臺大管理論叢第31卷第3期

175 NTU Management Review Vol. 31 No. 3 Dec. 2021 gender (male, 34 %; female, 66%), age (less than or equal to 30 years of age, 30%; 3140 years of age, 27%; 41-50 years of age, 28 %; more than or equal to 51 years of age, 15 %), annual income (less than or equal to US $20,000, 8%; US $20,001-US $40,000, 34%; US $40,001-US $60,000, 40%; more than or equal to US$60,001, 18%), and relationship length (less than or equal to 1 year, 21%; 2-5 years, 30%; 6-10 years, 26%; more than or equal to 11 years, 23%), mutlichannel retailers (Zara, 13%, Gap, 26%, Uniqlo, 30%, H & M, 21%, others, 10%). 3.2 Measures We measure multichannel integration, perceived affordances, economic value, relational value, online satisfaction, offline satisfaction, online loyalty and offline loyalty by a five-point Likert-type scale. First, we revise six items from Lee and Kim (2010) for the measurement of multichannel integration. We adapt persistence, interactivity (active control, two-way communication, and synchronicity) and selectivity (customization) from Fang (2019) to measure perceived affordances. Three items are adapted from Chan et al. (2010) to measure both economic value and relational value. Four items taken from Srinivasan, Anderson, and Ponnavolu (2002) are used to measure online satisfaction. Another four items are adapted from Srinivasan et al. (2002) and Garbarino and Johnson (1999) to measure offline satisfaction. Second, since the boundaries between online and offline are becoming blurred— online shopping will also refer to offline services, and offline purchases will also refer to online information, for the measurement of online loyalty, we want to measure customers’ future behavioral intention and thus adapt four items from Yang and Peterson (2004) and Srinivasan et al. (2002). For the measure of offline loyalty, since we take it as customer loyalty, we again adapt four items from Yang and Peterson (2004). 3.3 Validation of Measures We estimate the Structural Equation Modeling (SEM) for hypotheses testing. The fit of the model is suitable (chi-square (467) =1414.368, p = 0.00; GFI = 0.85; CFI = 0.96; PNFI = 0.82; NNFI = 0.96; RMSEA = 0.08; RMR = 0.06). NNFI and CFI are greater than the suggested 0.9. These two fit indices indicate a good fit of measurement model, particularly for a model with such a large number of constructs (Hoyle and Panter, 1995; Taylor and Todd, 1995). The Cronbach’s alphas of all constructs are all greater than 0.80, which support the reliability of the measurement. In addition, all composite reliabilities

RkJQdWJsaXNoZXIy ODg3MDU=