with six directly testable hypotheses that characterize the resulting consumer-satisfaction function in terms of empirically meaningful properties.

We first generalize from the TAM literature and propose that service quality in Internet banking and resulting consumer satisfaction depend on individual perceptions with regard to usefulness and ease of use. The empirical importance of these considerations to consumer attitudes toward Internet banking was investigated and established in [ 5]. Combining these results and applying them to the core framework, we obtain the hypotheses:

 

H1. Perceived usefulness (USE) is a positive determinant of CSIBS; and

H2. Perceived ease of use (EOU) is a positive determinant of CSIBS.

 

Reliability—a basic category in the Servqual protocol—has been found to be an empirically important determinant of service quality in many situations [ 12]. In Internet banking, concern over reliability would tend to focus on whether information access and transaction processes are expected to be operationally consistent and accurate. Applying these results to the core framework yields the hypothesis:

 

H3. Perceived reliability (REL) is a positive determinant of CSIBS.

Under Servqual modeling, security is understood in physical and financial terms, as well as in terms of privacy and the protection of data against unauthorized disclosure, modification, and destruction. In particular, privacy enters the analysis in the sense of individuals and organizations determining for themselves when, how, and to what extent personal and sensitive data is to be transmitted to others [ 9]. In Internet banking, security has been found to be a matter of intense concern, especially with regard to the acquisition and dissemination of personal and sensitive data. Perceptions regarding this aspect of service quality are generally operationalized in the form of transaction security, as represented directly by the safe and accurate transfer of funds and payment-credit information and indirectly by transaction risk [ 5]. These observations suggest the hypothesis:

 

H4. Perceived security (SEC) is a positive determinant of CSIBS.

In Servqual modeling, service responsiveness is generally captured in terms of the vendor’s ability to supply information with minimal time lag to make available problem-solving mechanisms, as well as provide guarantees when difficulties emerge [ 12]. As applied to e-service quality, responsiveness has been operationalized and studied in terms of promptness and efficiency [ 6]. These observations suggest an extension of Servqual modeling to the case of Internet banking in terms of the hypothesis:

 

H5. Perceived responsiveness (RES) is a positive determinant of CSIBS.

 

The Servqual idea of continuous improvement was proposed to depict service quality in relation to the vendor’s expected ability to meet changing consumer needs and requirements [ 10]. Such an attribute would be fundamental to competitive advantage in business areas characterized by rapid technological and institutional change (such as Internet banking), especially with regard to product-service innovation and enhancement to increase demand. Applying these ideas and results to the core framework suggests the hypothesis:

 

H6. Continuous improvement (IMP) is a positive determinant of CSIBS.

 

METHODOLOGY AND RESULTS

Our research methodology involvedthe standard areas of questionnaire design, survey implementation, and quantitative analysis. Our questionnaire was designed to allow Likert-scale measurement of the core framework’s perception-based constructs and service-quality attributes: consumer satisfaction with Internet banking, usefulness, ease of use, reliability, responsiveness, security, and continuous improvement. In 2005, we dispatched 500 questionnaires to individuals with experience in Internet banking in Hong Kong. A research sample of 182 meaningful replies was obtained.

We first performed a Cronbach test to determine the internal consistency of data obtained from multi-ple-item measurement of {USE, EOU, REL, RES, SEC, IMP}. The values we obtained ranged from 0.796 to 0.907, indicating satisfactory internal consistency with reference to the standard criterion of 0.7 (see Table 1). Correlation coefficients ranging from 0.457 to 0.758 indicate the existence of significant relationships (at the 0.01 level) among {USE, EOU, REL, RES, SEC, IMP} in the data, thereby supporting the combination of such attributes under linear modeling of the core framework.