Influential constructs, mediating effects, and ...

Transport Policy 30 (2013) 207–219

Contents lists available at ScienceDirect

Transport Policy journal homepage: www.elsevier.com/locate/tranpol

Influential constructs, mediating effects, and moderating effects on operations performance of high speed rail from passenger perspective Jui-Sheng Chou n, Chun-Pin Yeh Department of Construction Engineering, National Taiwan University of Science and Technology, 43, Section 4, Keelung Road, Taipei 106, Taiwan

art ic l e i nf o

Keywords: Operations performance Policy making High-speed rail Structural equation model Mediating and moderating effects Passenger perspective

a b s t r a c t In a competitive society with diverse consumer needs, service quality, customer satisfaction, customer loyalty, and corporate image determine the sustainability of service-oriented industries. However, management and leadership, employee satisfaction, and employee loyalty also influence company growth and profit. This study applied a theoretical model and findings from related literature to investigate the constructs and observed indices for measuring operations performance in the highspeed railway (HSR) from the passenger perspective. Cause and effect relationships between constructs and operations performance were quantified, and structural equation modeling was used to verify the hypothetical relationships proposed in this study in order to identify constructs, to measure the effects of indices on the constructs, and to measure mediating and moderating effects between constructs. The analytical results showed that leadership and employee cognition have a greater influence on longterm profitability compared to service quality, customer recognition, and corporate image. Notably, employee cognition mediates the effect of leadership on service quality. Further, mediating and moderating effects of corporate image and customer recognition significantly affect operations performance. By using the confirmatory findings of this study as a policy making reference and for clarifying resource use, the HSR can enhance passenger perceptions. Improving the identified evaluation indicators can increase passenger loyalty and improve operating performance in the high-speed rail service. & 2013 Elsevier Ltd. All rights reserved.

1. Introduction The increasing national income of Taiwan and its growing role as an open market for foreign capital have improved quality of life and have increased public awareness of the importance of leisure time, especially after the government implemented regulations regarding a two-day weekend. Thus, service quality requirements of transportation systems and the importance of reducing space value have increased. Competing against time constraints has apparently become the mainstream in various industries. The number of tourist visits to Taiwan has exceeded 150 million since the Taiwan High Speed Rail (THSR) began operations. However, according to its financial report, the THSR substantially reduced its liability only after the government intervened by lowering loan interest rates in 2009 (http:// www.npf.org.tw/post/2/5668). Although the THSR continuously increased the number of operating shifts during 2007 and 2011,

n

Corresponding author. Tel.: þ 886 2 2737 6321; fax: þ886 2 2737 6606. E-mail addresses: [email protected], [email protected] (J.-S. Chou). 0967-070X/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tranpol.2013.09.014

the costs increased with income, and the average passenger rate was only slightly higher than 50% of the total capacity. Therefore, the long-term financial status of the THSR has substantial room for improvement. However, most THSR stations are located in suburban areas far from city centers and downtown areas. Thus, although the THSR enables rapid long-distance southward and northward travel, passengers incur additional time and costs to reach their final destinations. In recent years, the THSR has collaborated with the Taiwan Railway Administration and bus companies to achieve mutual benefits. However, studies suggest that the final overall time and financial costs of using the THSR may actually be greater than that of other transport methods. The THSR sets schedules according to transfer and shuttle services in order to satisfy customer requirements for convenient transportation. Besides airline industries, numerous empirical studies have attempted to measure service quality, customer satisfaction, and customer loyalty in service industries. However, empirical studies of employee satisfaction and loyalty are rarely performed in operating units of service industries (Anderson and Fornell, 2000; Jia and Ping, 2005; Joo and Sohn, 2008; Lin, 2007). The development of information technologies in the twenty-first century enables technology-, economy-, and service-oriented

208

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

industries to understand whether service provided by management and corporations reaches the standards required by customers. To establish customer trust and loyalty, the services provided must also resolve customer doubts and satisfy their service quality requirements. Increased customer revisit rates, improved corporate image, and positive word of mouth can then increase sustainable operating benefits. Research in this area is critically needed by operation units. This study explored the perceived feelings regarding contact with first-line employees from the perspective of passengers (i.e., customers). Quantitative methods were used to analyze service quality-customer satisfaction-customer loyalty (QScLc), including leadership-employee satisfaction-employee loyalty (LSeLe). The effects of these two main linkages on operations performance were analyzed to measure the overall perceptions of THSR passengers and to establish a behavioral model. The THSR and the transportation industry can use the empirical data obtained by an index evaluation of each construct in questionnaire surveys to develop effective marketing plans and reference directions for improving quality. This paper is organized as follows. Section 2 defines and classifies the research constructs, including leadership, employee satisfaction, employee loyalty, operations performance, service quality, corporate image, customer satisfaction, and customer loyalty, based on a literature review. Based on the literature, various research hypotheses are then proposed. Section 3 establishes the research structure and related analysis methods, including the reliability, validity, and SEM test methods. Section 4 describes the questionnaire sampling and statistical analyses. The SPSS software is used for descriptive analysis, and AMOS is used for data analysis to test the research model hypotheses. This section also discusses the modification needed to meet the goodness-of-fit standard. Finally, Section 5 concludes the analysis of results and suggests future research directions.

2. Literature review This section first describes the recent development situation of the survey subject, i.e., the THSR. Constructs related to employees are then reviewed, and information related to customer constructs is collected. The employee and customer constructs are then linked to build the research hypothesis model. Finally, empirical procedure of analyzing the mediating and moderating effects is reviewed.

2.1. Overview of the THSR Since it became operational in 2007, the THSR had run 211,099 trains by December 31, 2011. The average on-time rate during that period was 99.40%. The accumulated passenger number has surpassed 157 million. Fig. 1 shows the operation stations. All THSR stations cooperate with bus companies, taxis, rental cars, temporary parking areas, parking lots, the Taiwan Railways Administration, and travel agencies to facilitate customer transfers. However, undesirable events that have occurred since operations began include a lower than expected use of the automated ticket machines, which has caused customer queues and long ticket purchase times; malfunctioning ticket machines and exit turnstiles at certain stations; and other problems such as inadequate customer safety and inadequate transfer facilities (Chou et al., 2011). Additionally, since several THSR stations in central and southern Taiwan are located in remote or undeveloped locations, the crime rate has increased significantly, which has greatly influenced the corporate operations image of the company.

Taipei

Nangang

Banciao Taoyuan Hsinchu Miaoli

Taichung

Changhua Yunlin Chiayi

Tainan

Zuoying

Legend Station Base

Fig. 1. THSR stations.

