Application of structural equation modeling to evaluate the intention of ...

3 downloads 98802 Views 423KB Size Report
(Dunn et al., 1994; Stank et al., 2001), marketing .... protection in business-to-business (B2B) Internet ... market, Internet services are valuable for global car-.
European Journal of Operational Research 180 (2007) 845–867 www.elsevier.com/locate/ejor

O.R. Applications

Application of structural equation modeling to evaluate the intention of shippers to use Internet services in liner shipping Chin-Shan Lu a

a,*

, Kee-hung Lai b, T.C.E. Cheng

b

Department of Transportation and Communication Management Science, National Cheng Kung University, No. 1, University Road, Tainan City 701, Taiwan, ROC b Department of Logistics, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Received 5 March 2005; accepted 2 May 2006 Available online 21 June 2006

Abstract Operations research (OR) is the application of modeling techniques to formulate and analyze systems and problems for management decision-making. Structural equation modeling (SEM) is a modeling technique applied to social or behavioral systems to understand and explain relationships that may exist among elements of systems. Recently, the measurement of unobservable variables has gained increasing attention in operations management (OM) research, and the OR discipline has begun to recognize the value of applying SEM to analyze behavioral-related OR problems. To provide OR researchers with a better understanding of the application of this useful statistical modeling technique, this paper presents a tutorial on the application of SEM. Specifically, we investigate the key factors that affect the adoption of Internet services in the context of liner shipping services. Although [Fishbein, M.A., Ajzen, I., 1975. Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research, Addison-Wesley, Reading, MA; Davis, F.D., 1989. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly 13 (3), 319–339; Ajzen, I., 1985. From intention to actions: A theory of planned behavior. In: Kuhl, J., Bechmann, J. (Eds.), Action Control: From Cognition to Behavior. Springer Verlag, New York, pp. 11–39; Ajzen, I., 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes 50, 179–211] have made important contributions to understanding users’ behavior of technology acceptance, shippers’ resistance to end-user systems is still a common problem in the liner shipping industry. To better predict, explain, and increase shippers’ acceptance of technology, we need to understand why shippers accept or reject Internet services provided by their liner shipping carriers. Another objective of this paper is to propose and empirically test a theoretical framework that relates the intention of shippers to use Internet services in liner shipping with its antecedents such as perceived usefulness, perceived ease of use, and the perceptions of security protection. Tests of the structural model confirm Davis’s (1989) notion that perceived ease of use explains the intention of shippers to use Internet services, and that perceived ease of use has a strong positive effect on perceived usefulness. The results also indicate that security protection influences perceived ease of use. The SEM analyses in this study offer OR researchers a methodological guide on how to assess the efficacy of both a measurement model that relates observed indicators to latent factors and a structural model that poses relationships between constructs. Ó 2006 Elsevier B.V. All rights reserved.

*

Corresponding author. Tel.: +886 6 2757575x53243; fax: +886 6 275 3882. E-mail address: [email protected] (C.-S. Lu).

0377-2217/$ - see front matter Ó 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.ejor.2006.05.001

846

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

Keywords: Internet services; Liner shipping; Structural equation model

1. Introduction Structural equation modeling (SEM) has been widely used in a number of disciplines, including banking (Cheng et al., in press), healthcare (Babakus and Mangold, 1992), information management (Etezadi-Amoli and Farhoomand, 1996), logistics (Dunn et al., 1994; Stank et al., 2001), marketing (Steenkamp and Baumgartner, 2000), psychology (Agho et al., 1992; Shen et al., 1995), and tourism management (Reisinger and Turner, 1999). While operations research (OR) researchers have recently begun to recognize this new statistical approach, SEM has become a preferred data analysis method for operations management (OM) empirical research. Studies that employ SEM as the primary analytic tool have frequently appeared in major OM journals (Shah and Goldstein, 2006). For example, Koufteros (1999) and Koufteros et al. (1998) provided an insightful overview of SEM application in the context of time-based manufacturing practices. In their further research, Koufteros et al. (2002) examined a framework of product development practices, and discussed the relations among different variables of interest, as well as their relationships with several important performance variables (product innovation, quality, premium pricing, and profitability) using SEM. Nahm et al. (2003) used SEM to examine the impact of organizational structure on time-based manufacturing and plant performance. It is rather difficult to find an issue of a major operations management journal in which SEM is not used in at least one of the articles (Shah and Goldstein, 2006). Following the trend in OM research, the scope of OR should be broadened in that before OR researchers develop useful decision models, they need to construct relevant theories to underpin their models, and SEM has much to offer in the area of theory building. There is a tradition in OR to incorporate managerially estimated parameter values into operations models. In principle, it is possible to use judgmentally derived parameter values in SEM by fixing specific structural or measurement parameters to certain values, as determined by managerial judgment. More complex managerial assessments to evaluate the relative effects of several variables can

be considered, too. A useful feature of SEM is that it provides, for each judgmentally determined parameter value, a significance test of its appropriateness, and an estimate of the predicted change in the parameter if it were freely estimated. This allows managers to assess to what extent their assessment fits with the data. In the spirit of the decision calculus approach, this information can lead to updated judgmental estimates, re-estimation of the model, and new assessment of the appropriateness of judgmental estimates (Steenkamp and Baumgartner, 2000). Therefore, SEM has potential for decision support and operations modeling. It is useful for OR researchers to test and develop operations models. This study applies SEM to investigate the impact of three sets of antecedent factors, namely security protection, perceived usefulness, and perceived ease of use, on the intention of shippers to use Internet services in liner shipping. SEM is a statistical modeling technique that can handle a large number of endogenous and exogenous variables, as well as latent (unobserved) variables specified as linear combinations (weighted averages) of the observed variables (Golob, 2003). SEM encompasses many different terms (Rigdon, 1998), such as causal models (Hullnad et al., 1996), covariance structure analysis, latent variable analysis (Dunn et al., 1994), confirmatory factor analysis, path analysis, and LISREL analysis (the name of one of the more popular software packages for SEM). From a liner shipping perspective, despite a growth in the use of Internet services, a survey conducted in 2001 on the status of 150 information and communications technology companies in the field of transport indicated that about one-third of these companies had gone bankrupt, 18% were inactive, 17% had been acquired by other companies, and 16% were revising their business models. Only 16% were conducting business as originally planned (United Nations Conference on Trade and Development, UNCTAD, 2002). What contributes to business failure is not always clear. Relatively little previous research has employed systematic analysis to investigate the reasons of failure. One plausible reason that firms fail to pay attention to what their users want and to make an effort to understand the factors that affect users’ intention to use Internet

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

services (Olson and Boyer, 2003). To avoid this pitfall, it is important for liner shipping companies to understand the factors that affect the use of nternet services from the perspective of the shippers. Beyond Fishbein and Ajzen (1975), Davis (1989) and Ajzen (1985, 1991), who have introduced important notions such as ‘‘perceived ease of use’’ and ‘‘perceived usefulness’’ to explain users’ behavior of technology acceptance, another important factor that is recognized as key to the use of Internet services is ‘‘security protection’’. A security threat has been defined as a ‘‘circumstance, condition, or event with the potential to cause economic hardship to data or network resources in the form of destruction, disclosure, modification of data, denial of service, and/or fraud, waste, and abuse’’ (Belanger et al., 2002; Suh and Han, 2003). Security, then, is the protection against these threats. Under this definition, threats can be made either through network and data transaction attacks, or through unauthorized access by means of false or deceptive authentication. The perceptions of security threats originate from a customer’s standpoint. Security protection in business-to-business (B2B) Internet services is reflected in the technologies used to protect customer data and secure legal protection of transactions. This study demonstrates the use of SEM and empirically tests a model based on the theory of reasoned action (TRA) (Fishbein and Ajzen, 1975), the technology acceptance model (TAM) (Davis, 1989; Davis et al., 1989), and the theory of planned behavior (TPB) (Ajzen, 1985, 1991) in the context of shippers’ use intention of Internet services in Taiwan’s liner shipping industry. Typically, the Taiwanese liner shipping industry includes liner shipping companies and shipping agencies. Liner shipping companies are businesses provided by a container transport company whereby cargo-carrying ships are operated between scheduled, advertised ports of loading and discharging on a regular basis. Liner shipping agencies are businesses that represent shipowners to look after marketing and the interests of a container ship while it is in port. Among the major liner shipping companies, Evergreen Marine Corporate was ranked the fourth largest container carrier in the world in 2004, while Yang Ming Lines and Wan Hai Line were ranked 18th and 22nd, respectively. In an increasingly globalized liner shipping market, Internet services are valuable for global carriers to enhance their cost and service advantages in

847

the competitive marketplace. This study provides an empirically validated model to explain the adoption of Internet services for liner shipping services from the shippers’ perspective. We empirically test the model using survey data collected from the top 500 manufacturers in Taiwan, which are the major customers of liner shipping companies in this export-oriented maritime economy. In the next section we review the literature and provide justifications for investigating the intention of shippers to use Internet services in liner shipping, and discuss the conceptual framework and the related research hypotheses. Section 3 discusses the construct measures and sampling techniques adopted in this study. An assessment of traditional methods for assessing measurement scales is discussed in Section 4, while the findings from SEM are examined in Section 5. Finally, the conclusion, discussions and future research directions are provided in Section 6. 2. Theory and research hypotheses The research framework of this study, based on behavioral intention to use Internet services in liner shipping, is shown in Fig. 1. The theory and research hypotheses of this study are presented in four parts. First, a fundamental intention-based theory, the Theory of Reasoned Action (TRA) (Fishbein and Ajzen, 1975) is briefly reviewed. According to TRA, beliefs influence attitudes, which in turn shape intentions, then guide or dictate behaviors. The technology acceptance model (TAM) and the theory of

Security protection

Use intention

Perceived ease of use

Perceived usefulness

Fig. 1. A research framework of shippers’ use intention for Internet services in liner shipping.

