ABSTRACT The deployment of cryptocurrencies in e-commerce has reached a significant number of transactions and continuous increases in monetary circulation; nevertheless, they face two impediments: a lack of awareness of the technological utility, and a lack of trust among consumers. E-commerce carried out through social networks expands its application to a new paradigm called social commerce. Social commerce uses the content generated within social networks to attract new consumers and influence their behavior. The objective of this paper is to analyze the role played by social media in increasing trust and intention to use cryptocurrencies in making electronic payments. It develops a model that combines constructs from social support theory, social commerce, and the technology acceptance model. This model is evaluated using the partial least square analysis. The obtained results show that social commerce increases the trust and intention to use cryptocurrencies. However, mutual support among participants does not generate sufficient trust to adequately promote the perceived usefulness of cryptocurrencies. This research provides a practical tool for analyzing how collaborative relationships that emerge in social media can influence or enhance the adoption of a new technology in terms of perceived trust and usefulness. Furthermore, it provides a significant contribution to consumer behavior research by applying the social support theory to the adoption of new information technologies. These theoretical and practical contributions are detailed in the final section of the paper.
PLS-SEM uses a similar approach to factorial analysis of major components. PLS combines analysis of principal components, path and regression . A path model includes a set of latent variables (or constructs). It is dened by two components: a structure model and a measurement model . PLS-SEM is useful for analyzing constructs and, at the same time, to evaluate the structural model , . Therefore, PLS-SEM is appropriate for the proposed analysis and research. The rst step is to validate the measurement model, to then generate the estimates using the bootstrapping technique with 500 resamples. The software used in the evaluation is SmartPLS version 3 . C. EVALUATION OF THE MEASUREMENT MODEL The evaluation of the measurement model is important because it guarantees that the results to obtain in later phases will be: (i) reliable and (ii) valid.
id: 7ed103544d729afe64c655e50100c7b2 - page: 7
(i) In order to evaluate the reliability of the constructs, two indices are used: Cronbachs alpha and composite reliability. The internal consistency reliability represents the homogeneity in the constructs. According to Hair et al. , the suggested value for both indices should be greater than 0.7; although for exploratory investigations a value of 0.60 is considered to be acceptable. The reliability of the indicators is evaluated through their factor loading, whose value must be greater than 0.5 . VOLUME 6, 2018 50743 J. C. Mendoza-Tello et al.: S-commerce as a Driver to Enhance Trust and Intention to Use Cryptocurrencies for Electronic Payments TABLE 3. Construct reliability and convergent validity.
id: 00a4b8dbb33b2733da3c8307a6d169ad - page: 7
TABLE 4. Discriminant validitys. In the proposed model, reective indicators are used. The results obtained demonstrate that the indices of internal consistency reliability of the constructs exceed the value of 0.81 (Table 3), and that the factor loading of the indicators is between 0.744 and 0.983 (appendix B). Therefore, the measurement model satises the reliability criteria. D. EVALUATION OF THE STRUCTURAL MODEL The structural model is evaluated after proving that the measurement model satises the criteria of reliability and validity. One of the tasks in evaluating the structural model is testing the hypothesis, whose results make it possible to analyze (i) the standardized path coefcients and (ii) the coefcient of determination (R2).
id: 6dc65f025e28683d68aea5347ada1d55 - page: 8
(ii) A construct should satisfy the criteria of validity, that is high correlations between the items within the same construct (convergent validity), and low correlations between the items in different constructs (discriminant validity) . The Average Variance Extracted (AVE) is a convergent measure of validity, whose suggested value should be above 0.5 . As show in Table 2, AVE values for each construct exceed the 0.5 threshold and demonstrate that the model also meets the convergent validity criteria. Discriminant validity is evaluated using the criteria suggested by Chin . As show in Table 4, the correlation between any two constructs is less than the square root of the AVE shared by the indicators within the construct. The discriminant validity can also be evaluated through a cross-loading analysis . As show in Appendix B, each block of indicators has a greater load within their respective construct than in the others. According to these two criteria (Fornell and cr
id: d15ac2fca69fe3117b69aeae98bb1635 - page: 8