2.2. Literature related to employees 2.2.1. Correlation between LSeLe and operations performance Because of the diverse customer requirements and the intense competition among numerous enterprises in recent years, employees have become the most crucial asset of enterprises. Cultivating supervisors and other employees with leadership potential is essential for improving operations performance. Successful leadership requires a certain personality type such as an inspirational personality. People with leadership skills can inspire a group and are skilled in coordinating and persuading employees and in executing or implementing tasks (Chen and Chen, 2007). Cavazotte et al. (2011) emphasized that effective leadership not only shows the value of an employee to an enterprise, it also inspires other employees to improve organizational performance (Cavazotte et al., 2011). Additionally, employees who trust the management of an organization attribute positive behavioral intentions to the organization, which generates employee loyalty and is an authentic reflection of employee (Turkyilmaz et al., 2011). Employee satisfaction is the attitudes and emotional reflections of work satisfaction of employees after they compare their expected value with the actual value obtained (Burke et al., 2005). Therefore, we propose the following hypotheses H1 , H2 , H3 , H4 , and H5 : H1 : Leadership significantly and positively affects employee satisfaction. H2 : Leadership significantly and positively affects operations performance. H3 : Employee satisfaction significantly and positively affects employee loyalty. H4 : Employee satisfaction significantly and positively affects operations performance. H5 : Employee loyalty significantly and positively affects operations performance.

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

2.2.2. Correlation of customer satisfaction with employee satisfaction, employee loyalty, and service quality Employee loyalty inspires the subjective activity and creativity of employees needed to enable them to reach their full potential (Yee et al., 2010). Employees who are satisfied with their work extend their perceptions and emotions to customers. Customers who receive satisfactory service quality, experience, and perceptions then recommend the service to others and have increased repurchase willingness (Chi and Gursoy, 2009). Thus, hypotheses H6 , H7 , and H8 are proposed: H6 : Employee satisfaction has a significant positive effect on service quality. H7 : Employee satisfaction has a significant positive effect on customer satisfaction. H8 : Employee loyalty has a significant positive effect on service quality. 2.3. Literature related to customers 2.3.1. Correlation between corporate image and QScLc Burke et al. (2005) contended that service quality is a customer attitude and the overall measurement and evaluation of a service provider by customers. Service quality also affects corporate image (Burke et al., 2005). Customer satisfaction with service quality depends on the service quality level anticipated by the customer, the process of receiving services, and the service quality actually received and perceived by the customer. Customer satisfaction is measured by comparing expected and actual service quality levels (Finn, 2011). A study of the intercity bus service industry in Wen et al. (2005) and a study of THSR service in Chou et al. (2009, 2011) confirmed that the paired dimensions of service quality and customer satisfaction and customer satisfaction and customer loyalty are positively related (Chou and Kim, 2009; Chou et al., 2011; Wen et al., 2005). Therefore, the following hypotheses H9 , H10 , H11 , and H12 are proposed: H9 : Service quality significantly and positively affects corporate image. H10 : Service quality significantly and positively affects customer satisfaction. H11 : Corporate image significantly and positively affects customer satisfaction. H12 : Customer satisfaction significantly and positively affects customer loyalty.

2.3.2. Correlation of operations performance with corporate image and QScLc Hsu et al. (2006) indicated that companies generally use conventional methods such as advertising to affect the mental recognition of the enterprise by the customer and to affect customer anticipation. They confirmed that corporate image positively affects customer satisfaction, customer loyalty, and operating performance (Hsu et al., 2006). Yee et al. (2010) further found that creative and diverse services constitute a unique value provided by an enterprise and result in outstanding operations performance and long-term customer loyalty (Yee et al., 2010). This value enhances competitiveness and cannot be matched by competitors. Thus, hypotheses H13 , H14 , and H15 are proposed: H 13 : Corporate image significantly and positively affects operations performance. H14 :Customer satisfaction significantly and positively affects operations performance.

209

H15 : Customer loyalty significantly and positively affects operations performance. 2.4. Mediating and moderating effects 2.4.1. Mediating effect An independent variable that significantly affects a dependent variable through another construct is called a mediating variable. According to Baron and Kenny (1986), statistical methods that can be used to measure mediating effects are hierarchical regression analysis and SEM. They also proposed the following four conditions (Baron and Kenny, 1986): Condition 1: Exogenous variables significantly affect endogenous variables. Condition 2: Exogenous variables significantly affect mediator variables. Condition 3: When conducting the SEM of endogenous variables using exogenous variables and mediator variables as predicator variables, the mediator variables must significantly affect the endogenous variables. Condition 4: In the SEM of Condition 3, the regression coefficients of exogenous variables for the endogenous variables are smaller than the regression coefficients when only exogenous variables are used to predict endogenous variables. Mediating effects can be further classified as partial mediating effects and full mediating effects. Fig. 2 shows the test procedure. 2.4.2. Moderating effect Moderating variables are independent variables that directly affect the direction or strength of the relationship between another independent variable and dependent variables (Baron and Kenny, 1986). Moderating variables can be qualitative (e.g., gender or ethnicity) or quantitative. Moderating variables, which are independent variables, significantly affect dependent variables and have moderating effects on the relationship between independent and dependent variables (Ro, 2012). In the context of regression, moderating variables moderate the relationship equation between independent and dependent variables, including the direction and size (Baron and Kenny, 1986). Generally, moderating-effect models with latent variables only consider two situations: (1) a moderating variable is the categorical variable and an independent variable is the latent variable, and (2) both the moderating and dependent variables are latent variables. The regression analysis in this study first determined the average score of independent variables and then renamed the variable. The average scores for constructs that showed moderating effects in the test path were then determined. Finally, independent variables and moderating variables were standardized and multiplied to determine the values of interacting terms. A test path result that met the test standard (n o0.05, nn o0.01, and nnn o0.001) was considered an indication of a moderating effect (Baron and Kenny, 1986; Kim et al., 2009; Ro, 2012). 3. Methods 3.1. Research variables and structure This study investigated the overall operations performance of the THSR in terms of service quality and customer revisit intention. The theoretical bases were the American Customer Satisfaction Index (ACSI) combined with the findings of a review of the literature related to various THSR characteristics, including service quality, satisfaction, and customer loyalty. Nine major dimensions were established: customer recognition-related (service quality, customer satisfaction, customer loyalty, customer complaints, and corporate image); and employee cognition-related (leadership,

210

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

Test coefficient C X

Y C Test coefficients a&b

M a

b Partial mediator

X

Y C

Test coefficient C'