848

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

planned behavior (TPB) are derived from TRA and both have gained substantial empirical support through studies on a wide array of users and technologies, which are described in the second and third parts, respectively. The research hypotheses based on the theories of TRA, TAM and TPB are presented in the final part. 2.1. The theory of reasoned action (TRA) TRA is a widely studied model from social psychology that is concerned with the determinants of consciously intended behaviors (Fishbein and Ajzen, 1975; Ajzen and Fishbein, 1980) (see Fig. 2). According to TRA, a person’s performance of a specified behavior is determined by his or her behavioral intention to perform the behavior, and behavioral intention is jointly determined by the person’s attitudes and subjective norms. Attitude towards behavior is defined as an individual’s positive or negative feeling about performing the target behavior (Fishbein and Ajzen, 1975, p. 216) while subjective norm refers to a person’s perception that most people who are important to him or her think he or she should or should not perform the behavior in question (Fishbein and Ajzen, 1975, p. 302). In addition, according to TRA, a person’s attitude towards a behavior is determined by his or her salient beliefs and evaluations. An individual’s subjective norm is determined by a multiplicative function of his or her normative beliefs and motivation to comply with perceived expectations. TRA, being a general model, does not specify the beliefs that are operative for a particular behavior (Davis et al., 1989).

Beliefs and evaluations

2.2. The technology acceptance model (TAM) TAM was proposed by Davis (1986) and Davis et al. (1989) as an extension of TRA. TAM replaces TRA’s attitudinal determinants with a set of two variables, i.e., perceived ease of use and perceived usefulness, which are derived separately for each behavior and employed in many computer technology acceptance contexts (see Fig. 3). Many researchers have conducted studies to examine the relationships among perceived ease of use, perceived usefulness, attitudes, and the use of Internet or other information technologies (Olson and Boyer, 2003; Eastin, 2002; Venkatesh and Davis, 2000; Davis et al., 1989). Agarwal and Prasad (1999) examined the effects of attitudes, of beliefs about perceived usefulness, and of perceived ease of use on a user’s behavioral intentions towards the acceptance of new technologies. Their results provide support for the TAM model, including the mediating role of beliefs. Lederer et al. (2000), based also on TAM, examined users’ behaviors towards the World Wide Web (WWW). Their research found that information quality in websites predicts perceived usefulness for websites users. Mehrtens et al. (2001) created a model to examine the factors that influence the adoption of the Internet in smalland medium-sized enterprises. They concluded that three factors significantly affect the adoption of the Internet by small firms, namely perceived benefits, organizational readiness, and external pressure. Igbaria et al. (1997) investigated the factors that affect the acceptance of personal computing using SEM. They found that perceived usefulness and perceived ease of use are important predictors of the

Attitude toward behavior Behavioral intention

Normative beliefs and motivation to comply

Subjective norm

Source: From Fishbein and Ajzen (1975) Fig. 2. Theory of reasoned action.

Actual behavior

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

849

Perceived usefulness External variables

Attitude toward using

Behavioral intention to use

Actual system use

Perceived ease of use

Source: From Davis (1986) Fig. 3. Theory of technology acceptance model.

use of a computer system. TAM and TRA have been found useful in predicting intentions and usage of information technology in previous studies (Igbaria et al., 1997; Agarwal and Prasad, 1999; Chau and Hu, 2001). However, TAM has been found to be simpler and easier to use, as well as a more powerful model to predict user acceptance of computer technology (Davis et al., 1989; Igbaria et al., 1997). Yet, there is a void of studies on the applicability of TAM in maritime firms. The theoretical insights of TAM thus provide a strong basis on which to examine the factors that contribute to the adoption of Internet services in liner shipping.

2.3. The theory of planned behavior (TPB) TPB is an extension of TRA, which provides a useful tool to predict a wide range of behaviors in many different studies in the information systems literature (Ajzen, 1991; Taylor and Todd, 1995a,b; Harrison et al., 1997; Chau and Hu, 2001) (see Fig. 4). TRA and TPB have provided the basis for several studies on Internet purchasing behavior (Song and Zahedi, 2001; George, 2002, 2004; Khalifa and Limayem, 2003; Suh and Han, 2003). For TRA and TPB, attitude towards the target behaviors and subjective norms about engaging in the

Attitude toward the behavior

Subjective norm

Intention

Perceived behavior control

Source: From Azjen (1991) Fig. 4. Theory of planned behavior.

Behavior

850

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

behaviors are thought to influence intention, but TPB includes perceived behavioral control over engaging in the behaviors as an additional factor influencing intention (George, 2002). According to TPB, an individual’s behavior can be explained by his or her behavioral intention, which is jointly influenced by attitude, subjective norms, and perceived behavioral control. Perceived behavioral control has a direct effect on behavioral intention, too. Attitude refers to an individual’s positive or negative disposition towards performing a particular behavior. Perceived behavioral control is a construct unique to TPB, and it refers to an individual’s perceptions of the presence or absence of the requisite resources or opportunities necessary for performing a behavior (Ajzen and Madden, 1986; Chau and Hu, 2001). Subjective norms reflect the individual’s perception of social support for or opposition to his or her performance of the behavior (Ajzen and Fishbein, 1980). Subjective norms include two separate components, namely normative beliefs and motivation to comply with others (Wiethoff, 2004). Normative beliefs refer to an individual’s perception of other relevant persons’ opinions on whether or not he or she should perform a particular behavior, whereas motivation to comply with others represents the relative importance of the referent person to the actor. TPB provides a useful framework to examine an individual’s behavior to use Internet services. Complementary to TPB, this research applies SEM to examine the intention of shippers to use Internet services in liner shipping. The primary customers of liner shipping companies are shippers, which include manufacturers and trading companies. The focus of this study is on B2B Internet services. There are relevant B2B studies on Internet services. For instance, Cheng et al. (2002) investigated the use of the Internet in Hong Kong in manufacturing and service industries, Olson and Boyer (2003) discussed the factors influencing the utilization of Internet purchasing in small organizations, and Grandon and Pearson (2004) examined the determinant factors of the adoption of electronic commerce in small and medium sized US businesses. One common point in these previous B2B studies is that subjective norm was dropped. Following this line of research, and based on the TAM, we have excluded the variables in TPB related to subjective norms and perceived behavioral control from our study. Many researchers have examined the factors influencing the acceptance and use of information

technology or Internet services in a specific industry. However, an important factor that should be taken into account in the adoption and provision of Internet services – security protection – seems to have been ignored in the previous studies. Security protection of Internet services is related to the technologies and legal means used to protect and secure consumer data and benefits (Belanger et al., 2002), and to safeguard users against such threats as denial of service, fraud and disclosure of commercial secrets (Kalakota and Whinston, 1996). Ajzen (1991) defined perceived behavioral control as people’s perception of the ease or difficulty of performing the behavior. Behaviors are more likely to result from intention when people believe they have the resources to perform the behavior and are likely to be successful in doing so. Perceived behavioral control comprises control beliefs, or the belief that the required resources and opportunities are available to carry out the behavior, and perceived facilitation, or the assessment of the importance of those resources to successfully complete the behavior (Ajzen, 1991; Wiethoff, 2004). In our study a shipper’s intention to use carriers’ Internet services may be affected by perceived control factors such as security protection. While many corporations have achieved success in using the Internet to exchange data and information, there is a lack of empirical studies investigating factors that affect the use of Internet services in the context of liner shipping. This is also the issue that will be investigated in the model. 2.4. Research hypotheses Based on the research framework (see Fig. 1), the network of relationships among the constructs in the model and the rationale for the proposed linkages are elaborated below. Security protection is an important concern of customers who are considering the use of Internet services, and many studies have focused on this issue (e.g., Elofson and Robinson, 1998). The risk in carrying out commercial activities over the Internet lies in the transmission of data via the Internet and WWW. When a customer uses the Internet, anyone from anywhere around the world may be able to access the information being transmitted. The possibility that information may be stolen, data corrupted, and fraud committed may become a reality (Suh and Han, 2003). Strader and Shaw (1997) described how customers may refrain from using

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

the Internet if they feel that the level of risk is not acceptable. Customers may incur damages such as invasion of privacy and financial loss if security problems arise. Firms will suffer severe losses ranging from the loss of valuable information to a bad public image and even to legal penalties imposed by regulatory agencies (Suh and Han, 2003). The security concerns of shippers may require carriers to provide technological security protection, such as encryption, authentication, and legal recognizance. In this study the application of specific security protection technologies in transactions and legal recognizance are categorized as security protection. Our delineation of security protection is similar in nature and scope to that identified and discussed in the studies of Kalakota and Whinston (1996), Hoffman et al. (1999), Belanger et al. (2002), and Suh and Han (2003). Miyazaki and Fernandez’s (2001) study has provided support for a relationship between concerns over security and a consumer’s willingness to purchase online. Belanger et al. (2002) and Suh and Han (2003) also studied the impact of security on the acceptance of electronic commerce and website use. However, the effects of security protection on perceived ease of use and perceived usefulness have not been tested by previous studies. Shippers would tend to perceive transactions as less risky if liner shipping companies provide security protection in their Internet services. It is therefore reasonable to expect that security protection may have a positive effect on shippers’ perceptions of ease of use and usefulness when transacting with an Internet service provider. The hypotheses of this research seek to extend the research by examining the influence of perceived security protection on shippers’ perceptions of the usefulness and ease of use of Internet services in liner shipping. Thus, we hypothesize that: H1a: Perceived security protection has a positive effect on a shipper’s intention to use Internet services in liner shipping. H1b: Perceived security protection has a positive effect on a shipper’s perception of the usefulness of Internet services in liner shipping. H1c: Perceived security protection has a positive effect on a shipper’s perception of the ease of using Internet services in liner shipping. Perceived ease of use refers to ‘‘the extent to which a person believes that using the system will be free of effect’’ (Venkatesh and Davis, 2000, p.

