Chapter IV Data Analysis
4.5 Regression Analysis
tolerance” (.445). The lowest correlation coefficient occurs between “product satisfaction” and “recommendation” (.443) (See Table 4-4-7).
Table 4-4-7: Hierarchical-Stepwise Correlation Analysis of Customer Satisfaction and Loyalty
Customer Loyalty Customer
Satisfaction
Recommendation Repurchase Price Tolerance Product will further predict the influence of independent variables on dependent variables by the calculation of regression in the section. Stepwise analysis on the variables in first order and second order will be conducted. Specifically, the influence of brand image on perceived value, brand image on customer satisfaction, brand image on customer loyalty, perceived value on customer satisfaction, perceived value on customer loyalty, and the influence of customer satisfaction on customer loyalty will be tested and examined one after another.
A regression model can be conceived as a prediction on a response for a given set of predictor variables (Jain, 2008). To determine the value of parameters for a function which best fits a set of observed data is the goal of regression analysis. To investigate
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
83
how response／dependent variables is influenced by predictor/independent variables in a linear function, linear regression is applied. When there is only one predictor／
independent variable, we use simple linear regression for estimating the variance of the dependent variable. If there are more than one predictor/independent variable, multiple linear regression is applied.
Before building the regression model, it is reasonable to examine all the variation of the possible outcome from the systematic sources; however, it is not realistic for practice for the sampling is usually not census. Thus, given the model structured by only some parameters, it is necessary to test the goodness beforehand, for its main feature is to evaluate the agreement between the model predictions and the observed data (Hagquist and Stenbeck, 1998). For the reason, 𝑅2 is chosen for the goodness of fit test in the study since it is considered as a summary measure for the assessment of agreement. The value of 𝑅2 should be higher than 0.1 for achieving the explanatory power of regression model. To test the significance of regression model, F test is applied. Specifically, to achieve the significance, the observed p value, the corresponding value of F value, should be less than .05.
When predictor variables (independent variables) in the regression model are more highly correlated with other predictor variables than with the dependent variable, multi-collinearity in occurs, which causes a problem for differentiating the effects of each predictor variables on the dependent variables. To avoid it, the correlation coefficient is higher than 0.8 should be singled out.
Other measures for assessing the multi-collinearity are tolerance and variance inflation factor (VIF). Tolerance value equals to 1-𝑅2 (𝑅2 means the variance of
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
84
independent variable explained by other variables), which ranges to 0 to 1. The larger the tolerance value is, the less likely the multi-collinearity occurs. On the contrary, since VIF is the reciprocal of tolerance, the lesser the VIF is, the less likely the multi-collinearity occurs. Usually, VIF statistic should be less than 10.
To ensure the error terms are random and the residual of the regression model is independent to each other, Durbin-Watson test is applied. If and D-W statistic ranges 1.5 to 2.5, there will not be a problem of autocorrelation.
1. First-Order Regression Analysis
Regression Analysis of Brand Image, Perceived Value on Customer Satisfaction and Customer Loyalty: Based on the research structure (See Table 3-1-1), the study means to examine the effects of brand image and perceived value on customer satisfaction and loyalty. Two regression models will be built by testing the effects of brand image and perceived value on customer satisfaction and customer loyalty. The other is about the effect of all the independent variables including brand image, perceived value, and customer satisfaction on customer loyalty.
The results of regression analysis are shown as Table 4-5-1: Brand image and perceived value are significantly associated with customer satisfaction (Adjucted 𝑅2= .795, F= 901.466, p=.000, VIF=2.563<10, D-W= 2.214). Brand image and perceived value are significantly associated with customer loyalty (Adjusted 𝑅2=.639, F= 411.825, p=.000, VIF= 2.563<10, D-W= 2.247). To sum up, both brand image and perceived value are significantly associated with customer satisfaction and customer loyalty, which are related in a positive linear sense.
