Data was processed and analyzed using the multivariate statistical approach of partial path least squares (PLS-MP). This technique was chosen because it is recommended when the purpose is to confirm a theory from the constructs and the sample size is small and/or the data is not normally distributed(Ringle et al., 2014). All the constructs of the measurement model predicted in Fig. 2, as well as their interactions, are based on the theoretical survey, and can be classified as reflective(Hair, Black, Babin, and Anderson, 2014a). For reflective measurement models, we use reliability analysis of internal consistency, convergent validity and discriminant validity(Hair, Hult, Ringle, and Sarstedt, 2014b). The reliability of the internal consistency can be analyzed by composite reliability (CR) and Cronbach's alpha, and both should exceed 0.70 (Ringle et al., 2014). The convergent validity was analyzed by the average variance extracted (AVE) and should exceed 0.50 (Hair, Hult, Ringle, and Sarstedt, 2014b). For discriminant validity, the square root of the AVE should be greater than the correlation between the latent variables (Hair,Hult, Ringle, and Sarstedt, 2014b)