As mentioned earlier, in the first part of the model (focused on switching costs’ antecedents) we hypothesize that:
H1: The 5Ps are positively related to perceived positive switching costs
Following this approach, we suppose (in the second part of the model, focused on the consequences of the customer perception of positive switching costs) that:
H2: Perceived positive switching cost are positively associated with customer satisfaction, repurchase intentions, cognitive and behavioral loyalty
H3a: Customer satisfaction and cognitive loyalty (in a digital environment) are positively associated with cognitive lock-in strategies
As well as:
H3b: Repurchase intentions and behavioral loyalty (in a digital environment) are positively associated with behavioral lock-in strategies
Our empirical analysis followed two steps: in the first part standard scale development procedures were followed in the development of the multidimensional switching costs scale. Scale items for the 5Ps were developed based on the guidelines suggested by Churchill (1979) and Gerbing and Anderson (1988). But also in-depth interviews with managers from a sample of 15 firms from IT (B2B) sector (three e-suppliers and 12 of their e-customers [that had experienced shopping online experiences with all the three esuppliers]) were conducted to define the scale items. A panel of five marketing faculty reviewed the items for clarity and face validity. Then item-total correlation, Cronbach’s Alpha and exploratory factor analysis were examined: thus the original items were refined and items not meaningfully were deleted. At this step (concluded the exploratory analysis) we were ready to administer the questionnaire to a sample of 180 e-customers [that had experienced shopping online from at least two of the original 3 e-suppliers]. The answering rate was quite high (about 86%). Our analysis followed mainly the structural equation modeling procedure.
The hypotheses were tested using multiple multivariate analysis methodologies (we used SPSS 11.0 and LISREL 8.54). As mentioned earlier, firstly we conducted an exploratory factor analysis to determine whether the scale items loaded as expected. We then calculated Cronbach’s alphas for the scale items to ensure that they exhibited satisfactory levels of internal consistency (Table 16-1). We refined the scales by deleting items that did not load meaningfully on the underlying construct and those that did not highly correlate with other items measuring the same construct.
|
PSC |
P1 |
P2 |
P3 |
P4 |
P5 |
|
|---|---|---|---|---|---|---|
|
Positive Perceived Switching Costs (PSC) |
0.93 | |||||
|
Perceived Usefulness (P1) |
0.67 |
0.84 | ||||
|
Perceived Ease of Use (P2) |
0.62 |
0.69 |
0.79 | |||
|
Perceived Simplicity of Web site Design (P3) |
0.73 |
0.65 |
0.61 |
0.87 | ||
|
Best Perceived Customer Service (P4) |
0.79 |
0.72 |
0.67 |
0.75 |
0.91 | |
|
Positive Perceived Feeling with the Web site (P5) |
0.79 |
0.63 |
0.72 |
0.84 |
0.62 |
0.85 |
|
*Cronbach’s Alpha for the constructs are given in red colour in the diagonal elements. |
Hypothesis H1 refers to the impact of the 5Ps on the customer perceived feeling of positive switching costs (during his online shopping expeditions), stemming from the Web site’s elements. To parsimoniously capture the joint impact of the 5Ps, we used the following multiplicative model:
PSC = g0 * P1g1 * P2g2 * P3g3 * P4g4 * P5g5 (1)
Where: PSC = customer perceived feeling of positive switching costs; P1 = perceived usefulness; P2 = perceived ease of use; P3 = perceived simplicity of Web site interface design; P4 = best perceived customer service; P5 = positive perceived feeling with the Web site. Equation (1) was linearized by logarithmic transformation and the log-transformed model was estimated using ordinary least squares (OLS). Correlations between the variables are reported in Table 16-1.
Parameter g0 is the intercept, and parameters g1 to g5 capture the impact of the 5Ps on PSC Results of the regression analysis are presented in Table 16-2. A summary consideration of the results indicates that all the parameters estimated are significant at p < .05, and in the predicted direction. The adjusted R2 of the model is 0.73. We calculated the variance inflation factor to check for multicollinearity: the average VIF is 2.66, ranging from 2.11 and 3.21, well below the recommended cutoff of 10 (Neter, Wasserman & Cutner, 1985). Thus our hypothesis (H1) that the 5Ps are positively related to customer feeling of positive switching costs stemming from Web site’s elements is supported. The elasticity of PSC respect to the 5Ps ranges from 0.07 for perceived ease of use to 0.42 for positive perceived feeling with the Web site.