M

Partial Full mediator mediator

a

Stop mediator analysis

b Full mediator

X

Y Fig. 2. Mediating effect test flowchart.

employee satisfaction, employee loyalty), and operations performance. The effects of observation variables and latent constructs of these dimensions were then tested for correlations with operations performance. Mediating and moderating constructs were also considered when establishing the research structure and were used for reference in further long-term data collection. To investigate the operating performance of THSR in terms of service quality and revisit rates, the authors consulted references such as service quality measurement indices and other relevant literature to design a questionnaire for collecting data from THSR customers. The final questionnaire design included all operational definitions and measurement items of constructs. The construct development phase was performed in two steps: a literature review and an experience survey. The questionnaire was developed in two steps. The authors first reviewed the literature on operations performance before organizing constructs and question items for measuring operations performance. The scale format, item number, and categories were then determined. Prior to conducting the actual survey on site, the authors performed a pilot survey of 30 THSR customers to examine the question clarity regarding how operations performance affected their satisfaction and the THSR corporate image when the customers traveled on the THSR and received related services. The experience survey method was used to strengthen the previous item combination and to investigate these combinations regarding the items and conditions of operations performance. The results of the literature review and survey of THSR customer experience were summarized by using explorative theory constructs such as “service quality,” “customer satisfaction,” “customer loyalty,” “corporate image,” “leadership,” “employee satisfaction,” “employee loyalty,” and “operations performance.” Table 1 shows the constructs and measurement variable items.

3.2. Research hypotheses Fig. 3 shows the model of the influence that THSR customer perceptions and employee cognition had on operations performance from the perspective of passengers based on the constructs introduced above. Notably, the relationships in the structural model are based on the hypotheses described in Section 2.

3.3. Multivariate analysis 3.3.1. Reliability analysis Reliability represents the consistency and stability of measured results. Reliability also refers to the similarity between measurements of the same object or similar objects for two or more iterations. Since a high reliability indicates a high stability, the measured results are considered reliable (Joseph et al., 2010). Thus, Eq. (1) was used for the Cronbach α reliability test of reliability between variables.
 ∑s2 k 0 Cronbach S α ¼ 1  2i ð1Þ k1 s where s2i is the variance of each index, s2 is the total variance of the measured index, and k is the number of indices of the measured construct. 3.3.2. Validity analysis Validity is the degree to which researchers can actually measure the object under investigation. Restated, validity measures whether the measured items approach the object attributes being measured during the measurement process. A high validity measurement indicates that the measured result is close to the characteristics of the object being measured. Construct validity can be either convergent or discriminant. Unlike convergent validity, discriminant validity, which is also referred to as divergent validity, exists when multiple indices of a dimension congregate or correspond to each other. They typically have low correlation with measurement indices of the opposing dimension. Validity is often measured in terms of factor loadings, average variance extracted (AVE), and composite reliability (CR) (Fornell and Larcker, 1981). Factor loadings, which are obtained by factor analysis, mainly test the consistency in indices and dimensions. The variance extracted measures the variance of measurement dimensions when extracting variables. Eq. (2) shown below is used to evaluate the explanatory power of each measurement variable regarding the variance of a certain latent variable: AVE ¼

∑ni¼ 1 λ2i n

ð2Þ

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

211

Table 1 Construct and measurement variable definitions. Research construct

Sub-construct

Measurement variable

Leadership

Personnel management

The THSR directs employees in a timely manner. The THSR gives adequate authorization to its employees. The THSR develops various marketing strategies. The THSR hosts promotional activities. The THSR uses information technology appropriately. The THSR analyses operational conditions using statistical data. The THSR generates innovative and impressive ideas. The THSR always adopts appropriate decorations according to festivals and holidays. The THSR provides satisfactory welfare systems. The THSR provides satisfactory recreation and entertainment benefit and preference systems. The THSR provides high-speed rail employees with travel benefits and preference systems. The THSR provides basic etiquette training to cultivate service skills. The THSR invests sufficient resources and effort in employee training plans. The THSR provides employee risk management training. The THSR promotes employees and increases the salaries of employees who show exceptional performance. The THSR promotes employees and adjusts salaries based on periodic performance evaluations. Sense of accomplishment from work Recognition of the work service content Employee suggestions are valued by THSR managers. The THSR treats all employees fairly. The THSR has high employee loyalty OR has high employee morale. Employees are proud to work at THSR. Employee suggestions are valued by THSR managers. The THSR treats all employees fairly. The air conditioning in the carriages is comfortable. The noise decibel level in the carriage is comfortable. The THSR service staffs always provide services in a pleasant manner. The THSR service staffs are willing to provide additional services. Seats in the carriage are spacious, comfortable, and clean. The directional signs at the station are easy to understand. The cleanliness of the stations is satisfactory. The station locations are convenient Transfer services are convenient Schedule information is easily available The THSR actively provides customers with correct information. The THSR rapidly and effectively responds to questions from customers. The THSR has various complaint channels. Using THSR for transportation is the right decision. Selecting the THSR is comfortable and safe. The THSR provides services that meet customer expectations and requirements. The THSR ticket prices are reasonable. The THSR handles customer complaints appropriately. The THSR has satisfactory complaint processing procedures. The THSR listens to customer complaints and accepts complaints in an open-minded manner. The THSR responds to complaints by implementing improvement measures.

Strategic planning Information analysis Innovative changes Employee loyalty

Employee welfare

Employee training

Encouragement and development Work environment Employee satisfaction

–

Service quality

Comfort of service

Overall environment

Convenience

Response ability

Customer satisfaction

–

Customer loyalty

–

The THSR listens to customer suggestions I will use THSR the next time I travel. I will recommend the THSR to family and friends. Despite the increased THSR ticket prices, I am still willing to use the THSR.