851

187). Perceived ease of use was identified as an important determinant of the use of a system in Davis et al. (1989) study. According to TAM, perceived usefulness is also influenced by perceived ease of use because the easier the system is to use the more useful it can be (Venkatesh and Davis, 2000). Perceived ease of use was suggested as a causal antecedent to perceived usefulness in Venkatech and Davis’s study. Igbaria et al. (1997) found that both perceived usefulness and perceived ease of use are important factors in the acceptance of personal computing. Several studies using TAM have suggested perceived ease of use influence perceived usefulness (Igbaria et al., 1997; Heijden, 2003; Hu et al., 2003). In our study we also investigate this relationship. We hypothesize that shippers who perceive that systems having a high speed to link to the shipping companies’ services and that they are easy to use will also perceive the services to be more useful, as suggested by previous research findings (Lucas and Spitler, 1999; Lederer et al., 2000). Perceived ease of use is a shipper’s assessment that his or her interaction with the Internet services offered by a carrier will be relative free of cognitive burden. In other words, shippers do not need to spend significant time and effort to learn and use the services. Perceived ease of use represents an intrinsically motivating component of the shippers and carriers’ Internet services interaction. Accordingly, we surmise that: H2a: Perceived ease of use has a positive effect on the perceived usefulness of Internet services in liner shipping. H2b: Perceived ease of use has a positive effect on a shipper’s intention to use Internet services in liner shipping. Venkatesh and Davis (2000) defined perceived usefulness as ‘‘the extent to which a person believes that using the system will enhance his or her job performance’’ (p. 187). Igbaria et al. (1997) demonstrated the importance of perceived usefulness, and argued that it has a direct effect on the acceptance of personal computing due to the reinforcement value of outcomes. Hu et al. (2003) and Heijden (2003) suggested, too, that there are relationships between perceived usefulness and intention to use websites and information technology. Other studies have as well demonstrated that perceived usefulness is positively associated with performance, usage, or use intention (e.g., Lucas and Spitler, 1999; Lederer et al., 2000; Olson and Boyer, 2003). In our work,

852

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

perceived usefulness is defined as the extent to which shippers believe that using the Internet services in liner shipping would enhance their operational performance or benefits. These benefits include reductions in documentation and communication cost and errors. It is expected that shippers believe that if they use the shipping companies’ Internet services, their operational performance will improve. Therefore, we posit that: H3: Perceived usefulness has a positive effect on a shipper’s intention to use Internet services in liner shipping.

3. Construct measures and sample technique The data for this study were collected via a questionnaire survey. The design of the questionnaire follows the stages outlined by Churchill (1991). Content validity was ensured through a comprehensive review of the literature and interviews with practitioners, i.e., the indicators in the questionnaire were based on previous studies (Davis, 1986; Igbaria et al., 1997; Agarwal and Prasad, 1999; Chau and Hu, 2001; Kalakota and Whinston, 1996; Belanger et al., 2002; and Suh and Han, 2003) interviews with practitioners, and discussions with a number of executives and experts in liner shipping. The items in the questionnaire were judged as relevant by 15 liner shipping executives. The interviews resulted in minor modifications to the wording and examples provided in some measurement items, which were finally accepted as possessing content validity. The refined measurement items were included in the final survey questionnaire administered to the target respondents. 3.1. Construct measures Numerous previous studies have identified an array of indicators for the acceptance of the Internet, technology or personal computing (Davis, 1986; Igbaria et al., 1997; Lucas and Spitler, 1999; Lederer et al., 2000; Olson and Boyer, 2003). Based on previous studies, the measurement factors and indicators for the acceptance of the Internet in liner shipping are identified and shown in Appendix A. The respondents were asked to indicate the level of importance of the factors affecting the use of Internet services in liner shipping, such as perceived usefulness, perceived ease of use, and security pro-

tection. The measures of perceived ease of use from the perspective of shippers are based on the functions available from the websites of the service providers, i.e., the liner shipping companies. These factors were scored using a seven-point Likert scale, where 1 corresponds to ‘‘very unimportant’’ and 7 to ‘‘very important.’’ It should be noted that the measures of use intention of Internet services were selected from the Internet services attributes listed in the websites of major liner shipping companies. Based on Jeffery’s (1999) study, an evaluation of the attributes of Internet services provided by the main liner shipping companies was undertaken by searching the Internet. These companies included Maersk SeaLand, P&O Nedlloyed, Evergreen, Hanjin/DSRSenator, MSC, NOL/APL, K-Line, MOL, Yang Ming, COSCO, and OOCL. An examination of the Internet service attributes listed in the websites of the major liner shipping companies reviewed in this study reveals that many are associated with enquiries on sailing schedules, tracking cargo, booking space, and responding to customers. This indicates that the key attributes of Internet services considered extremely important by liner shipping companies are interactive in nature. Therefore, the measures of use intention of Internet services were created on the basis of these key service attributes. Similarly, the respondents were asked to indicate their use intention of Internet services in liner shipping on a seven-point Likert scale, where 1 corresponds to ‘‘I would strongly dislike to use’’ and 7 to ‘‘I would strongly like to use.’’ 3.2. Sampling techniques The sample of shippers was selected from the List of Leading Firms with Good Export & Import Performance, published by the Board of Foreign Trade of the Ministry of Economic Affairs in Taiwan. We used the key informant approach to collect data (Phillips and Bagozzi, 1986). The survey questionnaire in Chinese language was sent to the shipping division of the top 500 export firms (shippers). By definition, an informant’s role is to report on organizational processes, events, or outcomes that are aggregate in nature; thus, informants should be sampled according to their knowledge of or involvement with the research issues under investigation. Accordingly, the informants sampled were familiar with the requirements of their company for Internet services in liner shipping. The potential effective

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

population size was reduced to 487, as 13 managers had left the companies or the businesses were no longer in operations. The initial mailing elicited 46 usable responses. A follow-up mailing was sent two weeks after the initial mailing. An additional 39 usable responses were returned. The total number of usable responses was 85. Therefore, the overall response rate for this study was 17.45%. An analysis of a non-response bias was conducted to assess the extent of the potential bias in the data collected. The procedure requires that responses be numbered sequentially in the order of the date they were received. We administered the survey twice and set a cut-off date to classify the responses into one of the two mailings. A comparison of the early (those who responded to the first mailing) and the late (those who responded to the second mailing) respondents was carried out to test for the nonresponse bias (Armstrong and Overton, 1977). The 85 respondents were divided into two groups based on their response wave (first and second waves). ttests were performed on the responses of the two groups. At the 5% significance level, there were no significant differences between the responses of the two groups. Although the results do not rule out

853

the possibility of a non-response bias, they suggest that non-response may not be a problem to the extent that the late respondents represent the opinions of non-respondents. 3.3. Characteristics of respondents The sample used for analysis was received from 85 firms, which rank in the top 500 export firms in Taiwan. The profiles of the respondent companies and their characteristics are displayed in Table 1. The results show that 32.94% of the participants in the survey were clerks, 25.8% directors/vice directors, 11.76% and 10.59% managers/assistant managers and sales representatives, respectively, and 16.47% fell into other categories. Generally, the shipping division is in charge of a company’s logistics activities in Taiwan. Therefore, the views of managers or employees in shipping divisions on Internet service requirements would be more useful than those of personnel in other divisions. However, in many firms the shipping division is low in the organizational hierarchy, while shipping or logistics activities are handled by staff in the business division. This explains why no staff of the rank of vice

Table 1 Profile of the respondent companies (n = 85) Characteristics of the respondents

Frequency

%

Job title

Vice president or above Manager/assistant manager Director/Vice director Sales representative Clerk Others Not available

0 10 22 9 28 14 2

0.00 11.76 25.88 10.59 32.94 16.47 2.35

Nature of business

Trading company Manufacturer

10 75

11.76 88.24

Main export cargo

Electrical machineries and apparatus Metal products Textile products Rubber and plastic products Chemical products Machineries Paper products Processed food Wooden products Others

22 10 9 7 4 4 3 2 1 23

25.88 11.76 10.59 8.24 4.71 4.71 3.53 2.35 1.18 27.06

Level of turnover (US $)

Below 10 million 11–20 million 21–30 million Greater than 30 million Unknown

7 6 6 60 6

9.24 7.06 7.06 70.59 7.06

854

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

president or higher and few managers/assistant managers participated in the survey. The vast majority of the shippers were manufacturers. The remaining respondents were trading companies (11.76%). Table 1 shows the main items of the cargo of the responding shipping firms. Electrical machinery and apparatus were indicated as the major cargo items by 25.88% of the responding firms. In addition, almost 12% and 11% of the respondents stated that metal products and textile products, respectively, were their primary items of cargo. For less than 10% of the respondents, the main cargo items were rubber and plastic products, chemical products, machinery, paper products, processed food and wooden products. The respondents were also asked to provide information about the turnover of their firms. The results in Table 1 indicate that over 70% of the respondents reported that their turnover was over US$30 million. Only 9.2% reported that their turnover was below US$10 million, while 14.1% had a turnover of between US$11 million and US$30 million. 4. Research methods Based on the studies of Koufteros (1999) and Koufteros et al. (2001), our research steps and methods included instrument development, an exploratory analysis, a confirmatory factor analysis, and a test of a structural model as shown in Fig. 5. Instruments in the form of measurement items in our survey questionnaire were developed to capture the dimensions of Internet services. Item generation began with theory development and a literature review. Items were evaluated through interviews with practitioners. Furthermore, the methods employed for the development and exploratory evaluation of the measurement scales for the latent variables in this study included corrected item-total correlations (CITC), exploratory factor analysis within block, exploratory factor analysis on the entire set, and reliability estimation using Cronbach’s alpha. CITC were used for purification purposes because unimportant items may confound the interpretation of the factor analysis. Within block factor analysis may confirm that one factor can be identified in a given block of items and in essence addresses the unidimensionality. Exploratory factor analysis (EFA) was used to determine how many latent variables

Fig. 5. Analytical steps.

underlie the complete set of items. Cronbach’s alpha is one of the most widely used metrics for reliability evaluation (Koufteros et al., 2001). These techniques are useful in the early stages of empirical analysis, where theoretical models do not exist and the basic purpose is exploration. However, these traditional techniques do not assess unidimensionality (Segar, 1997; O’Leary-Kelly and Vokurka, 1998), nor can unidimensionality be demonstrated by either mathematical or practical examinations (Gerbing and Anderson, 1988; Koufteros, 1999). Several researchers have suggested the use of confirmatory factor analysis (CFA) with a multiple-indi-