The regression equations are as follows:
Customer Satisfaction= 13.357+ 0.769*Brand Image+0.151*Perceived Value
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
85
Customer Loyalty= 2.325+0.130*Brand Image+0.695*Perceived Value
Table 4-5-1: First-Order Regression of Brand Image and Perceived Value on Customer Satisfaction and Customer Loyalty
Predictor Variables Dependent Variables
Customer Satisfaction Customer Loyalty
Brand Image .769*** .130***
Perceived Value .151*** .695***
（Constant） 13.357 2.325
R 2 .796 .641
Adjusted R 2 .795 .639
F 901.466 411.825
p .000 .000
VIF 2.563 2.563
D-W 2.214 2.247
p＜.05*, p＜.01**, p＜.001***
Regression Analysis of Brand Image, Perceived Value, Customer Satisfaction on Customer Loyalty: According to results of Table 4-5-2, brand image, perceived value and customer satisfaction are significantly associated with customer loyalty (Adjusted 𝑅2=.652, F= 290.647, p=.000, VIF= 2.675~5.462<10, D-W= 2.463). Hence, brand image, perceived value and customer satisfaction are significantly associated with and customer loyalty. The following equation reveals that brand image and perceived value are associated with customer loyalty in a positive linear sense; whereas customer satisfaction is associated with customer loyalty in a negative sense.
The regression equation is as follows:
Customer Loyalty= 4.250+0.169*Brand Image+0.893*Perceived Value -0.257*Customer Satisfaction
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
86
Table 4-5-2: First-Order Regression of Brand Image, Perceived Value and Customer Satisfaction on Customer Loyalty
Predictor Variables Dependent Variables
Customer Loyalty
Brand Image .169***
Perceived Value .893***
Customer Satisfaction -.257***
（Constant） 4.250
R 2 .654
Adjusted R 2 .652
F 290.647
p .000
VIF 2.675~5.462
D-W 2.463
p＜.05*, p＜.01**, p＜.001***
2. Second-Order Regression Analysis
(1) Regression of Brand Image on Perceived Value H1: Brand image has a positive impact on perceived value
Regression models as a tool to examine the adequacy of each hypothesis in research will be applied in the study for testing the effects of independent variables on dependent ones. H1 will be tested with two constructs (SCTI and HRI) of brand image as independent variables and four constructs (Social Value, Quality, Price Value, Emotional Value) of perceived value as dependent variables via multiple regression analysis. The results are shown as follows:
According to Table 4-5-3, “SCTI” of brand image is significantly and positively associated with perceived value, which includes “Social Value,” “Quality,” “Price
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
87
Value,” and “Emotional Value” four constructs. Specifically, “SCTI” is significantly associated with “Social Value” (β= .740, p=.000), “Quality” (β= .216, p=.000), “Price Value” (β= .329, p=.000), and “Emotional Value” (β= .293, p=.000) in a positive sense.
“HRI,” the other construct of brand image, is significantly and positively associated with perceived value, too. Specifically, “HRI” is significantly associated with “Social Value” (β= .304, p=.000), “Quality” (β= .575, p=.000), “Price Value” (β= .183, p=.000), and “Emotional Value” (β= .273, p=.000) in a positive sense.
The regression equations are as follows:
Social Value= -9.684+0.74*SCTI+0.304*HRI Quality= 4.014+0.216*SCTI+0.575*HRI Price Value= 5.383+0.329*SCTI+0.183*HRI Emotional Value= 3.497+0.293*SCTI+0.273*HRI
The results of regression analysis indicate that H1 is supported.
Table 4-5-3: Second-Order Regression of Brand Image on Perceived Value
Predictor Variables Perceived Value
Social Value Quality Price Value Emotional Value Brand
Image
SCTI .740*** .216*** .329*** .293***
HRI .304*** .575*** .183*** .273***
（Constant） -9.684 4.014 5.853 3.497
R 2 .848 .492 .197 .234
Adjusted R 2 .847 .490 .194 .230
F 1285.501 223.808 56.723 70.488
p .000 .000 .000 .000
VIF 1.270 1.270 1.270 1.270
D-W 2.049 2.377 1.972 2.439
p＜.05*, p＜.01**, p＜.001***
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
88
(2) Regression of Brand Image and Perceived Value on Customer Satisfaction
H2: Brand image has a positive impact on customer satisfaction
To verify H2, the two constructs (SCTI and HRI) of brand image as independent variables and the two constructs (product satisfaction and service satisfaction) of customer satisfaction are subjected to regression analysis.
