Abstract
Work on how consumers evaluate electronic service quality is both topical and important due to the well-accepted criticality of electronic channels in selling products and services. However, most of the relevant research on electronic research quality is preoccupied with the website Internet context and most of the studies are single-country studies, inhibiting conclusions of generalizibility. Theoretically rooted in the Nordic Model of perceived service quality, this exploratory study uses an e-service quality scale to measure mobile Internet service quality in different national settings. Consistent with the available e-service quality literature, results indicate that e-service quality is a second-order factor, with three reflective first-order dimensions: efficiency, outcome, and customer care. Most important, cross-validation investigations using samples drawn from Korean, Hong Kong, and Japanese mobile Internet user populations, support the factorial structure invariance of the construct. Following CitationCheung and Reynolds's (2002) suggestions, factor means differences between the three countries contributing to the scarce cross-national electronic service quality literature are tentatively examined. These initial empirical findings imply that although consumers in different countries use the same dimensions to evaluate mobile Internet services, importance weightings assigned on these dimension are probably not the same.
Notes
1A marketing website that enables the creation of online panels consisting of mobile Internet (e.g. MIN i-mode monitors) and stationary internet users. MIN is an official research partner of the Electronic Commerce Promotion Council of Japan.
2Further support for the superiority of a second-order factor model can be found in the structural equation modeling literature. CitationChen et al. (2005) pointed to the next set of advantages: (a) a second-order model can test whether a hypothesized second-order factor can actually account for the pattern of relations between the first-order factors; (b) puts a structure on the pattern of covariance between the first-order factors; (c) separates variance due to specific factors (these specific factors are represented by the disturbance of each first-order factor), leading to a theoretically error-free estimation of the specific factors; and (d) can provide useful simplification of complex multitrait–multimethod models.
3We also checked for discriminant validity employing the most stringent criterion of CitationFornell and Larcker (1981). We tested whether AVE from each latent is greater than its shared variance with the other two latents (γ2). Results indicate that for each pair of latents AVE> γ2, though the AVE of “customer care” quality is marginally greater than squared correlation of “customer care” quality and “efficiency.”
4Regarding the high RMSEA value, we build from the structural equation modeling literature and point that the interpretation of any fit index in isolation could be problematic because trade-offs between Type I and Type II errors call for the interpretation of combinations of indexes in various model contexts. Another related issue is statistical power, which has to be taken into account when interpreting fit indices. In studies where power is overly great (i.e., > 0.9, as is the case with the present study) may require a more relaxed interpretation of fit than is typical. Conversely, a more stringent interpretation of fit statistics is required when power is low. The high statistical power of the present study and the acceptable values for the CFI and SRMR indices, seem to mitigate the somewhat high RMSEA values (see CitationZeithaml et al., 2002).
5Compared to people-processing services and possession-processing services.
6In terms of not being overly sensitive to small errors of approximation.