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

cator measurement model to assess unidimensionality (Anderson, 1987; Segar, 1997). Exploratory techniques can help us develop hypothesized measurement models that can subsequently be tested using confirmatory factor analysis (Koufteros, 1999). Confirmatory factor analysis is performed on the entire set of items simultaneously. Anderson et al. (1987) suggested that assessment of unidimensionality for sets of measurement items be made in the same model as the one that the researcher is interested in making statements about the unidimensionality of those measurement items. Convergent validity was assessed by examining the significance of individual item loadings through ttests. The overall fit of a hypothesized model can be tested by using the maximum likelihood Chisquare statistic provided in the Amos (a software package for SEM) output and other fit indices such as the ratio of Chi-square to degrees of freedom, goodness-of-fit index (GFI), adjusted goodness-offit index (AGFI), comparative fit index (CFI), root mean square residual (RMSR), the root mean square error of approximation (RMSEA), standardized residual, and modification index (MI). Each item’s completely standardized expected change in Kx was examined with regard to potential misspecifications in the measurement model. Items exhibiting change in Kx greater than 0.4 should be investigated for their lack of unidimensionality and possible misspecifications in the model (Koufteros et al., 2001). Discriminant validity was assessed by comparing the average variance extracted (AVE) to the squared correlation between constructs. The AVE estimate is a complimentary measure to the measure of composite reliability (Fornell and Larcker, 1981a; Koufteros et al., 2001). We tested our model on the adoption of Internet services in liner shipping by shippers with survey data collected from them. To assess the fit of the model to the data, Chi-square per degrees of freedom, GFI, AGFI, CFI, RMSR, RMSEA, and MI were computed. If the model fits the data adequately, the t-values of the structural coefficients will be evaluated to test the research hypotheses. 5. Exploratory measurement results 5.1. Corrected item-total correlations Item-total correlation refers to a correlation of an item or indicator with the composite score of all the items forming the same set. Corrected

855

item-total correlation (CITC) does not include the score of the particular item in question in calculating the composite score, thus it is labeled ‘corrected’ (Koufteros, 1999). The CITC analyses were performed for each construct. Table 2 shows the CITC scores, with the exception of accurate information in Internet services (PE3), which ranged from 0.524 to 0.822. Using the traditional cutoff value 0.50 for evaluating CITC, PE3 was eliminated from subsequent analyses. 5.2. Exploratory factor analysis An EFA analysis was used to reduce the twelve items to a smaller, more manageable set of underlying factors. This is helpful for detecting the presence of meaningful patterns among the original variables and for extracting the main service factors. Two common methods of factor extraction are principal component analysis (PCA) and principal axis factoring (PAF). PAF is a form of factor analysis that seeks the least number of factors that can account for the common variance (correlation) of a set of variables, whereas the more common PCA in its full form seeks the set of factors that can account for all the common and unique (specific plus error) variance in a set of variables. PCA is generally used when the research purpose is data reduction (to reduce the information in many measured variables to a smaller set of components). PAF is generally used when the research purpose is to identify latent variables that contribute to the common variance of the set of measured variables, excluding variablespecific (unique) variance (Kline, 1998; Fabrigar et al., 1999; Hutcheson and Sofroniou, 1999). PAF is preferred for purposes of structural equation modeling (SEM). PAF accounts for the covariation among variables, whereas PCA accounts for the total variance of variables. Because of this difference, in theory it is possible under PAF but not under PCA to add variables to a model without affecting the factor loadings of the original variables in the model. Widaman (1993) suggested that PCA not be used if a researcher wishes to obtain parameters reflecting latent constructs or factors. Thus, this study was based on PAF to extract the factors. An important tool for interpreting factors is the rotation of factors. Two methods can be used to identify the factors, namely the orthogonal rotation method and the oblique rotation method (Churchill, 1991). The choice of an orthogonal or oblique

856

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

Table 2 Corrected item-total correlations Variable

Item

Security protection

SE1.

SE2. SE3. SE4. SE5. Perceived usefulness

PE1. PE2. PE3. PE4.

Perceived ease of use

PU1. PU2. PU3. PU4.

Use intention

UI1. UI2. UI3. UI4.

Corrected item-total correlation The ability to provide technical protection for on-line transactions such as quoting, booking space and issuing B/Ls for shippers The legal recognizance of electronic B/L The ability to provide protection of commercial secrets (e.g., from thefts of data and disclosures) The ability to provide legal protection of transactions (e.g., from denial of service and fraud) The ability to provide protection of data transmission

0.658

0.627 0.776 0.822 0.705

The speed to link shipping companies’ Internet services The speed of responses for my enquires in shipping companies’ Internet services The accurate information in shipping companies’ Internet services Ease to use shipping companies’ Internet services

0.619 0.605

Using shipping companies’ Internet services to reduce documentation cost Using shipping companies’ Internet services to reduce communication cost (e.g., telephone and fax) Using shipping companies’ Internet services to improve documentation errors Using shipping companies’ Internet services to make it faster to receive bill(s) of lading

0.727

0.327 0.524

0.663 0.616 0.660

Using the tracking of containers function in shipping companies’ Internet services Enquiring sail schedules in shipping companies’ Internet services Checking the condition of customs clearance of my firm’s cargo in shipping companies’ Internet services Booking space in shipping companies’ Internet services

rotation should be made on the basis of the particular needs of a given research problem. Hair et al. (1995) suggested that if the goal of the research is to reduce the number of original variables, regardless of how meaningful the resulting factors may be, the appropriate solution would be an orthogonal one. However, if the ultimate goal of the factor analysis is to obtain several theoretically meaningful factors or constructs, an oblique solution is appropriate. An orthogonal rotation provides information that no correlations exist between the factors or components, whereas an oblique rotation assumes that the factors are actually correlated with one another. Moreover, in order to ascertain the dimensionality within block before subjecting all items of all constructs to a single factor, a within-block factor analysis was conducted. Table 3 shows that there is a single factor in each block with relatively high loadings. The loadings of the security protection block ranged from 0.716 to 0.924, while the loadings

0.746 0.756 0.582 0.701

Table 3 Within-block EFA (within-block loadings) Block 1

Block 2

Block 3

Block 4

Item

Factor

Item

Factor

Item

Factor

Item

Factor

SE1 SE2 SE3 SE4

0.716 0.804 0.906 0.924

PU1 PU2 PU3 PU4

0.852 0.808 0.791 0.822

PE1 PE2 PE4

0.852 0.859 0.802

UI1 UI2 UI3 UI4

0.883 0.880 0.742 0.835

of the use intention block ranged from 0.742 and 0.883. The loadings of the items under perceived usefulness and perceived ease of use blocks were 0.791 and 0.852, as well as 0.802 and 0.859, respectively. Similar to the results in many previous studies on Internet related topics, we conclude that we have sufficient evidence of the unidimensionality. However, Koufteros (1999) indicated that while the results from within-block factor analysis may tell us whether more than one factor may be present in a given block of items, they do not tell us how the

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

items of one block relate to the factor(s) or items of the other blocks. Moreover, this study is based on TAM to identify the factors that influence the intention of shippers to use Internet services, so the oblique rotation method with Oblimin was employed and the results are shown in Table 4. A factor loading can be used as an indicator in interpreting the role each item plays in defining each factor. Factor loadings are in essence the correlation of each item to their underlying factor. Kim and Muller (1978) suggested factor loadings of 0.30 as a cut-off for significance. Nunnally (1978) suggested that it is doubtful that loadings of a smaller size be taken seriously when accounting for less than ten percent of the variance of the factor. According to Hair et al. (1998), in a sample of 85 respondents, factor loadings of value greater than 0.60 are required to retain an item. Alternatively, Lewis-Beck (1994) stated eigenvalue is a measure attached to factors and indicates the amount of variance in the pool of original variables that the factor explains. Hence, the number of factors extracted in this research was based on the cumulative percent of variation explained (Churchill, 1991). This rule suggests that only factors that explain more variance than the average amount explained by one of the original items should be retained. The resulting exploratory solution with an oblique rotation indicates a three-factor solution (see Table 5). Table 5 indicates that the factor loadings of SE5, PE4, PU3, and PU4 were below 0.60 and were subsequently eliminated. The loadings for the first factor (perceived ease of use), after the elimination of the items as stated above, ranged from 0.813

Table 4 Exploratory factor analysis and the factor loadings Item

Factor 1

Factor 2

Factor 3

SE1 SE2 SE3 SE4 SE5 PE1 PE2 PE4 PU1 PU2 PU3 PU4

0.403 0.162 0.067 0.012 0.493 0.825 0.813 0.385 0.134 0.045 0.572 0.429

0.633 0.836 0.869 0.886 0.583 0.009 0.041 0.200 0.020 0.174 0.011 0.029

0.262 0.119 0.168 0.059 0.169 0.008 0.114 0.360 0.786 0.850 0.382 0.534

Eigenvalues Percentage of variance

5.76 47.99

1.62 13.51

1.07 8.91

857

Table 5 Cronbach alpha values for each factor Measures

Cronbach alpha

Factor 1: Security protection (SE1, SE2, SE3, SE4) Factor 2: Perceived ease of use (PE1, PE2) Factor 3: Perceived usefulness (PU1, PU2) Use intention (UI1, UI2,UI3, UI4)

0.877 0.758 0.818 0.862

to 0.825, the second factor (security protection) ranged from 0.633 to 0.886, and the third factor (perceived usefulness) ranged from 0.786 to 0.850. In this study, only a few items were considered in the research design. It is critical to observe whether the cross-loadings are high enough to be alarming. Overall, items loaded strongly on their intended factors as the lowest factor loading stood at 0.633. Because the item SE1 is an important measure (mean = 6.59 in Table 2) for the security protection construct, its loading was higher than factor 1 (perceived ease of use, factor loading = 0.403) and factor 3 (perceived usefulness, factor loading = 0.262). The percentages of variance explained of the three factors were 47.99, 13.51, and 8.91. These three factors accounted for 70.41% of the variance. 5.3. Coefficient alpha and reliability Cronbach’s alpha is one of the most widely used measures for evaluating reliability (Koufteros, 1999). The Cronbach’s alpha value for each measure is shown in Table 5. The reliability value for each construct was well above the value of 0.75, which is considered satisfactory for basic research (Nunnally, 1978; Churchill, 1991; Litwin, 1995). Nevertheless, Cronbach’s alpha has several disadvantages, including the fact that it is inflated when a scale has a large number of items, and it assumes that all of the measured items have equal reliabilities (Gerbing and Anderson, 1988). In addition, Cronbach’s alpha cannot be used to infer unidimensionality (Gerbing and Anderson, 1988). 6. Structural equation modeling 6.1. Confirmatory factor analysis The importance of unidimensionality has been highlighted by Gerbing and Anderson (1988), who stated that ‘‘because the meaning of a measure

