Table 6: Probit regression result for marginal effect on risk perception 

The marginal effects of the selected variables on risk perception towards e-commerce are shown at Table 6. The variables experienced of using internet, social networking, and market orientation have significant marginal effect on risk perception of the e-commerce customers at 5% significance level. The variable experienced of using internet is significant under 5% level of significance as its corresponding p-value is 0.012 which is less than 0.05. The value of its corresponding marginal effect is -0.205 which indicates that customers risk perception significantly decreases 20.5% towards e-commerce for an individual who is experienced at using internet compared to an individual who is not experienced at using internet. Customers risk perception significantly (p-value <0.001) increases 24.2% for an individual who use internet for social networking compared to others, keeping all other variables at a fixed level. Meanwhile, customers risk perception significantly (p-value 0.015) increases by 14.5% towards e-commerce for those who agree about market orientation compared to those who disagree about market orientation. 

Before proceeding for PCA, one has to check whether the data is appropriate for the analysis. For that reason, this study uses two statistics: the Bartlett test of Sphericity and the Kaiser-Meyer-Olkin Measure of Sampling Adequacy (Chiciudean and Chiciudean, 2013).The Kaiser-Meyer-Olkin tests whether the partial correlations among the considered variables are small or not (Trujillo-Ortiz et al., 2006). High values (close to 1.0) generally suggest that PCA may be useful with that data.

Table 7: KMO and Barletts Test 

Here at Table 7, Kaiser-Meyer-Olkin measure is 0.711 which indicates that the data set is valid for PCA. On the other hand, Bartletts test investigates the hypothesis that the underlying correlation matrix is an identity matrix, which would indicate that variables are not related (Tobias and Carlson, 1969). Small values (less than 0.05) of the significance level permit the use of PCA with the data. Here the p-value is less than 0.001 for Bartletts test that is under 5% level of significance there is enough evidence that PCA is useful for the data. So, this study can approach for the PCA with the dataset.

Table 8:  Descriptive Statistics 

Table 9: Communalities

The term communality means the total influence on  single observed variable from all the associated factors (Borůvka et al., 2005). Communality for “getting access to data without permission” variable is 0.637 that means 63.7% variation of this variable is explained by the associated factors. Similarly, communality for “risky when the products may be available at lower price” variable is 0.628 that means 62.8% variation of this variable is explained from the factors associated with the study. Here, it is observed that the communalities of all variables ranges from 0.387 to 0.637, i.e. the component extracted can explain the variables well.

Table 10 is showing the Eigen values (Vajda, Valko, & Turanyi, 1985) and the explained percentage of variance in the original variables by the extracted factors. Eigen values bigger than 1 is the criteria to select the factors (Cliff, 1988). From the obtained Eigen values of components, it is seen that first 6 components Eigen value is greater than 1. So, the first 6 components should be considered as 6 factors. Here, factor 1 to factor 6 are explaining 16.057%, 10.278%, 7.038%, 6.605%, 6.272%, and 6.104% of variation of data set, respectively. These 6 factors altogether can explain 52.354% of variation which is nearly 53% of total variation of the data set.

Table 10: Total variance explained

Fig 1: Scree plot of components.

A scree plot graph display the amount of variation explained by each component (Bro and Smilde, 2014). Fig 1 indicates that only 6 components Eigen values lie above 1 and rest of the components Eigen value lie below 1, suggesting considering only 6 components that explain the variation of data set to a specific extent.

Table 11: Rotated component matrix

From rotated component matrix in Table 11, first factor is highly correlated with three variables: technologies are effective in keeping the accurate data, performing the outmost of the customers benefit, and websites are good at collecting customers information. These three variables are included in factor 1. Second factor has high correlation with five variables: when the new one is dealing, I had a pleasant experience, customers opinion can influence website, risky when the products delivered may fail the expected time, getting access to the data without permission, and predict the customers. Third factor 3 has strong correlation with three variables: purchasing is solved effectively and satisfactorily, technology obstacles should not be a major concern, and capable of processing a large number of transactions. 

In the case of fourth factor, there is high correlation with two variables: risky when the products may be available at lower price and risky when the products have inferior quality. From the above table we can see that factor 5 is highly correlated with two variables: websites review customers opinions effectively, and websites encourage customers to send their feedback. The last factor, factor 6, has high correlation with two variables: receiving a timely response, and care, concern, and goodwill to their customers. These findings are presented at Table 12.

These factors are representing potential elements of customers trustworthiness and risk perception on e-commerce. The factor loading is showing that factor 1 is explaining the variation of the data set most. If the e-commerce companies focus on efficient technologies in keeping the accurate data, performing the outmost of the customers benefit and the websites that are good at collecting customers information then e-commerce will attain more trust. Factor 2 implies that the customers give importance to the pleasant experience, their personal opinion, and time duration to deliver product. From factor 3, it is obvious that effective and satisfactory purchasing, having no technology obstacles, and capacity of large number of transactions have similar type of effect on trust-worthiness and risk perception. 

According to factor 4, those who believe that products of lower price are risky, also believe that products of inferior quality are risky. Factor 5 suggests that reviewing customers opinion and encouraging the customers to send feed-back are almost equally important. The last factor is explaining that for betterment of e-commerce trust worthiness and risk perception, receiving timely response is necessary, and care, concern, and goodwill should be shown.

Table 12: Included variables in factors