Corporate image

–

Operations performance

–

The THSR punctual at departure and arrival. The THSR provides a service schedule that satisfies time requirements. The THSR projects a positive corporate image. The THSR operating growth rate has already increased significantly. The THSR profit rate has already increased significantly. The THSR has a complete organization structure. The THSR develops new markets continuously. The THSR values the latent after-sales requirements of customers. All THSR departments cooperate effectively. The THSR values employee opinions about the company. The THSR improves the professional abilities of employees through education and training. The THSR establishes an active organizational atmosphere for employees.

where AVE is the average variance extracted, λ represents the factor loadings of each index for the dimension, and n is the no. of measurement items. A high factor loading indicates that the given index adequately explains the measured dimension. According to Fornell and Larcker, the standard AVE value must exceed 0.5 (Fornell and

Larcker, 1981). The composite reliability or CR value of latent variables comprises the reliabilities of all measured variables, which indicate the internal consistency of the construct indices. A high CR indicates a high internal consistency of the latent variables. Fornell et al. (1996) suggested that the standard CR value should exceed 0.6 (Fornell et al., 1996) whereas the

212

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

Employee Loyality

Service Quality

Corporate Image

Employee Satisfaction

Customer Satisfaction

Operations Performace

Leadership

Customer Loyality

Fig. 3. Original research model.

threshold value recommended by Hair (2010) is 0.7. Eq. (3) shows the calculation. CR ¼

ð∑ni¼ 1 λi Þ2

ð3Þ

ð∑ni¼ 1 λi Þ2 þ ð∑ni¼ 1 ei Þ

where CR is the composite reliability, λ denotes the factor loadings of each index for the construct, and e represents the error variance term for the measured variable. Discriminant validity is defined as a low correlation between two measured constructs; in this case, both constructs are viewed as having discriminant validity. The determinant principle is that the AVE of each construct exceeds the squared correlation coefficients between the constructs. 3.3.3. Structural equation modeling Structural equation modeling integrates characteristics of path analysis, confirmatory factor analysis (CFA), and structural regression models. In SEM, latent variables are not directly observable. That is, further analysis requires explicit indicators. Latent variables in the causal model can also be divided into exogenous and endogenous variables. Exogenous variables are variables that are not included in the causal model but still influence effects of other variables. Endogenous variables are variables influenced by exogenous variables or other system variables. The two models used for SEM analysis are the measurement model and structural model. The measurement model establishes linear relationships between explicit and latent variables whereas the structural model defines the linear relationship between latent independent and latent dependent variables. Additionally, the structural model can simultaneously complete the measurement problems and causal relationship equations of the evaluation system and solve measurement errors. The measurement model equation is as follows: X ¼ Λx ξ þ δ

ð4Þ

Y ¼ Λy η þ ε

ð5Þ

This form can also be transformed into the following matrix: " # " # X1

¼ ½Λx qn  ½ξn1 þ

⋮ Xq

"

Y1 ⋮ Yp

q1

#

δ1 ⋮ δq

ð6Þ q1

" # ¼ ½Λy pm  ½ηm1 þ p1

ε1 ⋮ εp

ð7Þ p1

Eq. (7) is the structural model equation: η ¼ Bη þ Γξ þ ζ This general form can also be transformed into a matrix as follows: " # " # " # η1 ζ1 Bmm ⋮ ¼ ð8Þ  ½ηm1 ξn1  þ ⋮ Γ mn ηm ζm m1

m1

In the above equations, X is the row vector for q observation variables of ξ; Λx is the regression coefficient matrix (q  m) for exogenous constructs; ξ is the row vector for n exogenous constructs; δ is the row vector for the measurement error in q observation variables; Y is the row vector for p observation variables of η; Λy is the regression coefficient matrix (p  m) for endogenous constructs; η is the row vector for m endogenous constructs; ε is the row vector for measurement error in p observation variables; η is the row vector for m endogenous constructs; ξ is the row vector for n exogenous variables; B is the regression coefficient matrix (m  m) for endogenous variables; Γ is the regression coefficient matrix (m  n) for exogenous variables; and ζ is the explained residual term row vector for endogenous variables.

4. Research analysis and results 4.1. Sampling method and descriptive analysis This study used the questionnaire survey method to explore passenger travel behavior and then constructed a model of customer requirements and expectations from the THSR. The research subjects were passengers who had used one of the eight THSR stations (Taipei, Banqiao, Taoyuan, Hsinchu, Taichung, Chiayi, Tainan, and Zuoying). The actual questionnaire survey was performed by distributing 279 paper questionnaires throughout Taiwan. Of these, 261 valid questionnaires were recovered. In addition to the 31 valid online questionnaires recovered, the analysis included 292 valid questionnaires. The valid return rate was 94.19%. The questionnaire design was based on the composite item scale (1–7) used in previous studies. The questionnaire was also used to test the proposed hypotheses. Table 2 shows the results of descriptive statistical analysis of the THSR passenger data for socioeconomic background and characteristic, which were derived from the questionnaire survey of THSR passengers. Regarding the gender of the THSR customers,

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

213

Table 2 Socioeconomic characteristics of THSR customers. Item

Description

Distribution

Percentage (%)

Gender

Male Female 18 years and below 19–25 years 26–32 years 33–39 years 40–46 years 47–53 years 54–60 years 61 years and above Elementary school and below Junior high school Senior high school or vocational school Associate degree or bachelor degree Graduate degree Student Commercial and service industries Industrial and manufacturing industries Soldiers, public servants, and teachers Home caretakers Agricultural, forestry, fishery, husbandry, and mining industries Freelancer Retired people Others US$ 0–500 US$ 500–1,000 US$ 1,000–1,667 US$ 1,667–2,333 US$ 2,333 and above

151 141 12 67 46 65 39 28 23 12 2 30 56 138 66 45 75 49 38 23 16 17 14 15 82 45 89 51 25

51.7 48.3 4.1 22.9 15.8 22.3 13.4 9.6 7.9 4.1 0.7 10.3 19.2 47.3 22.6 15.4 25.7 16.8 13.0 7.9 5.5 5.8 4.8 5.1 28.1 15.4 30.5 17.5 8.6

Age

Education level

Career

Average monthly income

a slight majority (51.7%) were male. The age distribution of the respondents was mainly between 19 and 46 (74.4%). The data showed that most respondents were young to middle-aged, and most had intentions to travel via THSR in the future. Most (69.9%) respondents had a bachelor degree or above. This indicated that people who use the THSR and people who have intention to use the THSR tend to be well educated. Regarding employment, most were working in the business or commercial and services industries (25.7%) followed by those employed in the industrial and manufacturing industry (16.8%), students (15.4%), and those employed as soldiers, public servants, and teachers (13.0%). These data indicated that most THSR customers were involved in mobile business affairs and had jobs requiring frequent commuting. Finally, most customers had an average monthly income of either US$ 1,000 to US$ 1,667 (30.5%) or below US$ 5,000 (28.1%).