858

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

intended by the researcher may not be the same as the meaning inputted to it by the respondents, the scale development process must include an assessment of whether the multiple measures that define a scale can be acceptably regarded as alternative indicators of the same constructs’’ (p. 186). Traditionally, item-total correlations and exploratory factor analysis can only offer preliminary analyses, particularly without an adequate theoretical base, because they fail to directly assess unidimensionality (Gerbing and Anderson, 1988). In relation to this respect and following the work of Churchill (1979), Peter (1979, 1981) and Jo¨reskog and So¨rbom (1996) delineated a paradigm for scale development that employs confirmatory factor for the assessment of unidimensionality. The confirmatory factor analysis (CFA) approach to scale estimations and construct reliability has overcome the limitations of the exploratory factor model in which the researcher is allowed to determine on the basis of theories (1) which pairs of common factors are correlated, (2) which observed variables are affected by which common factors, (3) which observed variables are affected by an error term factor, and (4) which pairs of error terms are correlated. Statistical tests can then be employed to determine whether the data confirm the substantively generated model (Long, 1983; Garver and Mentzer, 1999). CFA involves the specification and estimation of one or more hypothesized models of factor structure, each of which proposes a set of latent variables (factors) to account for covariances among a set of observed variables (Koufteros, 1999). The path diagram presented in Fig. 6 implies a measurement model where there are four latent variables (constructs) made up of their corresponding multiple indicators (measures or items). Following the convention of AMOS analysis (Arbuckle, 1997), observed variables are represented by squares and labeled with letters X. Latent variables are represented by circles and labeled with the Greek letters n, which are also called common factors. At the left of the figure, the Greek letters d are seen as errors in manifest or observed variables. A straight arrow pointing from a latent variable to an observed variable indicates the causal effect of the latent variable on the observed variable. The Greek letter Uij represents the correlation between the latent variables, whereas the Greek letter k coefficients are the factor loadings of the observed indicators on the latent variables. Curved arrows between two latent variables indicate that those variables are correlated.

δ1 X1

δ2 δ3 δ4

1

X2

X3

X4

δ5 X5

δ6

ξ1 Security protection

1

ξ2 Perceived ease of use

X6

Φ ij δ7 X7

δ8

δ9

1

ξ3 Perceived usefulness

X8

X9 1

δ 10 δ 11

X10

ξ4 Use intention

X11

δ 12 X12

Fig. 6. Path diagram representing the measurement model.

There is an important note on the estimation of the measurement model for constructs with more than one item. Because of the estimation procedure, the construct must be made ‘‘scale invariant’’, meaning that the indicators of a construct must be ‘‘standardized in a way to make constructs comparable’’ (Long, 1983; Jo¨reskog and So¨rbom, 1996; Koufteros, 1999). One of the loadings in each construct can be set to a fixed value of 1.0 in order to make the constructs comparable (Koufteros, 1999). 6.2. Convergent validity and item reliability Convergent validity can be assessed by examining the loadings and their statistical significance through t-values (Dunn et al., 1994). In the AMOS text output file, the t-value is the critical ratio (C.R.), which represents the parameter estimate divided by its standard error. A t-value greater than 1.96 or smaller than 1.96 implies statistical significance (Segar, 1997; Byrne, 2001). The larger the factor loadings or coefficients, as compared with their standard errors, the stronger is the evidence that there is a relationship between the observed indicators to their respective latent factors (Bollen,

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

859

1989; Koufteros, 1999). Table 7 shows that each item exceeds the critical ratio at the 0.05 level of significance. Thus, all indicators were significantly related to their specified constructs, verifying the posited relationships among the indicators and latent variables. Item reliability refers to the R2 value in the observed variables that are accounted for by the latent variables influencing them. The R2 can be used to estimate the reliability of a particular observed variable (item) (Koufteros, 1999). R2 values above 0.50 provide evidence of acceptable reliability (Bollen, 1989). The squared correlations for the twelve items are listed in Table 6. An examination of the results reveals that two items (i.e., SE1 and UI3) did not meet the 0.50 criterion. Due to the fact that UI3 (Customs responses) is an important item for explaining use intention of Internet services, only item SE1 was eliminated in the revised model. Table 7 shows that the results are marginally acceptable as all items exhibited an R2 value greater than 0.338 and the critical ratios were all higher than 1.96, providing evidence of convergent validity.

of the Chi-square (v2 = 44.403, df = 38) was 0.220, and it was statistically non-significant. This provides evidence of model fits as the hypothesized model can represent adequately the observed data. Moreover, the goodness-of-fit index (GFI) and adjusted goodness-of-fit index (AGFI) had values of 0.923 and 0.866, which are acceptable. The root mean square residual (RMSR) indicates that the average residual correlation was 0.036. The root mean square error of approximation (RMSEA) was 0.045. Both estimates provide evidence of model fit as they were below 0.05. The Tucker Lewis Index (TLI) was 0.980, while the CFI was 0.986. Both are incremental fit indices and their values exceeded the recommended level of 0.90, further supporting acceptance of the model. The normed Chi-square (v2/df) had a value of 1.169. This falls well within the recommended range for conditional support to be given for model parsimony. In summary, the various index of overall goodness-of-fit for the model lent sufficient support for the results to be deemed an acceptable representation of the hypothesized constructs.

6.3. Assessment of the fit and unidimensionality of the model

6.4. Standardized residuals and modification indices

AMOS provides absolute goodness-of-fit measures, and these are shown in Table 7. The p-value

The model may be modified by examining the standardized residuals and the modification indices. The standardized residuals (normalized) are

Table 6 Parameter estimates, standard errors, critical ratios, and R2 for the proposed model Latent variable

Item

Unstandardized factor loading

Completely standardized factor loading

Standard errora

Critical ratiob

R2 (item reliability)

n1

SE1 SE2 SE3 SE4

1.0 1.324 1.681 1.603

0.575 0.688 0.915 0.944

–c 0.265 0.285 0.269

– 4.997 5.906 5.955

0.331 0.473 0.836 0.892

n2

PE1 PE2

1.0 0.883

0.810 0.765

– 0.174

– 5.073

0.586 0.655

n3

PU1 PU2

1.0 1.125

0.806 0.860

– 0.219

– 5.136

0.649 0.740

n4

UI1 UI2 UI3 UI4

1.0 0.929 0.737 0.857

0.899 0.866 0.582 0.733

– 0.095 0.129 0.110

– 9.765 5.709 7.819

0.808 0.750 0.338 0.537

Fit indices: v2 = 55.038 (p = 0.226), df = 48, v2/df = 1.147, GFI = 0.913, AGFI = 0.859, CFI = 0.986, RMSR = 0.038, RMSEA = 0.042, TLI = 0.980. a S.E. is an estimate of the standard error of the covariance. b C.R. is the critical ratio obtained by dividing the estimate of the covariance by its standard error. A value exceeding 1.96 represents a level of significance of 0.05. c Indicates a parameter fixed at 1.0 in the original solution.

860

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

Table 7 Parameter estimates, standard errors, critical ratios, and R2 for the revised model Latent variable

Item

Unstandardized factor loading

Completely standardized factor loading

Standard errora

Critical ratiob

R2 (item reliability)

n1

SE2 SE3 SE4

1.0 1.277 1.219

0.685 0.946 0.915

–c 0.169 0.160

– 7.564 7.611

0.469 0.838 0.895

n2

PE1 PE2

1.0 0.886

0.764 0.803

– 0.176

– 5.037

0.584 0.658

n3

PU1 PU2

1.0 1.133

0.811 0.863

– 0.221

– 5.132

0.644 0.746

n4

UI1 UI2 UI3 UI4

1.0 0.929 0.737 0.857

0.899 0.866 0.581 0.733

– 0.095 0.129 0.110

– 9.764 5.708 7.819

0.808 0.750 0.338 0.537

Fit indices: v2 = 44.403 (p = 0.220), df = 38, v2/df = 1.169, GFI = 0.923, AGFI = 0.866, CFI = 0.986, RMSR = 0.036, RMSEA = 0.045, TLI = 0.980. a S.E. is an estimate of the standard error of the covariance. b C.R. is the critical ratio obtained by dividing the estimate of the covariance by its standard error. The value of exceeding 1.96 represents a level of significance of 0.05. c Indicates a parameter fixed at 1.0 in the original solution.

