The use of Survey Monkey allowed for the participation of a wide range of consumers from different backgrounds. Table 2, below, illustrates the participants’ demographic characteristics.
Respondents banking profile Tables 3a and 3b summarise respondents’ banking profile.
Scale Measurement Construct reliability
Using the Cronbach’s alpha (a) coefficient, the measuring instruments were tested for internal consistency and construct reliability. Item selection and scale purification using inter-item and item-to-total correlations were used to measure internal consistency for questions. With the exception of Section D (demographic information), all other scales (see Table 4) were tested for reliability. According to Tan and Teo, perception scales yielding a Cronbach’s alpha of at least 0.6 are regarded as reliable and internally consistent.
Cronbach’s alphas of the sub-scales ranged from 0.690 (Relative Advantage) to 0.925 (Self-efficacy) (Table 4), which indicate an acceptable internal consistency and reliability measures for the questionnaire. Factor analysis using the varimax rotation was used to ascertain construct validity of each of the constructs (sub-scales) with multiple-item measures. The 17 retained-items (Table 4) yielded a five-factor structure with eigenvalues greater than 1, explaining 73 per cent of the variance. As in Brown et al., risk and complexity loaded onto Factor 1. A correlation analysis revealed that risk and complexity were separate constructs, as the coefficient between them was 0.427 (p < .05). Self-efficacy loaded onto Factor 2. Factor 3 contained items from Trialability and Compatibility. Factor 4 also had cross loading with items from Facilitating Conditions and Compatibility loading on this factor. Relative advantage loaded onto Factor 5. After dropping, one item due to cross loadings, a subsequent factor analysis yielded yet another 5-factor structure accounting for 74 per cent of the variance. Overall these findings suggest that these factors were indeed separate constructs. Table 5 illustrates these constructs as well as their means, standard deviations, and correlations.
Tests of model and hypotheses
To test the model and hypotheses, the authors utilised multiple regression and logistic regression analysis. A series of direct logistic regression was performed to assess the impact of independent variables on the dependent variable (adopting/ rejecting CB). The model contained 7 independent variables (risk, complexity, self-efficacy, trialability, compatibility, facilitating conditions, and relative advantage) identified the factor analysis and depicted in Table 5. The binary regression model’s goodness of fit model was significant (x2 = 30.392; p < .0005), indicating that the model was able to distinguish between respondents who reported to have adopted CB from those who did not. The model explained between 24.9 per cent (Cox & Snell R squared) and 33.5 per cent (Nagelkerke R squared) of the variance, with 73.6 per cent classification accuracy. This indicates that approximately three-quarters of the sample were correctly classified as either using CB or not. Of the seven predictor variables, only two made unique statistically significant contribution to the model: relative advantage (B = 1.004; Wald = 11.203; p < .0001) and complexity.
To further test these predictors, we analysed results using an independent-samples t-test to determine whether the means for each of the predictor variables identified in the model differed significantly between the trial and non-trial groups. The mean scores for relative advantage were 3.9151 and 3.0758 for CB users and non-CB users, respectively (t-value = -5.104, p < .0001). Mean scores for complexity were 1.9717 and 2.6288 for CB users and non-CB users, respectively (t-value = 3.542; p < .001).
The hypotheses formulated above were tested using multiple regression and logistic regression. There was support for two of the seven hypotheses, as follows:
H1: the greater the perceived complexity of using CB, the less likely that it will be adopted.
H3: the greater the perceived relative advantage of using CB, the more likely that it will be adopted.
Past research illustrates that compatibility, relative advantage, and complexity are consistently related to adoption.
Table 3 a. Types of Accounts used
Table 3b. Extent of channel use (1 = never, 6 = daily)
|Cell phone Banking||4.1||1||6||1.93|
Table 4. Cronbach’s alpha (a) for the sub-scales
|Sub-scale||Cronbach’s alpha (a)||Mean||No. of items retained|
Table 5. Means, standard deviations, and correlations among variables