CHAPTER V. RESULTS OF STUDY 2—AN EXPERIMENTAL STUDY OF
Phase 2: Structural equation modeling and results
The second part of the proposed model tests the relationships among an individual’s response to the fit review, response to the overall product information, and response to the online retailer (see Figure 3.2).
Correlations between the variables
The correlations between the three manifest variables and four latent variables are shown in Table 5.20. Attitude toward the retailer and overall confidence in purchase were highly correlated with site credibility (.57), overall confidence in purchase (.52), and future purchase intention (.63). Review evoked confidence in purchase and overall confidence in purchase were highly correlated (.52). Review credibility was moderately correlated with site credibility (.33) and attitude toward the review (.31). Future purchase intention was moderately correlated with site credibility (.37), overall confidence in purchase (.35), and review evoked confidence in purchase (.12). Review credibility was moderately correlated with review evoked confidence in purchase (.14), and site credibility was moderately correlated with overall confidence in purchase (.14). Attitude toward the review was moderately correlated with review evoked confidence in (.12) and site credibility (.15).
Table 5.20. Correlations Among Three Manifest Variables (1-3) and Four Latent Variables (4-7)
1 2 3 4 5 6 7
1. Review credibility 1 2. Review evoked confidence
in purchase .14* 1
3. Attitude toward the review .33*** .12* 1
4. Site credibility .31*** .07 .15** 1 5. Overall confidence in
purchase .11 .52*** .14* .47*** 1
6. Attitude toward the retailer .10 .11 .20*** .57*** .52*** 1
7. Future purchase intention .01 .14* -.01 .37*** .35*** .63*** 1 Note. Significant at p < .05*, p < .01**, p < .001***
Measurement model
A measurement model with three manifest variables and four latent variables (14 indicators) was performed to evaluate the quality of measures prior to testing the second part of the hypothesized model (H4-H14). The three manifest variables were review credibility, review evoked confidence in purchase, and attitude toward the review. Because those three variables were repeated measures, provided twice (once for positive fit review and once for negative), the summed scores for each review were added together to create a single manifest variable for each.
As presented in Table 5.21, standardized factor loadings for the 14 items to the four factors (i.e., site credibility, overall confidence in purchase, attitude toward the retailer, and future purchase intention) were higher than .50, ranging from .72 to .95, and were significant at p
< .001 level. Chi-square test and fit indices (CFI, RMSEA, TLI, and SRMR) show that the measurement model fits the data well (χ2 = 275.77, df = 101, p < .000, RMSEA = .07, CFI = .96, TLI = .95, SRMR = .03).
Table 5.21. Parameter Estimates for the Four Latent Variables and 14 Indicators in the Measurement Model
Structural model results and hypothesis testing
Figure 5.7. Structural Model Showing the Hypothesized Relationship between the Variables The results of the chi-square test and the fit indices show that the hypothesized model fits the data well (χ2 = 305.538, df = 108, p < .000, RMSEA = .07, CFI = .97, TLI = .96, SRMR = .08). Figure 5.7 shows the variables, paths, and goodness-of-fit measures used in testing the structural model. This procedure used maximum-likelihood estimation to test Hypotheses 4 to 14 of the proposed model. Six out of 11 structural paths in the SEM model were statistically
significant (p < .05).
The structural model supported six of the 11 hypotheses (4, 6, 7, 10, 12, and 14).
Hypothesis 4 predicted that review credibility is positively related to site credibility. The
standardized path coefficient between review credibility and site credibility was .32 (t = 7.36, p <
.001), providing support for Hypothesis 4. Hypothesis 5 posited that review credibility is positively related to overall confidence in purchase. The standardized path coefficient between
Review credibility
review credibility and overall confidence in purchase was -.02 (t = .39, p = .70), thus Hypothesis 5 was not supported. Hypothesis 6 predicted that review evoked confidence in purchase is positively related to site credibility. The standardized path coefficient between confidence in purchase decisions and site credibility was .11 (t = 2.13, p < .05), supporting Hypothesis 6.
Hypothesis 7 posited that review evoked confidence in purchase is positively related to overall confidence in purchase. The standardized path coefficient was .57 (t = 15.97, p < .001),
supporting Hypothesis 7. Hypotheses 8 and 9 predicted that attitude toward the review is positively related to site credibility (H8) and overall confidence in purchase (H9). The standardized path coefficients were not significant (attitude toward the review and site credibility: b = .04, t = .82, p = .42; attitude toward the review and overall confidence in purchase: b = .04, t = .85, p = .40), thus Hypothesis 8 and 9 were not supported.
Hypothesis 10 proposed that site credibility positively affects attitude toward the retailer.
The standardized path coefficient was .37 (t = 7.37, p < .001), supporting Hypothesis 10.
Hypothesis 11 predicted that site credibility is positively related to future purchase intention. The standardized path coefficient was not significant (b = .07, t = 1.29, p = .20), thus Hypothesis 11 was not supported. Hypothesis 12 posited that overall confidence in purchase has a positive influence on attitude toward the retailer. The standardized path coefficient was .23 (t = 4.35, p <
.001), supporting Hypothesis 12. Hypothesis 13 proposed that overall confidence in purchase is positively associated with future purchase decisions. The standardized path coefficient was not significant (b = .05, t = .94, p = .35), thus Hypothesis 13 was not supported. Hypothesis 14 predicted that attitude toward the retailer is positively related to future purchase decisions. The standardized path coefficient was .60 (t = 14.39, p < .001), supporting Hypothesis 14.