Table 8 Modification indices for the model Items

n1 (security protection)

n2 (perceived ease of use)

n3 (perceived usefulness)

n4 (use intention)

SE1 SE2 SE3 SE4 PE1 PE2 PU1 PU2 UI1 UI2 UI3 UI4

– – – – 0.066 0.057 0.293 0.249 0.005 0.433 0.505 0.322

2.305 0.005 0.202 0.019 – – 0.139 0.118 0.689 0.067 0.206 1.796

0.116 0.124 1.464 1.978 0.070 0.061 – – 0.301 0.387 0.261 1.922

1.922 0.529 0.540 0.016 0.001 0.001 0.050 0.042 – – – –

provided by the AMOS program and represent the differences between the observed correlation/covariance and the estimated correlation/covariance matrix. Small fitted residuals indicate good fit, although their size depends on the units of items of the observed variables. To ease interpretation, residuals are standardized by dividing them by their asymptotic standard errors (Jo¨reskog, 1993). Residuals with values larger than 2.58 in absolute terms are considered statistically significant at the

0.05 level (Hair et al., 1998). Significant residuals indicate the presence of a substantial error for a pair of indicators. The results show that none of the standardized residual values exceeded 2.58 in absolute terms. This provides additional evidence of model fit and of no apparent misspecifications. Another indication of a possible re-specification of the model is the modification index (MI), which is calculated for each non estimated relationship. The MI can be used to decide which parameters should be added to the model. The MI are measures of the predicted decrease in the Chi-square value that results if a single parameter (fixed or constrained) is freed (relaxed) and the model re-estimated, with all other parameters maintaining their present values (Jo¨reskog and So¨rbom, 1996; Reisinger and Turner, 1999). For example, a high modification index for item SE1, 2 of lambda X suggests that X1 may share a significant amount of variance with construct 2, i.e., the indicator is not uni-dimensional. The lambda X is an indicator of exogenous indicator or construct. Therefore, to reduce the overall Chi-square value by the amount of the modification index, i.e., to improve the model fit, a path between this respective indicator and the construct can be estimated. Typically, small modification indices (i.e., approximately 4.0, p < 0.05) provide an

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

insignificant improvement in model fit relative to the loss of one degree of freedom from estimating the additional parameter (Anderson, 1987; Koufteros, 1999). In terms of modification indices for the twelve measures, none of the modification indices was significant. Table 8 shows the highest value of the modification index was only 2.305. This implies that there does not appear to be a reason for re-specification. It is also important to accompany the modification index statistic with the completely standardized expected changes in the loading with other latent variables. Items exhibiting changes in expected parameter change (EPC) of greater than 0.3 should be investigated for lack of unidimensionality (Koufteros, 1999). The results of the completely standardized expected changes in lambda X are summarized in Table 9. The highest completely standardized expected change in lambda X was 0.175 for item UI4 in n2 and this result does not justify an alternative specification. All other changes were below 0.17. The modification indices and the values of the completely standardized expected changes can be obtained from the AMOS software package. 6.5. Discriminant validity Assuming an adequate model fit, further psychometric analysis for composite scales can be performed with the measurement model. The test of discriminant validity is one of the important analyses to be performed (Koufteros, 1999). According to Arbuckle (1997) study, models are constructed for all possible pairs of latent variables within each

861

construct (containing the measurement items). These models are run: (1) with the correlation between the latent variables fixed at 1.0, and (2) with the correlation between the latent variables free to assume any value. The difference in Chi-square values for the fixed (for constrained) and free solutions indicate whether a one-dimensional model would be sufficient to account for the intercorrelation among the construct observed in each pair. A significantly lower Chi-square value for the model in which the trait correlations are not constrained to unity would indicate that the traits are not perfectly correlated and that discriminant validity can be inferred (Anderson, 1987). The results indicate that the differences in v2 between the fixed and free solutions were very significant (i.e., the minimum v2 = 26.460, p < 0.01, df = 1). This result provides evidence of discriminant validity. It is also possible to test discriminant validity by comparing the average variance extracted (AVE) with the squared correlation between constructs. Discriminant validity exists if the items share more common variance with their respective construct than any variance that construct shares with other constructs (Fornell and Larcker, 1981b; Koufteros, 1999). As can be seen in Table 10, the AVE for a construct should be substantially higher than the squared correlation between that construct and all other constructs. Evidence of discriminant validity is also provided by the AVE method presented. The highest squared correlation was observed between perceived ease of use and perceived usefulness and

Table 10 Correlations and squared correlation between perceived ease of use, perceived usefulness, security protection, and use intention Table 9 Completely standardized expected change in Kx Items

n1 (security protection)

n2 (perceived ease of use)

n3 (perceived usefulness)

n4 (use intention)

SE1 SE2 SE3 SE4 PE1 PE2 PU1 PU2 UI1 UI2 UI3 UI4

– – – –

0.167 0.008 0.031 0.008 – – 0.049 0.047 0.082 0.026 0.074 0.175

0.030 0.031 0.067 0.069 0.026 0.020 – – 0.044 0.049 0.067 0.145

0.122 0.064 0.040 0.006 0.003 0.003 0.023 0.022 – – – –

0.028 0.021 0.062 0.060 0.006 0.057 0.102 0.065

Measures

AVEa Security Perceived Perceived Use protection ease of use usefulness intention (SE) (PE) (PU) (UI)

SE PE

0.660 0.671

PU UI

1 0.496** (0.246)b 0.670 0.448** (0.201) 0.700 0.254* (0.064)

1 0.516** (0.266) 0.398** (0.158)

1 0.245 (0.060)

1

a Average variance extracted (AVE) = (sum of squared standardized loading)/[(sum of squared standardized loadings) + (sum of indicator measurement error)]. b Squared correlation. * Correlation is significant at the 0.05 level. ** Correlation is significant at the 0.01 level.

862

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

it was 0.266. This was significantly lower than their individual AVEs. The AVE for the latent variables was 0.671 and 0.670, respectively. The results have demonstrated evidence of discriminant validity for the study constructs. If there is a lack of evidence supporting discriminant validity, a revision of the scales along with a collection of new data may be warranted before further analysis can be undertaken with confidence (Koufteros, 1999). 6.6. Construct reliability and variance extracted measures Estimates of the reliability and variance extracted measures for each construct are needed to assess whether the specified items sufficiently represent the constructs. The reliability of a construct can be estimated using AMOS output. Construct reliability means that a set of latent indicators of constructs are consistent in their measurement. In more formal terms, this reliability is the degree to which a set of two or more indicators share the measurement of a construct. Highly reliable constructs are those in which the indicators are highly intercorrelated, indicating that they are all measuring the same latent construct. The range of values for reliability is between 0 and 1. Computations for each construct are shown in Table 11. The reliability of the constructs of security protection, perceived ease

Table 11 Descriptive statistics and construct reliability for each construct Measures

Meana

S.D.b

Construct reliabilityc

Security protection (SE2, SE3, SE4) Perceived ease of use (PE1, E2) Perceived usefulness (PU1, U2) Use intention (UI1, UI2, UI4)

6.541

0.363

0.890

6.377

0.451

0.776

5.877

0.595

0.820

6.165

0.783

0.874

a The mean scores of perceived usefulness, perceived ease of use and security protection are based on a seven-point scale where 1 = very unimportant to 7 = very important, whereas use intention is 1 = strongly dislike to use to 7 = strongly like to use. b SD = standard deviation. c Construct reliability = (sum of standardized loadings)2/[(sum of standardized loadings)2 + (sum of indicator measurement error)]. Indicator measurement error can be calculated as 1  (standardized loading)2.

of use, perceived usefulness, and use intention were 0.890, 0.776, 0.820, and 0.874, respectively. All constructs exceeded the recommended level of 0.70 (Hair et al., 1998). The average variance extracted measures the amount of variance in the specified indicators accounted for by the latent construct. Higher variance extracted values occur when the indicators are truly representative of the latent construct. The variance extracted value is a complementary measure for the construct reliability value (Koufteros, 1999). Table 10 shows that among the AVEs of the measures, security protection (SE) had the lowest value of 0.66, indicating that 66% of the variance in the specified indicators was accounted for by the construct. All of the constructs had a variance extracted value that was higher than the recommended level of 50%. In sum, the overall results of the goodness-of-fit of the model and the assessment of the measurement model lent substantial support to confirming the proposed model. 6.7. Results of hypothesis testing The model’s overall fit with the data was evaluated using common model goodness-of-fit measures estimated by AMOS 5.0. Overall, our model exhibited a reasonable fit with the data collected. Based on the data, the AMOS estimation of our model showed a value of 1.146 in the Chi-square to degree of freedom ratio, which is satisfactory with respect to the commonly recommended value of less than 2.0. We assessed the model fit using other common fit indices: goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), comparative fit index (CFI), root mean square residual (RMSR), root mean square error of approximation (RMSEA), standardized residual, and modification index (MI). The model exhibited a fit value exceeding or close to the commonly recommended threshold for the respective indices, e.g., values of 0.913 and 0.986 for the GFI and CFI, satisfactory with respect to the commonly recommended value of equal to 1.0. We also tested the hypotheses based on the model as shown in Fig. 1. As summarized in Table 12, the specified relationship between perceived ease of use and use intention was supported by the data, as indicated by a significant critical ratio (C.R. = 2.002). The C.R. is a t-value obtained by dividing the estimate of the covariance by its

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867 Table 12 Results of the structural equation modeling Variables Perceived ease of use ! Use intention Perceived usefulness ! Use intention Security protection ! Use intention Security protection ! Perceived ease of use Security protection ! Perceived usefulness Perceived ease of use ! Perceived usefulness

7. Conclusions and discussions S.E.a

C.R.b

0.406

0.203

2.002

0.036

0.149

0.241

0.074

0.157

0.470

0.474

0.126

3.772

0.297

0.160

1.850

0.473

0.194

2.437

Estimate c

863

Fit indices: v2 = 55.038 (p = 0.226), df = 48, v2/df = 1.146, GFI = 0.913, AGFI = 0.859, CFI = 0.986, RMSEA = 0.042. a S.E. is an estimate of the standard error of the covariance. b C.R. is the critical ratio obtained by dividing the covariance estimate by its standard error. c Underlined values are critical ratios exceeding 1.96, at the 0.05 level of significance.

standard error. A value exceeding 1.96 represents a level of significance of 0.05. Ease of use of Internet services may be conducive to reaching a higher level of use intention. This reflects that perceived ease of use was the most important determinant of Internet services in liner shipping throughout our investigation. On the other hand, the effects of perceived usefulness (C.R. = 0.241) and security protection (C.R. = 0.470) on use intention were not supported by the results. Security protection was found to have a significant effect on the level of perceived ease of use (C.R. = 3.772). Nevertheless, there was a lack of support for a positive relationship between security protection and use intention, but there was a significant positive relationship between perceived ease of use and security protection. The results imply that security protection may influence perceived ease of use, which, in turn, affects shippers’ use intention for Internet services. In addition, we found that the effect of perceived ease of use on perceived usefulness was significant (C.R. = 2.437). In sum, the tests of the structural model showed that perceived ease of use affects shippers’ intention to use Internet services of liner shipping. Perceived ease of use had a positive effect on perceived usefulness. The data also showed that security protection positively affects perceived ease of use. The above findings are consistent with those from the studies of Davis et al. (1989), and Venkatesh and Davis (2000).