|
Independent Variables |
Parameter Estimate |
Standard Error |
T Value |
|---|---|---|---|
|
Constant |
0.3984 |
0.0723 |
4.985 |
|
Perceived Usefulness (P1) |
0.1547 |
0.0406 |
4.321 |
|
Perceived Ease of Use (P2) |
0.0731 |
0.0324 |
3.214 |
|
Perceived Simplicity of Web site Design (P3) |
0.1532 |
0.0256 |
5.332 |
|
Best Perceived Customer Service (P4) |
0.1634 |
0.0452 |
6.845 |
|
Positive Perceived Feeling with the Web site (P5) |
0.4187 |
0.0321 |
7.053 |
|
*t value is significant at <.05 confidence level. |
Hypotheses H2 focuses on the consequences of customer-perceived positive switching costs. To test the impact of PSC on customer satisfaction (= CS), repurchase intentions (= RI), cognitive loyalty (CL) and behavioral loyalty (BL), the hypotheses were tested by the use of structural equation modeling (SEM, this time using LISREL 8.54); the path diagram, the separate structural equations and regression analyses with the latent variable scores will be reported. The collected evidence from all these analyses will be used for the conclusion regarding the stated hypothesis. The path diagram (Figure 16-2) shows that PSC have a positive direct effect, as hypothesized, on all the considered four dimensions. All estimates have t-values larger than three and are thus statistically significant. The structural equations (reduced models) as well as the multiple regression analysis (MRA) confirm this result, by generating a high R2 (0.87).
Figure 16-2: Relationship between customer perceived feeling of positive switching costs and customer satisfaction (CS), repurchase intentions (RI), cognitive loyalty (CI), behavioral loyalty (BI)
Although c2 measure should show a lower value (c2 = 12.52, df = 172, p < 0.005), since c2 is very sensitive to sample size, a large number of other indices indicate a good fit of the model. Table 16-3 shows some of those indices along with reported guidelines for a good model.
|
Fit Index |
Guidelines |
Model values |
|---|---|---|
|
c2 |
12.52, df = 172, p < 0.005 |
|
|
CMIN/DF |
2<CMIN/DF<3 |
2.563 |
|
NFI |
NFI>0.9 |
0.975 |
|
TLI |
TLI>0.9 |
0.975 |
|
GFI |
GFI>0.9 |
0.947 |
|
AGFI |
AGFI>0.9 |
0.911 |
|
Delta 2 |
Delta 2>0.9 |
0.982 |
Hypotheses H3a and H3b focus instead on the possible correlation between CS, RI, CL, BL and cognitive and behavioural strategies in a digital environment. In order to test their possible correlation, the hypotheses were also tested in this case by the use of SEM. The path diagram, the separate structural equations and regression analyses with the latent variable scores are reported below. The path diagram (Figure 16-3) shows, as hypothesized, that customer satisfaction and cognitive loyalty are correlated higher with cognitive lock-in strategies (than with behavioral lock-in), whereas repurchase intentions and behavioral loyalty better correlate with behavioral lock-in strategies (than with cognitive ones). All estimates have t-values larger than three and are thus statistically significant. The structural equations (reduced models) as well as the multiple regression analysis (MRA) confirm this result, by generating a high R2 (0.76).
Figure 16-3: Relationship between customer satisfaction (CS), repurchase intentions (RI), cognitive loyalty (CI), behavioral loyalty (BI) and cognitive (CLC)/behavioral (BLC) lock-in strategies (For simplicity we will show only the significant values [values <0.40 are omitted]. We report in large dashed arrows and irregular dashed arrows the significant pathways.)
Also in this case c2 measure should show a lower value (c2 = 14.73, df = 165, p < 0.005). But a large number of other indices indicate a good fit of the model (Table 16-4 shows some of those indices along with reported guidelines for good model.).
|
Fit Index |
Guidelines |
Model values |
|---|---|---|
|
c2 |
14.73, df = 165, p < 0.005 |
|
|
CMIN/DF |
2<CMIN/DF<3 | |
|
NFI |
NFI>0.9 | |
|
TLI |
TLI>0.9 | |
|
GFI |
GFI>0.9 | |
|
AGFI |
AGFI>0.9 | |
|
Delta 2 |
Delta 2>0.9 |