Research Methodology

---

Measures

Kettinger & Lee (1997) suggest three alternative service quality measures originally proposed by Parasuraman et al. (1994a) to improve IS-adapted service measures. They imply that (1) SERVPERF (one-column format), which measures direct ratings of the gap between the desired and the perceived service levels, is best for ease of application, situations requiring high predictive power, and possibility for a total service quality score, (2) SERVQUAL (two-column format), including the short form (Kettinger & Lee, 1994), may be most applicable when greater dimensional diagnostics are required, (3) SERVQUAL+ (three-column format), which generates separate ratings of desired, adequate, and perceived service with three identical, side-by-side scales, may prove to be an improved form over SERVQUAL and SERVPERF, but this format is longer in length.

In order to balance the tradeoff between ease of application, dimensional diagnostics, information enrichment, and scale length, this study adopted the two-column format SERVQUAL to develop the EC-SERVQUAL instrument. The SERVQUAL (Parasuraman et al., 1991) was slightly modified to apply to an e-commerce context for digital products marketing. For example, the third question was originally asked in terms of "employees." We changed the wording to "screen displays" because employees' dresses could not be perceived in a digital marketing environment.

This EC-adapted SERVQUAL was pre-tested through a series of interviews with e-business professionals and graduate students. After careful examination of the results of pre-testing, additional wording adjustments were made. Each item was then turned into two statements-one to measure expectations and one to measure perceptions (see Appendix A). Respondents were asked to rate each item on a seven-point scale anchored on strongly disagree (1) and strongly agree (7). For each item a difference (gap) score was produced through subtracting its corresponding expected item from perceived item.

Data Collection

To increase the generalizability of the results, data is from consumers of a variety of digital product providers. The respondents' comments concerned the following thirteen digital product categories, including paper-based information products, product information, graphics, audio, video, tickets and reservations, financial instruments, government services, electronic messaging, business value creation processes, auctions and electronic mark, remote education, and interactive entertainment. The categories were chosen to represent a broad range of digital products proposed by Choi et al. (1997). Data for validation and refinement of the EC-SERVQUAL instrument was gathered from a quota sample of 260 adult respondents (20 years of age or older) attending an e-business exposition and symposium held in Taiwan, with an equal quota of 20 responses from each category of digital products. Specific examples of digital products or services purchased by the respondents were shown in Table 3.

Table 3: Examples of Digital Products Purchased by the Respondents

Digital Product Categories	Examples
Paper-based information products	News, magazines
Product information	Software, user manuals
Graphics	Photographs, maps
Audio	Music, speeches
Video	Movies, television
Tickets and reservations	Airplane, hotel
Financial instruments	Securities, Internet banking
Government services	Tax payments
Electronic messaging	Email, faxes
Business value creation processes	Ordering
Auctions and electronic mark	Online biding
Remote education	Asynchronous e-learning
Interactive entertainment	Online games

Anonymous and self-administered questionnaires containing 22 items from the initial EC-SERVQUAL instrument were distributed to the 260 screened and qualified respondents who had experience in using the service of the web site providing the digital products in question during the past three weeks. For each question within the initial EC-SERVQUAL instrument, respondents were asked to circle the response which best described their level of agreement with the statements. The ratio of sample size to number of items (approximately 12:1) was well above the minimum 10:1 ratio for factor analysis recommended by Kerlinger (1978).

Confirmatory Factor Analysis

The growing importance of buyer and seller interactions in B2C e-commerce leads us to investigate the question concerning the measurement of service quality in digital marketing environments. The question was tested by analyzing the result of a refined EC-version of the 1991 instrument of service quality (SERVQUAL) (Parasuraman et al., 1991). Confirmatory factor analysis (CFA) is needed to accomplish the validation and refinement of the instrument. CFA involves the specification and estimation of one or more putative models of factor structure, each of which proposes a set of latent variables (factors) to account for covariance among a set of observed variables (Bagozzi, 1980; Bollen, 1989; Jöreskog & Sörbom, 1993). As Byrne (1998) contends, CFA is appropriately used when researcher has some knowledge of the underlying latent variable structure. Based on knowledge of the theory, empirical research, or both, researchers postulate relations between the observed measures and the underlying factors a priori, and then tests this hypothesized structure statistically.

The advantages of applying CFA as compared to classical approaches (Churchill, 1979; Nunnally, 1978), such as common factor analysis and multi-trait multi-method analysis, to determine construct validity are widely recognized (Anderson & Gerbing, 1988; MacCallum, 1986; Marsh & Hocevar, 1985; Segars & Grover, 1993). In addition, a more rigorous specification that is required for a confirmatory factor analysis of a multiple-indicator measurement model can afford a more rigorous evaluation of unidimensionality (Gerbing & Anderson, 1988). We attempted to establish strong construct validity of the EC-SERVQUAL, including its convergent validity and discriminant validity, which can be directly tested by higher order CFA (Marsh & Hocevar, 1988).

---