4.2. Confirmatory analysis To facilitate the subsequent SEM evaluation, CFA was first used to test the reliability and validity of correlations between latent variables and measurement variables. The method typically used to test the reliability of each construct is the Cronbach α coefficient (Joseph et al., 2010). Construct validity is tested using the factor loadings mentioned above, AVE, and CR coefficients. Table 3 shows that, except for service quality (0.436), customer satisfaction (0.438), and customer loyalty (0.444), the AVE values for all constructs exceeded 0.5. Since the objectives were to describe significant variables and to test the equilibrium of hypothesis models, observation variables with low factor loadings were individually deleted. An index was removed if doing so effectively increased the corresponding construct reliability and validity. Table 3 shows the results after indexes were deleted. The AVE for the service quality construct increased from 0.436 to 0.565. The AVE for the customer satisfaction construct increased

from 0.450 to 0.550. The AVE for customer loyalty increased from 0.441 to 0.591. The AVE for all other constructs exceeded 0.5. Discriminant validity compares two latent variable constructs to test for discrimination between the constructs (Fornell and Larcker, 1981). Table 4 shows that the AVE values for employee satisfaction and employee loyalty were 0.544 and 0.589, respectively, which were both smaller than the correlation coefficient for the two constructs, which was 0.815 (0.664 for squared value). The AVE of customer satisfaction and customer loyalty were 0.550 and 0.591, respectively, which were smaller than the correlation coefficient of the two constructs, which was 0.838 (0.702 for squared value). Table 4 shows that these constructs were not discriminant. Thus, the modified model combined the four constructs into pairs (Fig. 4). Regarding the correlation coefficients between all constructs, correlation coefficients exceeding 0.7 indicated a strong correlation, those between 0.3 and 0.7 indicated a moderate correlation, and those smaller than 0.3 represented a low correlation. Table 4 shows that all correlation coefficients exceeded 0.7, which indicated high correlations between leadership and employee satisfaction, between employee satisfaction and employee loyalty, between employee loyalty and operations performance, between service quality and customer satisfaction, between service quality and corporate image, and between corporate image and customer satisfaction. The coefficients of other constructs ranged from 0.3 and 0.7, which indicated a moderate correlation (Table 4). 4.3. Model fit and modification A parameter estimation of the original model showed that the data had an unfavorable goodness-of-fit to the hypotheses. Typically, researchers improve the goodness-of-fit of models by increasing or removing the paths and variables among the initial model variables. Possible explanations for the poor goodness-of-fit in the model were violation of the basic distribution settings, missing values, sequential errors, or a nonlinear relationship. The

214

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

Table 3 Modified confirmatory factor analysis. Construct

Measurement variable

λ

μ

AVE

CR

α

Leadership

Direction of employees in a timely manner (L_1) Full authorization of employees (L_2) Development of various marketing strategies (L_3) Promotional activities (L_4) Use of information technology appropriately (L_5) Use of information statistics (L_6)

0.861

0.892

Additional welfare systems (ES_1) Recreation and entertainment benefit and preference system (ES_2) Travel benefits and preference system (ES_3) Basic etiquette training to cultivate employee skills (ES_4) Sufficient resources and effort in training employees (ES_5) Employee risk management training (ES_6) Promotion based on exceptional employee performance (ES_7)

0.544

0.882

0.914

Employee loyalty

Suggestions are valued (EL_1) All employees are treated fairly (EL_2) Employees wish to always remain a part of the THSR company (EL_3) I am an employee of the THSR (EL_4) Operating amount growth rate increases significantly (OP_1) Profit rate increases significantly (OP_2) New markets are developed continuously (OP_4) Improve professional abilities through training (OP_8)

0.589

0.826

0.855

0.559

0.827

0.900

0.733 – – – 0.731 0.729

4.78 4.65 4.79 4.90 4.87 4.67 4.82 4.72 4.91 4.85 4.93 5.04 4.99 4.94 4.88 4.88 4.82 4.83 4.75 4.69 4.70 4.71 4.82 4.76 4.82 4.64 4.71 4.64 4.55 4.68 4.64

0.536

Employee satisfaction

0.755 0.710 0.720 0.70 0.762 0.745 – – 0.710 0.714 0.708 0.745 0.787 0.792 0.702 – – – 0.787 0.778 0.770 0.734 0.760 0.783

0.750 0.701 0.745 – 0.821 0.754 0.783 – – – – – – – – 0.700 0.748 0.737 0.722 0.753 0.768 – – – 0.702 – – 0.542 0.761 – –

5.25 5.07 5.10 4.75 5.25 5.26 5.36 4.70 4.70 4.99 4.95 4.93 5.01 5.06 5.29 5.10 5.23 5.16 5.27 5.25 5.11 5.11 4.36 4.79 4.74 4.79 4.76 4.63 4.98 4.92 4.46

0.565

0.868

0.923

0.542

0.761

0.779

0.550

0.737

0.738

0.591

0.73

0.742

Operations performance

Create an active atmosphere (OP_9) Service quality

Comfortable air conditioning (SQ_1) Comfortable noise decibel levels (SQ_2) Provide services with a pleasant manner (SQ_3) Comfortable and clean seats (SQ_5) Understandable directional arrangements (SQ_6) Cleanliness (SQ_7)

Corporate image

Customer satisfaction

Amicable service staff (SQ_16) Departure punctuality (CI_1) Schedule services meet customer requirements (CI_2) The THSR corporate image (CI_3) Selecting the THSR is the right decision (CS_1) Traveling by THSR, is comfortable and safe (CS_2)

Complaint processing procedures are satisfactory (CC_6)

Customer loyalty

Selecting the THSR is the right decision (CS_1) Customers traveling by THSR, feel comfortable and safe (CS_2)

literature generally agrees that model reconstruction should not only accommodate theoretical legitimacy, but should also establish the reliability and validity of the newly reconstructed model (Joseph et al., 2010). After checking discriminant validity, employee satisfaction and employee loyalty were combined into the employee cognition construct, and customer satisfaction and customer loyalty were

combined into the customer recognition construct to facilitate further SEM analysis. Further reliability and validity tests then obtained favorable results for all indices and constructs. Fig. 4 shows that the modified model comprised one endogenous variable, i.e., leadership ðη1 Þ, and five exogenous variables i.e., employee cognition ðξ1 Þ, operations performance ðξ2 Þ, service quality ðξ3 Þ, corporate image ðξ4 Þ, and customer recognition ðξ5 Þ.

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

215

Table 4 Correlation matrix for latent variables. Constructs

Service quality

Customer satisfaction

Customer loyalty

Corporate image

Leadership

Employee satisfaction

Employee loyalty

Customer satisfaction Customer loyalty Corporate image Leadership Employee satisfaction Employee loyalty Operations performance

0.731 0.671 0.728 0.375 0.490 0.585 0.503

─ 0.815 0.709 0.467 0.610 0.631 0.622

─ 0.578 0.380 0.497 0.514 0.507

─ 0.273 0.357 0.426 0.379

─ 0.766 0.642 0.684

─ 0.838 0.693

─ 0.723

Employee Cognition

Service Quality

Corporate Image

Leadership

Operations Performance

Customer Recognition

Fig. 4. Modified research model.

