel activities, and tourist demographics. Journal of Travel Management, 5(8), 34103423. 1233
1190 Research, 42(4), 372380. World Airline awards (2010). Malaysia Airlines wins staff service excellence Asia 1234
1191 Tan, G. W. H., Ooi, K. B., Chong, S. C., & Hew, T. S. (2014). NFC mobile credit card: The award at World Airline awards. Available from: <http:// 1235
1192 next frontier of mobile payment? Telematics and Informatics, 31(2), 292307.> (Accessed 1236
1193 Tan, G. W. H., Ooi, K. B., Leong, L. Y., & Lin, B. (2014). Predicting the drivers of 16.12.2013). 1237
1194 behavioral intention to use mobile learning: a hybrid SEMneural networks World Airline awards (2011). Air Asia wins the Worlds best low-cost Airline award 1238
1195 approach. Computers in Human Behavior, 36, 198213. for the 3rd consecutive year. Available from: <http:// 1239
1196 Tan, G. W. H., Siah, M. W., Ooi, K. B., Hew, T. S., & Chong, A. Y. L. (2015). The adoption> (Accessed 1240
1197 of PDA for future healthcare system: An emerging market perspective. 16.10.2013). 1241
1198 International Journal of Mobile Communications, 13(1), 128. Yi, Y., & La, S. (2003). The relative effects of three dimensions of service quality on 1242
1199 Teas, R. K. (1993). Expectations, performance evaluation, and consumers CS-A comparative study of existing vs. potential customers. Korean Marketing 1243
1200 perceptions of quality. Journal of Marketing, 57(4), 1834. Association Journal, 18(4), 6797. 1244
1201 Teo, A. C., Cheah, C. M., Leong, L. Y., Hew, T. S., & Shum, Y. L. (2012). What matters Zait, A., & Bertea, P. E. (2011). Methods for testing discriminant validity. 1245
1202 most in mobile payment acceptance? A structural analysis. International Journal Management & Marketing, 9(2), 217224. 1246
1203 of Network and Mobile Technologies, 3(3), 4969. Zopiatis, A., Constanti, P., & Theocharous, A. L. (2014). Job involvement, 1247
1204 Teo, A. C., Tan, G. W. H., Ooi, K. B., Hew, T. S., & Yew, K. T. (2015). The effects of commitment, satisfaction and turnover: Evidence from hotel employees in 1248
1205 convenience and speed in m-payment. Industrial Management & Data Systems, Cyprus. Tourism Management, 41, 129140. 1249
1206 115(2), 311331. 1250

Please cite this article in press as: Leong, L.-Y., et al. An SEMarticial-neural-network analysis of the relationships between SERVPERF, customer satisfac-
tion and loyalty among low-cost and full-service airline. Expert Systems with Applications (2015),