SEM modeling is a powerful tool that enables researchers to go beyond factor analysis into the arena of determining whether one set of unobserved constructs is related to another set of constructs. It also provides OR researchers with an effective means to analyze behavioral-related OR problems. In studies of travel behavior, it is often the case that the variables under study cannot be directly observed or measured (e.g., motivation, satisfaction, importance, perception) yet these unobserved variables might be hypothesized to be related to one another. SEM analysis is a methodology capable of handling this type of analysis, along with more conventional regression models, and simultaneous regression models, while accounting for multicollinearity and other assumptions of regression modeling (Reisinger and Turner, 1999). While it has been widely used in a number of disciplines, it has rarely been used in maritime research and to study OR problems. From the perspective of methodology, traditional techniques of assessing measurement scales may be inadequate when it comes to evaluating unidimensionality. This study has demonstrated that SEM offers many benefits over the traditional techniques of Cronbach’s alpha and EFA. The adequacy of the model in explaining the sample data can be statistically tested using a variety of model fit indices. The failure of the model to fit the data would alert the researcher to deviations from unidimensionality of one or both of the constructs. Such statistical tests of dimensionality are not generally available through EFA. Unidimensionality alone, however, does not ensure the usefulness of a scale. Other measurements such as convergent validity, discriminant validity and construct reliability may prove to be necessary for the unconfounded evaluation of structural analysis. This paper presents a step-by-step approach to testing the properties of measurement constructs. In particular, evaluating unidimensional properties requires a careful development of scales. Justifications for the scales should be based on theory and statistical methodology. A secondary purpose of this study was to develop and empirically test a model on the use of Internet services in liner shipping companies. The final model has a number of implications for research and practice. The influence of security protection on use intention (H1a) and perceived

864

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

usefulness (H1b) were not supported in this study. Why were these hypotheses not supported in this study? Based on our subsequent interviews with the respondent shippers, several reasons emerged. First, liner shipping companies provide a platform in their websites for communication between carriers and shippers. Shippers tend not to worry about their commercial secrets (e.g., thefts of data and disclosures) while they use their carriers’ website services. Second, shippers believe that the security measures of their carriers to protect on-line transactions and data transmissions through Internet services are reliable. Third, the legal protection of on-line transactions (e.g., denial of service and fraud) has been established in our study context, i.e., Taiwan. Finally, the issue of electronic billof-lading has been legally approved. Nevertheless, the proposed path from security protection to perceived ease of use was supported (H1c). This result suggests that shippers feel more secure easy in using the information technology if there is more legal and technical protection for their transactions, and would be more willing to use Internet services. In this respect, liner shipping practitioners could take steps to reduce the uneasy feelings about Internet security protection by advertising that their websites are safe and by publicly explaining what sort of security protection they utilize. Needless to say, shipping practitioners should continue to enhance the security of their Internet services, decrease the number of negative experiences, and raise shippers’ perceptions of ease of use in Internet services. Our results are consistent with those of Belanger et al. (2002) in which they suggested that security features are less important than perceived ease of use to stimulate customer intention to purchase via electronic commerce. In addition, our results indicate that the hypothesis of a positive relationship between perceived ease of use and the perceived usefulness of Internet services in liner shipping (H2a) was supported. The findings reveal that shippers’ perceptions of the usefulness of Internet services could be raised through perceived ease of use of Internet services by offering such desirable design functions as speedy links and prompt responses. A positive, significant relationship was found between the perceptions of shippers on the ease of using Internet services and their intention to use them (H2b). This implies that the functionality of the information system system must be emphasized to stimulate more shippers to use carriers’ Internet services.

Education and training programs could aim to raise the awareness of shippers to potential applications and emphasize the benefits of using Internet services. Also, when developing Internet services, software developers must address the issue of quick responses as an important design objective. Conversely, the effect of perceived usefulness on use intention (H3) was not found in this study. This reflects that shippers do not perceive usefulness of carriers’ Internet services and these might include reductions of documentation cost and communication cost. This paper makes an important contribution to the field of OR in three ways. First, given the timely need for OR researchers to comprehend and adopt SEM in examining behavioral-related problems, this paper will help in this instructional process. Second, because there seems to be a lack of SEM applications in the OR literature to date, this paper should provide the reader with a useful methodological guide to follow when he or she employs SEM in their research. Finally, this paper contributes to the body of OR literature with findings on Internet service adoption in liner shipping context. The findings of this study indicate that the model explaining the intention of shippers to use Internet services in liner shipping is acceptable. However, the test of indirect effects among the factors was not conducted in this study. Further testing of such indirect effects through comparative models is required or using different samples (e.g., smalland medium-sized enterprises). In addition, researchers may build on this model to identify and examine other factors that may influence shippers to use Internet services, such as the organizational readiness of shippers, including the level of information technology in the organization, computer resources and external pressure. The integration of these constructs into the model will help researchers and shipping practitioners to further grasp the factors influencing the development of Internet services or electronic commerce in the industry. Acknowledgements We are grateful for the helpful comments of an anonymous referee on earlier versions of this paper. This research was supported in part by a research grant (G-YE25) from The Hong Kong Polytechnic University.

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

865

Appendix A. Prior research on Internet services and the measurement items used for liner shipping in this study Previous studies

Measure

Elofson and Robinson (1998), Suh and Han (2003), Strader and Shaw (1997), Miyazaki and Fernandez (2001)

Please indicate, from your point of view, how important each items affecting your company to use Internet services of liner shipping companies (from 1 = very unimportant to 7 = very important) Security protection: SE1: The ability to provide technical protection for on-line transactions such as quoting, booking space and issuing B/Ls for shippers SE2: The legal recognizance of electronic B/L SE3: The ability to provide protection of commercial secrets (e.g., from thefts of data and disclosures) SE4: The ability to provide legal protection of transactions (e.g., from denial of service and fraud) SE5: The ability to provide protection of data transmission

Agarwal and Prasad (1999), Venkatesh and Davis (2000), Igbaria et al. (1997), Lederer et al. (2000), Lucas and Spitler (1999), Olson and Boyer (2003), Grandon and Pearson (2004)

Perceived usefulness: PU1: Using shipping companies’ Internet services to documentation cost PU2: Using shipping companies’ Internet services to communication cost (e.g., telephone and fax) PU3: Using shipping companies’ Internet services to documentation errors PU4: Using shipping companies’ Internet services to to receive bill(s) of lading

reduce reduce improve make it faster

Agarwal and Prasad (1999), Venkatesh and Davis (2000), Davis et al. (1989), Igbaria et al. (1997), Lederer et al. (2000), Lucas and Spitler (1999), Olson and Boyer (2003), Grandon and Pearson (2004)

Perceived ease of use: PE1: The speed to link shipping companies’ Internet services PE2: The speed of responses for my enquires in shipping companies’ Internet services PE3: The accurate information in shipping companies’ Internet services PE4: Ease to use shipping companies’ Internet services

Agarwal and Prasad (1999), Cheng et al. (2002), Olson and Boyer (2003)

Please indicate, from your point of view, how likely each items is to your company to use Internet services of liner shipping companies (from 1 = extremely unlikely to use to 7 = extremely likely to use) Use intention: UI1: Using the tracking of containers function in shipping companies’ Internet services UI2: Enquiring sail schedules in shipping companies’ Internet services UI3: Checking the condition of customs clearance of my firm’s cargo in shipping companies’ Internet services UI4: Booking space in shipping companies’ Internet services

References Agarwal, R., Prasad, J., 1999. Are individual differences germane to the acceptance of new information technologies?. Decision Sciences 30 361–391. Agho, A., Price, J., Mueller, C., 1992. Discriminant validity of measures of job satisfaction, positive affectivity and negative

affectivity. Journal of Occupational and Organizational Psychology 65 (3), 185–196. Ajzen, I., Fishbein, M., 1980. Understand Attitudes and Predicting Social Behavior. Prentice-Hall, Englewood Cliffs, NJ. Ajzen, I., 1985. From intention to actions: A theory of planned behavior. In: Kuhl, J., Bechmann, J. (Eds.), Action Control:

866

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867

From Cognition to Behavior. Springer Verlag, New York, pp. 11–39. Ajzen, I., 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes 50, 179–211. Ajzen, I., Madden, T.J., 1986. Prediction of goal-directed behavior: Attitudes, intention and perceived behavioral control. Journal of Experimental Social Psychology 22, 453– 474. Anderson, J.C., 1987. An approach for confirmatory measurement and structural equation modeling of organizational properties. Management Science 33 (4), 525–541. Anderson, J.C., Gerbing, D.W., Hunter, J.E., 1987. On the assessment of unidimensional measurement: Internal and external consistency, and overall consistency criteria. Journal of Marketing Research XXIV (November), 432–437. Arbuckle, J.L., 1997. Amos Users’ Guide Version 3.6. SmallWaters Corporation, Chicago, IL. Armstrong, S.J., Overton, T.S., 1977. Estimating nonresponse bias in mail survey. Journal of Marketing Research 14 (August), 396–402. Babakus, E., Mangold, G., 1992. Adapting the SERVQUAL scale to hospital services. Health Services Research 26 (6), 767. Belanger, F., Hiller, J.S., Smith, W.J., 2002. Trustworthiness in electronic commerce: The role of privacy, security, and sit attributes. Strategic Information Systems 11, 245–270. Bollen, K.a., 1989. Structural Equations with Latent Variables. Wiley-Interscience Publication. Byrne, B.M., 2001. Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming. Lawrence Erlbaum Associates, New Jersey. Chau, P.Y.K., Hu, P.J.H., 2001. Information technology acceptance by individual professionals: A model comparison approach. Decision Science 32 (4), 699–719. Cheng, C.H., Cheung, W., Chang, M.K., 2002. The use of the Internet in Hong Kong: Manufacturing vs. service. International Journal of Production Economics 75, 33–45. Cheng, T.C.E., Lam, D.Y.C., Yeung, A.C.L., in press. Adoption of Internet banking: An empirical study in Hong Kong. Decision Support Systems. Churchill, G.A., 1979. A paradigm for developing better measures of marketing constructs. Journal of Marketing Research 16 (November), 360–375. Churchill, G.A., 1991. Marketing Research: Methodological Foundation, fifth ed. The Dryden Press, New York. Davis, F.D., 1986. A technology acceptance model for empirically testing new end user information systems: Theory and results. Doctoral Dissertation, Sloan School of Management, Massachusetts Institute of Technology. Davis, F.D., 1989. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly 13 (3), 319–339. Davis, F.D., Bagozzi, R.P., Warshaw, P.R., 1989. User acceptance of computer technology: A comparison of two theoretical models. Management Science 35, 982–1003. Dunn, S.C., Seaker, R.F., Waller, M.A., 1994. Latent variables in business logistics research: Scale development and validation. Journal Business Logistics 15 (2), 145–172. Eastin, M.S., 2002. Diffusion of e-commerce; An analysis of the adoption of four e-commerce activities. Telematics and Informatics 19, 251–267.