Table 5 Path evaluation results for research model. Research path

Path level of coefficient significance

H1 leadership-employee cognition H2 leadership-operations performance H3 employee cognitionoperations performance H4 employee cognition-service quality H5 employee cognition-customer recognition H6 service quality-corporate image H7 service quality-customer recognition H8 corporate image-customer recognition H9 corporate image-operations performance H10 customer recognitionoperations performance

0.668

nnn

Yes

0.075

0.406

No

0.365

nnn

Yes

0.794

nnn

Yes

0.325

nnn

Yes

0.697

nnn

Yes

0.430

nnn

Yes

0.179

0.027nn

Yes

0.334

0.037

nn

Yes

0.497

nnn

Note:

n

Hypothesis supported

To evaluate the capability of the established theory model to explain the observed data, the following five goodness-of-fit measurement indices were used as the determination standard: χ (a) Chi-square/degree of freedom ratio ðdof Þ 2 The chi-square test value ðχ Þ was used to test the goodnessof-fit of the proposed theoretical model to the observed data. Unlike conventional statistical methods, the model has a desirable goodness-of-fit when the null hypothesis is accepted. A high χ 2 indicates a poor goodness-of-fit of the theoretical models to the actual data. However, χ 2 is extremely sensitive to sample size. When the sample size is large, χ 2 increases relatively and easily rejects null hypotheses. Conversely, if the sample size is small, changes in χ 2 do not tend to reach statistical significance and therefore do not support the null hypothesis. Generally, 150–400 is an appropriate sample size (Kline (2011); Hair et al. (2010); Byrne (2001)). (b) Goodness-of-fit index (GFI) Like the explained variance (R2) used in statistical regression analysis (Wallgren and Hanse, 2007), GFI indicates the capability of the hypothesis model to explain the ratio of variance and covariance in the actual data. When the GFI value approaches one, the goodness-of-fit of the model increases; otherwise, the goodness-of-fit of the model declines. The relevant equation is Eq. (9). 2

Yes

indicates the value reached the 0.1 level of significance.

nn

indicates the value reached the 0.05 level of significance. nnn indicates the value reached the 0. 01 level of significance.

Table 5 shows the results of CFA to remove non-significant paths identified by SEM. Since H2 , which hypothesized that leadership has a statistically significant influence on operations performance, did not achieve the significance level of 0.1, it was removed.

trðs^ W _ sÞ trðs′WsÞ 0

GFI ¼

0

ð9Þ

where trðs WsÞ is the theoretically replicated sum of the weighted 0 variance, trðs^ W _ sÞ is the total sum of the weighted variance

216

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

when covariance is replicated through actual sample observation, and W is the weighted matrix. (c) Incremental fit index (IFI) The IFI is a structural equation GFI. The calculation method uses other models as the reference point to compare and evaluate the goodness-of-fit of the proposed hypothesis model to the empirical data. Generally, an IFI approaching 1 indicates that the model has a desirable goodness-of-fit (Joseph et al., 2010). The equation is expressed as Eq. (10). χ 2indep  χ 2test 2 χ indep  dof test where χ 2indep is the IFI ¼

Table 6 Fitness determination standards and output values. Index

Suggested requirement standard

Result

χ 2 =df GFI CFI IFI RMSEA

o3 40.8 40.9 40.9 o 0.1

1.893 0.833 0.920 0.921 0.055

ð10Þ

chi-square value of the null model and χ 2test is the chi-square value of the proposed model. (d) Comparative fit index (CFI) This index indicates not only the difference between the proposed model and the null model, but also the divergence of the tested model from the central chi-square distribution. The calculation method uses the ratio of improvement in noncentrality to obtain a non-centrality parameter. Because the null model is the least desirable model, any hypothesis model can be superior to the null model in terms of goodness-of-fit. A CFI value approaching 1 is the most desirable and indicates that non-centrality can be improved (Bagozzi and Yi, 1988). (e) Root mean square error of approximation (RMSEA) The RMSEA measures the difference regarding the observed covariance matrix and the estimated covariance matrix per each degree of freedom (dof). A low RMSEA value is desirable. Values for RMSEA lower than 0.10, 0.08, and 0.05 indicate acceptable, reasonable, and favorable model goodness-of-fit, respectively (Joseph et al., 2010). Eq. (11) is used to calculate RMSEA. sffiffiffiffiffiffiffiffi χ 2  dof F^ 0 ^ RMSEA ¼ ð11Þ ; F0 ¼ n dof where F^ 0 is the estimated value of the discrepancy function obtained by fitting a model to the population moments; χ 2 denotes the chi-square; n is the sample size. Table 6 shows the output values of the modified model goodness-of-fit. Since the output values meet the standard requirements, the goodness-of-fit of the modified model in this study is considered adequate. 4.4. Examining mediating and moderating effects 4.4.1. Mediating effect test The modified model revealed six sets of models with mediating effects. The following tests were performed for these models: (a) Leadership-employee cognition-service quality (L-CE-Q) Fig. 5 shows that, after the mediating variable of employee cognition was included in the tested L-CE-Q model, the path coefficient of leadership to service quality declined from 0.595 to 0.182 and was not significant. This confirmed that employee cognition mediates the effect of leadership on service quality. (b) Employee cognition-customer recognition-operations performance (CE-RC-P) Likewise, after the mediating variable of customer recognition was included in the tested CE-RC-P model, the path coefficient of customer recognition to operations performance was 0.606 and the t value was 0.486, which was not statistically significant. That is, the customer recognition construct showed no mediating effects in this model. (c) Employee cognition-service quality-corporate image (CE-Q-I) The test result shows that including the mediating variable of service quality in the tested CE-Q-I model slightly decreased

Leadership

Service Quality

Leadership

Employee Cognition

Leadership

Employee Cognition

Service Quality

Fig. 5. L-CE-D mediating effect test flowchart.

the path coefficient of employee cognition to corporate image from 0.636 to 0.618. This confirmed that the effect of employee cognition on corporate image is partially mediated by service quality. (d) Service quality-corporate image-customer recognition (Q-I-RC) After including the mediating variable of corporate image in the tested Q-I-RC model, the path coefficient of service quality to operations performance moderately decreased from 0.884 to 0.635. This confirmed that the effect of service quality on customer recognition is partially mediated by corporate image. (e) Customer recognition-corporate image-operations performance (RC-I-P) The test result shows that including the mediating variable of corporate image in the tested RC-I-P model reduced the path coefficient of service quality for operations performance from 0.595 to 0.162, which was not a statistically significant. This confirmed that the effect of customer recognition on operations performance is mediated by corporate image. (f) Service quality-corporate image-operations performance (Q-I-P). After including the mediating variable of corporate image in the tested Q-I-P model, the path coefficient of service quality for operations performance decreased from 0.501 to 0.127, but the decrease was not statistically significant. This confirmed that corporate image mediates the effect of service quality on operations performance.

