Using exploratory factor analysis with a Varimax rotation, the reliability of the constructs with Cronbach’s alpha coefficient is presented in Table 1. After deleting ten (factor loading < 0.7) of the original sixteen, six items were found for technical bonds. The factor loading of six items was larger than 0.735 and Cronbach’s alpha waslarger than 0.733. As expected, the eight items for social bonds were identified. After removing 4 items, the study
measured social bonds using eight items, which according to Cronbach’s alpha were satisfactory at above 0.89 and
likewise for factor loading, which ranged from 0.786 to 0.9097.
Factor analyses by Garbarino and Jonson [1999] and Doney and Cannon [1997] revealed two measurements of
e-trust. Since the reliability of Doney and Cannon’s measurement was low (< 0.6) we excluded it from our final
analysis. The factor loading was above 0.84 for the remaining Garbarino and Jonson’s measurements and was
deemed acceptable. Similarly, in factor analyses of word-of-mouse [Kiecker & Cowles 2002], three items had a high
factor loading (>0.81) and Cronbach’s alpha 0.89. This investigation identified two more dimensions from other
literature [e.g., Lee et al. 2001]. The first factor focused on the dimensions of behavioral e-loyalty, the second on
attitudinal e-loyalty. Overall, the factor loading and reliability were good.