Elofson, G., Robinson, W.N., 1998. Creating a customer massproduction channel on the Internet. Communications of the ACM 41 (3), 56–62. Etezadi-Amoli, J., Farhoomand, A.F., 1996. A structural model of end user computing satisfaction and user performance. Information and Management 30 (2), 65–73. Fabrigar, L.R., Wegener, D.T., MacCallum, R.C., Strahan, E.J., 1999. Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods 4, 272–299. Fishbein, M.A., Ajzen, I., 1975. Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research. AddisonWesley, Reading, MA. Fornell, C.R., Larcker, D.F., 1981a. Structural equation models with unobservable variables and measurement error. Journal of Marketing Research 18, 39–50. Fornell, C., Larcker, D.F., 1981b. Evaluating structural equation models with unoberservable variables and measurement error. Journal of Marketing Research 18, 186–192. Garver, M.S., Mentzer, J.T., 1999. Logistics research methods: Employing structural equation modeling to test for construct validity. Journal of Business Logistics 20 (1), 33–58. George, J.F., 2002. Influences on the intent to make Internet purchases. Internet Research 12 (2), 165–180. George, J.F., 2004. The theory of planned behavior and Internet purchasing. Internet Research 14 (3), 198–212. Gerbing, D.W., Anderson, J.C., 1988. An updated paradigm for scale development incorporating unidimensionality and its assessment. Journal of Marketing Research 25, 186–192. Golob, T.F., 2003. Structural equation modeling for travel behavior research. Transportation Research Part B 37, 1–25. Grandon, E.E., Pearson, J.M., 2004. Electronic commerce adoption: An empirical study of small and medium US businesses. Information & Management 42 (1), 197–216. Hair, J., Anderson, R., Tatham, R., Black, W., 1995. Multivariate Data Analysis with Readings, fourth ed. Prentice Hall International, Englewood Chiffs, NJ. Hair, J., Anderson, R., Tatham, R., Black, W., 1998. Multivariate Data Analysis with Readings, fifth ed. Prentice Hall International, Upper Saddle River, NJ. Harrison, D.A., Mykytyn, P.P., Riemenschneider, C.K., 1997. Executive decisions about adoption of information technology in small business: Theory and empirical tests. Information Systems Research 8 (2), 171–195. Heijden, H.V.D., 2003. Factors influencing the usage of websites: The case of a generic portal in The Netherlands. Information & Management 40, 541–549. Hu, P.J.H., Clark, T.H.K., Ma, W.W., 2003. Examining technology acceptance by school teachers: A longitudinal study. Information & Management 41, 227–241. Hoffman, E., Novak, T.P., Peralta, M., 1999. Building consumer trust online. Communication of the ACM 42 (4), 80–86. Hullnad, J., Chow, Y.H., Lam, S., 1996. Use of causal models in marking research: A review. International Journal of Research in Marketing 13 (2), 181–197. Hutcheson, G., Sofroniou, N., 1999. The Multivariate Social Scientist: Introductory Statistics using Generalized Linear Models. Sage Publications, Thousand Oaks, CA. Igbaria, M., Zinatelli, N., Cragg, P., Cavaye, A.L.M., 1997. Personal computing acceptance factors in small firms: A structural equation model. MIS Quarterly (September), 279– 305.

C.-S. Lu et al. / European Journal of Operational Research 180 (2007) 845–867 Jeffery, K., 1999. Electronic Commerce and container Shipping. Cargo Systems, London. Jo¨reskog, K.G., 1993. Testing structural equation models. In: Bollen, K.A., Long, L.S. (Eds.), Testing Structural Equation Models. Newbury Park Sage Publications, CA, pp. 94–316. Jo¨reskog, K.G., So¨rbom, D., 1996. LISREL 8: User’s Reference Guide. Scientific Software, Chicago, IL. Kalakota, R., Whinston, A.B., 1996. Frontiers of Electronic Commerce. Addison-Wesley, Reading, MA. Khalifa, M., Limayem, M., 2003. Drivers of Internet shopping. Communications of the ACM 46 (12), 233–239. Kim, J.O., Muller, C.W., 1978. Introduction to Factor Analysis: What It Is and How To Do It, Series: Quantitative Applications in the Social Sciences, Sage University Paper No. 9. Kline, Rex B., 1998. Principles and Practice of Structural Equation Modeling. Guilford Press, NY. Koufteros, X.A., 1999. Testing a model of pull production: A paradigm for manufacturing research using structural equation modeling. Journal of Operations Management 17, 467– 488. Koufteros, X.A., Vonderembse, M.A., Doll, W.J., 1998. Developing measures of time-based manufacturing. Journal of Operations Management 16, 21–41. Koufteros, X.A., Vonderembse, M.A., Doll, W.J., 2001. Concurrent engineering and its consequences. Journal of Operations Management 19, 97–115. Koufteros, X.A., Vonderembse, M.A., Doll, W.J., 2002. Integrated product development practices and competitive capabilities: The effects of uncertainty, equivocality, and platform strategy. Journal of Operations Management 20, 331–355. Lederer, A.L., Maupin, D.J., Sena, M.P., Zhuang, Y., 2000. The technology acceptance model and the World Wide Web. Decision Support Systems 29, 269–282. Lewis-Beck, M.S., 1994. Factor Analysis and Related Techniques. Sage Publication Ltd. Litwin, M.S., 1995. How to Measure Survey Reliability and Validity. Sage Publication, London. Long, S., 1983. Confirmatory Factor Analysis: A Preface to LISREL. Sage Publications, London. Lucas Jr, H.C., Spitler, V.K., 1999. Technology use and performance: A field study of broker workstation. Decision Science 30 (2), 291–311. Mehrtens, J., Cragg, P.B., Mills, A.M., 2001. A model of Internet adoption by SMEs. Information & Management 39, 165–176. Miyazaki, A.D., Fernandez, A., 2001. Consumer perceptions of privacy and security risks for online shopping. Journal of Consumer Affairs 35 (1), 27–44. Nahm, A.Y., Vonderembse, M.A., Koufteros, X.A., 2003. The impact of organizational structure on time-based manufacturing and plant performance. Journal of Operations Management 21, 281–306. Nunnally, J.C., 1978. Psychometric Theory, second ed. McGrawHill, New York. O’Leary-Kelly, S.W., Vokurka, R.J., 1998. The empirical assessment of construct validity. Journal of Operations Management 16 (4), 387–405. Olson, J.R., Boyer, K.K., 2003. Factors influencing the utilization of Internet purchasing in small organizations. Journal of Operations Management 21, 225–245.

867

Peter, J.P., 1979. Reliability: A review of psychometric basics and recent marketing practices. Journal of Marketing Research 16 (February), 6–17. Peter, J.P., 1981. Construct validity: A review of basic issues and marketing practices. Journal of Marketing Research 18 (May), 133–145. Phillips, L.W., Bagozzi, R.P., 1986. On measuring the organizational properties of distribution channels: Methodological issues in the use of key informants. In: Sheth, J. (Ed.), Research in Marketing, vol. 8. JAI Press, Greenwich, CT, pp. 313–369. Reisinger, Y., Turner, L., 1999. Structural equation model with Lisrel: Application in tourism. Tourism Management 20, 71– 88. Rigdon, E.E., 1998. Structural equation modeling. In: Marcoulides, G.A. (Ed.), Modern Methods for Business Research. Lawrence Erlbaum Associates, Mahwah, NJ. Segar, A., 1997. Assessing the unidimensionality of measurement: A paradigm and illustration within the context of information systems research. Omega 25 (1), 107–121. Shen, H., Bentler, P., Comerey, A., 1995. A comparison of models of medical schools student selection. Structural Equation Modeling 2 (2), 93–100. Shah, R., Goldstein, S.M., 2006. Use of structural equation modeling in operations management research: Looking back and forward. Journal of Operations Management, 148– 169. Song, J., Zahedi, F., 2001. Web design in e-commerce: A theory and empirical analysis. In: Proceedings of the 22nd International Conference on Information Systems, pp. 205– 220. Stank, T.P., Keller, S.B., Daugherty, P.J., 2001. Supply chain collaboration and logistics service performance. Journal of Business logistics 22 (1), 29–47. Steenkamp, J.E.M., Baumgartner, H., 2000. On the use of structural equation models for marketing modeling. International Journal Research in Marketing 17, 195–202. Strader, T., Shaw, M., 1997. Characteristics of electronic markets. Decision Support Systems 21, 185–198. Suh, B., Han, I., 2003. The impact of customer trust and perception of security control on the acceptance of electronic commerce. International Journal of Electronic Commerce 7 (3), 135–161. Taylor, S., Todd, P.A., 1995a. Understanding information technology usage: A test of competing models. Information Systems Research 6 (2), 144–176. Taylor, S., Todd, P.A., 1995b. Assessing IT usage: The role of prior experience. MIS quarterly 19 (4), 561–570. United Nations Conference on Trade and Development, 2002. Ecommerce and Development Report 2002. United Nations. Venkatesh, V., Davis, F.D., 2000. A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science 46 (2), 186–204. Widaman, K.F., 1993. Common factor analysis versus principal components analysis: Differential bias in representing model parameters. Multivariate Behavioral Research 28, 263–311. Wiethoff, C., 2004. Motivation to learn and diversity training: Application of the theory of planned behavior. Human Resource Development Quarterly 15 (3), 263–277.