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

4.4.2. Moderator effect test The modified models revealed four paths with possible moderating effects. Tests for each model are shown below. (a) Corporate image-customer recognition-operations performance (I-RC-P) The R2 of Table 7 is the coefficient of determination, which represents the explanatory power of the regression model. The ΔR2 represents the difference of R2 among the models. A statistically significant increase in ΔR2 indicates that including new variables increases the explanatory power of the model. Table 7 shows that including the corporate image variable in Model II significantly increased the total explained variance of operations (ΔR2 ¼0.044; p-value ¼0.000), which indicated that including the service quality variable in the model was statistically meaningful. Including the interaction term for customer recognition and corporate image in Model III also significantly increased the total explained variance of operations performance (ΔR2 ¼ 0.020; p-value ¼0.002). This indicated that corporate image has a moderating effect on the relationship between customer recognition and operations performance. (b) Employee cognition-service quality-customer recognition (CE-Q-RC) Similarly, after including the employee cognition variable in Model II, the total explained variance of customer recognition significantly increased (ΔR2 ¼0.079; p-value¼0.000), which confirmed that including the employee cognition variable in the model significantly increases its explanatory power. Including the interaction term for service quality and employee cognition in Model III did not affect the total explained variance of operations performance (ΔR2 ¼0.000; p-value¼0.684). This indicated that employee cognition does not moderate the relationship between service quality and customer recognition. (c) Service quality-corporate image-customer recognition (Q-I-RC) The test result shows that including the service quality variable in Model II significantly increased the total explained variance of customer recognition (ΔR2 ¼0.177; p-value¼0.000). That is, including the service quality variable was meaningful. However, including the interaction term for corporate image and service quality in Model III did not significantly increase the total explained variance of operations performance (ΔR2 ¼0.002; p-value¼0.915). This indicated that employee cognition does not moderate the relationship between corporate image and customer recognition. Employee cognition-customer recognition-operations performance (CE-RC-P)After including employee cognition in Model II, Table 7 Moderator regression analysis of corporate image for customer recognition and operations performance. Dependent variable

Operations performance

Statistics Independent variable

Model I

Independent variable Customer recognition Moderator variable Corporate image Interaction variable Customer recognition  corporate image R2 ΔR2 F p-value: n o 0.05, nnn

o 0.001.

nn

o 0.01.

0.581nnn

0.337 0.337 147.632nnn

Model II

Model III

0.418nnn

0.403nnn

0.266nnn

0.273nnn

0.381 0.044 20.614nnn

0.141 0.401 0.020 9.506nnn

217

the total explained variance of operations performance significantly increased (ΔR2 ¼0.285; p-value¼0.000). This indicated that including the employee cognition variable in the model was meaningful. Including the interaction term for customer recognition and employee cognition in Model III did not increase the explained variance of operations performance (ΔR2 ¼ 0.001; pvalue¼0.365). This indicated that employee cognition does not have a moderating effect on the relationship between customer recognition and operations performance. Table 8 summarizes the outcomes of mediating and moderating effects from the above analysis paths. 5. Conclusions This study explored THSR operations performance from the customer perspective. The focus of the study was the construct relationships related to customers and employees, including LSeLe and QScLc. The theoretical basis of these constructs was then used to establish a structural model of effects on customer perspectives of THSR operations performance. A tangible questionnaire and online survey methods were used for data collection. The SEM was used for empirical analysis of the validity and reliability of the proposed model. Mediating and moderating effects of model constructs were also analyzed. Based on the initial analysis, the proposed model was modified by combining employee satisfaction with employee loyalty and by combining customer satisfaction with customer loyalty. Finally, 10 paths were constructed as hypotheses. Tests of statistical significance revealed the following nine paths. Path 1: Leadership significantly and positively affects employee cognition. Path 3: Employee cognition significantly and positively affects operations performance. Path 4: Employee cognition significantly and positively affects service quality. Path 5: Employee cognition significantly and positively affects customer recognition. Path 6: Service quality significantly and positively affects corporate image. Path 7: Service quality significantly and positively affects customer recognition. Path 8: Corporate image significantly and positively affects customer recognition. Path 9: Corporate image significantly and positively affects operations performance. Path 10: Customer recognition significantly and positively affects operations performance.

A research model was also built to evaluate leadership, employee satisfaction, employee loyalty, service quality, corporate image, customer satisfaction, and customer loyalty for causal effects on operations performance from a customer perspective. A confirmatory factor analysis of model constructs and measurement variables and a path analysis of the causal relationships among constructs were then performed to restructure the model by eliminating constructs and by combining constructs and measurement variables of the original hypothesis model. The analysis results indicated that leadership significantly correlated with employee cognition (0.752); employee cognition significantly correlated with service quality and operations performance (0.667 and 0.724, respectively); service quality significantly correlated with corporate image (0.755); however, operations performance had only weak correlations with service quality, corporate image, and customer recognition (0.043, 0.078,

218

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

Table 8 Test table of mediating and moderating effects. Analysis path

Mediator / moderator construct

Mediator effects

Moderator effects

Leadership–employee cognition–service quality Employee cognition–customer recognition–operations performance Employee cognition–service quality–corporate image Service quality–corporate image–customer recognition Corporate image–customer recognition–operations performance Service quality–corporate image–operations performance Corporate image–customer recognition–operations performance Employee cognition–customer recognition–operations performance Employee cognition–service quality–customer recognition Service quality–corporate image–customer recognition

Employee cognition Customer recognition Service quality Corporate image Customer recognition Corporate image Corporate image Employee cognition Employee cognition Service quality