Covariate
This study used a particular shirt/blouse in the mock website, and participants’ attitude toward the product was measured to control for the effects of possible confounding variables. As shown in Table 5.22, the correlations between attitude toward the product, responses to the site (i.e., site credibility and overall confidence in purchase), and responses to retailer (i.e., attitude toward the retailer and future purchase intention) were statistically significant, ranging from .20 to .45. Attitude toward the product reflects an individual’s evaluative judgment of an object (Ajzen & Fishbein, 1980; Bagozzi et al., 1999). According to the IDM by Rogers (2003), a favorable or unfavorable attitude toward the innovation is the main outcome of the persuasion stage in the IDM model, which can be a significant determinant of whether to decide to adopt the product (Rogers, 2003). Furthermore, attitude toward the product plays an important role in determining an individual’s behavioral choices and intentions (Dabholkar, 1994). When an individual has a favorable attitude toward the product, she/he is likely to have positive attitude to the brand according to the empirical study of Mitchell (1986). Previous studies in Advertising also supported that positive beliefs in a product’s features had a positive influence on attitude to the brand (e.g., MacKenzie, et al., 1986). If an individual has a positive attitude toward the product, she/he is likely to have positive responses to the overall information provided by the retailer. Thus, attitude toward the product was used as a covariate variable to test post-hoc Hypotheses 4-7.
Post-hoc H4. Attitude toward the product will be positively related to site credibility.
Post-hoc H5. Attitude toward the product will be positively related to overall confidence in purchase.
Post-hoc H6. Attitude toward the product will be positively related to attitude toward the retailer.
Post-hoc H7. Attitude toward the product will be positively related to future purchase intention.
Table 5.22. Correlation Between Attitude Toward the Product and Four Dependent Variables
Attitude toward the product (shirt/blouse)
Site credibility .42***
Overall confidence in purchase .45***
Attitude toward the retailer .34***
Future purchase intention .20***
Note. ***p < .001, **p < .01, *p < .05 Post-hoc hypothesis testing
Figure 5.8. The hypothesized model (Phase 2) adding post-hoc hypotheses (effects of attitude toward the product on four dependent variables)
Review credibility
Using SEM analysis, post-hoc hypotheses 4-7 were tested by adding a covariate (attitude toward the product) based on the empirical evidence in the previous literature (see Figure 5.8).
Because the covariate had a significant effect on four dependent variables (see Table 5.22), the effect of attitude toward the product on four dependent variables were tested. Model fit indices and chi-square difference tests were performed to compare the two models with and without the covariate, and they proved to be significantly different (χ2diff = 141.57, dfdiff = 34, p < .001) (see Figure 5.8). Although there were changes in some standardized path coefficients after adding the covariate, the directions and significance of the paths were unchanged (see Table 5.22).
According to the results shown in Table 5.23, post-hoc H4 to H7 were supported, indicating that attitude toward the product was positively related to site credibility (b = .19, t = 3.71, p < .001), overall confidence in purchase (b = .11, t = 2.51, p < .001), attitude toward the retailer (b = .48, t
= 12.21, p < .001), and future purchase intention (b = .30, t = 5.95, p < .001).
Table 5.23. Summary of Casual Relationship Testing
Hypothesis Without covariate With covariate S. Est. (SE) t S. Est. (SE) t
H6 Review evoked confidence in purchase
à Site credibility
.11* (.05) 2.13 .11* (.05) 2.13
H7 Review evoked confidence in purchase decisions
à Overall confidence in purchase
.57*** (.04) 15.97 .57*** (.04) 15.97
H8 Attitude toward the review
à Site credibility .04 (.05) .82 .04 (.05) .82
H9 Attitude toward the review à Overall confidence in purchase
.04 (.04) .85 .04 (.04) .85
Table 5.23. Summary of Casual Relationship Testing (Continued)
H14 Attitude toward the retailer à Future purchase decisions
.60*** (.04) 14.39 .42*** (.06) 7.59 PH4 Attitude toward the product
à Site credibility
.19*** (.05) 3.71 PH5 Attitude toward the product
à Overall confidence in purchase
.11*** (.04) 2.57
PH6 Attitude toward the product
à Attitude toward the retailer .48*** (.04) 12.21
PH7 Attitude toward the product à Future purchase decisions
.30*** (.05) 5.95 Note. ***p < .001, **p < .01, *p < .05
Highlighted cells in light grey showed some changes after adding the covariate in the hypothesized model. PH = Post-hoc hypothesis.
Ad-hoc mediation testing
Mediation effects were tested to enhance understanding about the results. Preacher and Hayes’ (2008) bootstrap procedure was conducted in Mplus to examine the extent to which responses to the attitude toward the retailer mediates the effects of responses to the overall product information on future purchase decisions. In addition, other possible mediation effects were examined.
As shown in Table 5.24, the results indicated that there were indirect effects for site credibility (.20) and pvera;; confidence in purchase (.09) on future purchase intention through attitude toward the retailer. Attitude toward the product was indirectly related with future purchase decisions through attitude toward the retailer (.17) and through site credibility and
attitude toward the retailer (.02). There was an indirect effect of attitude toward the product on attitude toward the retailer through site credibility (.04).
Table 5.24. Results of Ad-hoc Mediation Effects
Predictors Mediators Dependent Site credibility Attitude toward the
retailer
Note. Standardized path coefficients are reported with t-values in parentheses.
*** p < .001. ** p < .01, * p< .05