Y N Y Y Y Y – – – –

─ ─ ─ ─ ─ ─ Y N N N

and 0.373, respectively). This indicates that rather concerning operation profits, customers actively consider whether management actualizes internal plans, arrangements, and adjustments and whether management comprehensively understands customer expectations about services, customer experiences, and customer perceptions. Additionally, the data indicate that these capabilities can increase customer acceptance. Notably, the SEM and regression analysis results showed that customer recognition had significant mediating effects on the association between corporate image and operations performance; and corporate image had significant mediating and moderating effects on the association between service quality and customer recognition/operations performance. Employee cognition had mediating effects on the association between leadership and operations performance. Service quality had mediating effects on the association between employee cognition and corporate image. Corporate image had mediating effects on the association between service quality and operations performance. By targeting the THSR, the questionnaire survey in this study revealed how causal relationships affect operations performance from the customer perspective. To enhance the professional and systemic management capabilities of the THSR, the following research paths are recommended for further studies. (a). Here, the research subjects were THSR passengers. Due to sampling difficulties in the paper-based questionnaire survey, non-systematic errors occurred when small numbers of older respondents (aged over 65 years) were surveyed, i.e., the sampled participants might not be evenly distributed. Larger samples are recommended in future studies to obtain a more reliable result. (b). The proposed model discusses the influence of related constructs on THSR operations performance from the customer perspective. Further studies are needed to develop a more comprehensive research model by considering both employee and customer perspectives. (c). The sample in this study was limited to THSR passengers. For model comparisons, future studies can sample high-speed rail customers in other countries such as China, South Korea, and Japan. A more comprehensive analysis can also be obtained by evaluating whether the relationship of model constructs among various countries differs in terms of influence and variation. (d). Future studies can also reapply the proposed model to obtain periodic measurements. Researchers can use the four major constructs of employee cognition, service quality, corporate image, and customer recognition to examine whether constructs improve significantly over time, to determine their long-term effects on operation profits, and to achieve the goal of sustainable management. Deployment on how the

analytical results can be applied in practice would be another valuable research direction. (e). Finally, this study did not consider the socioeconomic characteristics of THSR customers or their reasons, processes, riding frequency, and purposes for using THSR services. Therefore, future studies are needed to categorize these characteristics and to identify additional factors that affect customer use of these services.

References Anderson, E.W., Fornell, C., 2000. Foundations of the American customer satisfaction index. Total Quality Management 11 (7), 869–882. Kline, R., 2011. Principles and Practice of Structural Equation Modeling, Third Edition The Guilford Press, New York. Bagozzi, R.P., Yi, Y., 1988. On the evaluation of structural equation models. Journal of the Academy of Marketing Science, 74–94. Baron, R.M., Kenny, D.A., 1986. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology 51, 1173–1182. Burke, R.J., Graham, J., Smith, F., 2005. Effects of reengineering on the employee satisfaction-customer satisfaction relationship. TQM Magazine 17 (4), 358–363. Cavazotte, F., Moreno, V., Hickmann, M., 2011. Effects of leader intelligence, personality and emotional intelligence on transformational leadership and managerial performance. The Leadership Quarterly. Chen, J.-K., Chen, I.-S., 2007. The Relationships Between Personal Traits. Leadership Styles, and Innovative Operation. Chi, C.G., Gursoy, D., 2009. Employee satisfaction, customer satisfaction, and financial performance: an empirical examination. International Journal of Hospitality Management 28 (2), 245–253. Chou, J.-S., Kim, C., 2009. A structural equation analysis of the QSL relationship with passenger riding experience on high speed rail: an empirical study of Taiwan and Korea. Expert Systems with Applications 36 (3, Part 2), 6945–6955. Chou, J.-S., Kim, C., Kuo, Y.-C., Ou, N.-C., 2011. Deploying effective service strategy in the operations stage of high-speed rail. Transportation Research Part E: Logistics and Transportation Review 47 (4), 507–519. Finn, A., 2011. Investigating the non-linear effects of e-service quality dimensions on customer satisfaction. Journal of Retailing and Consumer Services 18 (1), 27–37. Fornell, C., Johnson, M.D., Anderson, E.W., Cha, J., Bryant, B.E., 1996. The American customer satisfaction index: nature, purpose, and findings. Journal of Marketing 60 (4), 7–18. Fornell, C., Larcker, D.F., 1981. Evaluating structural equations models with unobservable variables and measurement error. Journal of Marketing Research 18, 39–50. Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., 2010. Multivariate Data Analysis: A Global Perspective, Seventh Edition Pearson Education International, New Jersey. Hsu, S.H., Chen, W.H., Hsueh, J.T., 2006. Application of customer satisfaction study to derive customer knowledge. Total Quality Management and Business Excellence 17 (4), 439–454. 〈http://www.npf.org.tw/post/2/5668〉, National Policy Foundation Research Report. Jia, L., Ping, Z., 2005. Constructing quality-satisfaction-loyalty chain system: an empirical study of insurance industry in China, Services Systems and Services Management, 2005. Proceedings of ICSSSM '05. 2005 International Conference on. Joo, Y.G., Sohn, S.Y., 2008. Structural equation model for effective CRM of digital content industry. Expert Systems with Applications 34 (1), 63–71. Joseph, F., Hair, J., Black, W.C., Blain, B.J., Anderson, R.E, 2010. Multivariate Data Analysis. Pearson International, New Jersey.

J.-S. Chou, C.-P. Yeh / Transport Policy 30 (2013) 207–219

Kim, B.P., Murrmann, S.K., Lee, G., 2009. Moderating effects of gender and organizational level between role stress and job satisfaction among hotel employees. International Journal of Hospitality Management 28 (4), 612–619. Lin, W.-B., 2007. An empirical of service quality model from the viewpoint of management. Expert Systems with Applications 32 (2), 364–375. Byrne, B.M., 2001. Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming. Lawrence Erlbaum Associates. Ro, H., 2012. Moderator and mediator effects in hospitality research. International Journal of Hospitality Management 31 (3), 952–961. Turkyilmaz, A., Akman, G., Ozkan, C., Pastuszak, Z., 2011. Empirical study of public sector employee loyalty and satisfaction. Industrial Management and Data Systems 111 (5), 675–696.

219

Wallgren, L.G., Hanse, J.J., 2007. Job characteristics, motivators and stress among information technology consultants: a structural equation modeling approach. International Journal of Industrial Ergonomics 37 (1), 51–59. Wen, C.-H., Lan, L.W., Cheng, H.-L., 2005. Structural equation modeling to determine passenger loyalty toward intercity bus services. Transportation Research Record 1927, 249–255. Yee, R.W.Y., Yeung, A.C.L., Edwin Cheng, T.C., 2010. An empirical study of employee loyalty, service quality and firm performance in the service industry. International Journal of Production Economics 124 (1), 109–120.