According to Table 4-5-4, “SCTI” of brand image is significantly associated with customer satisfaction, which includes “product satisfaction” and “service satisfaction”
two constructs. Specifically, “SCTI” is significantly associated with “product satisfaction” (β= .171, p=.000) and “product satisfaction” (β= .469, p=.000) in a positive sense. “HRI,” the other construct of brand image, is significantly and positively associated with customer satisfaction. Specifically, “HRI” is significantly associated with “product satisfaction” (β= .646, p=.000) and “service satisfaction”
(β= .280, p=.000) in a positive sense.
The regression equations are as follows:
Product Satisfaction= 2.525+0.171*SCTI+0.646*HRI Service Satisfaction= 8.197+0.469*SCTI+0.280*HRI
The results of regression analysis imply that H2 is supported.
Table 4-5-4: Second-Order Regression of Brand Image on Customer Satisfaction Predictor
Variables
Customer Satisfaction Product Satisfaction
Model 1
Service Satisfaction Model 2
Brand SCTI .171*** .469***
Image HRI .646*** .280***
（Constant） 2.525 8.197
R 2 .548 .420
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
89
Adjusted R 2 .546 .417
F 280.216 166.947
p .000 .000
VIF 1.270 1.270
D-W 2.147 2.360
p＜.05*, p＜.01**, p＜.001***
H4: Perceived value has a positive impact on customer satisfaction
The four constructs (Social Value, Quality, Price Value, and Emotional Value) of perceived value as independent variables and the two constructs (product satisfaction and service satisfaction) of customer satisfaction are subjected to regression analysis for verifying H4.
The HRI of each construct of perceived value and customer satisfaction is revealed by the analysis results shown in Table 4-5-5. Accordingly, “Social Value” is significantly associated with “product satisfaction” (β= .118, p=.000) and “service satisfaction” (β= .406, p=.000) in a positive sense. “Quality” is significantly associated with “product satisfaction” (β= .496, p=.000) in a positive sense; while it is associated with “service satisfaction” (β= -.040, p=.357) in a negative sense. “Price Value” is significantly associated with “service satisfaction” (β= .374, p=.000) in a positive sense, but associated with “product satisfaction” (β= -.024, p=.295) in a negative sense. There is a significantly positive HRI between “Emotional Value” and both “product satisfaction” (β= .478, p=.000) and “service satisfaction” (β= .133, p=.000).
The regression equations are as follows:
Product Satisfaction= 10.148+0.118*Social Value+0.496*Quality-0.024*Price Value Service Satisfaction=12.406+0.406*Social Value-0.04*Quality+0.374*Price Value +0.133*Emotional Value
‧
By the observation of the above equations and the analysis results, H4 is partially supported.
Table 4-5-5: Second-Order Regression of Perceived Value on Customer Satisfaction Predictor
VIF 1.620~1.844 1.620~1.844
D-W 2.123 2.454
p＜.05*, p＜.01**, p＜.001***
(3) Regression of Brand Image and Perceived Value on Customer Loyalty H3: Brand image has a positive impact on customer loyalty
To verify H3, the two constructs (SCTI and HRI) of brand image as independent variables and the three constructs (recommendation, repurchase, and price tolerance) of customer loyalty are subjected to regression analysis.
The results are shown in the below table (See Table 4-5-6). Accordingly, “SCTI” is significantly and positively associated with each customer loyalty. In details, “SCTI” is significantly associated with “recommendation,” (β= .748, p=.000) “repurchase,”
(β= .267, p=.000) and “price tolerance” (β= .344, p=.000) in a positive sense. By contrast, “HRI” is only significantly associated with “repurchase” (β= .460, p=.000),
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
91
but its HRI with “recommendation” (β= .047, p=.156) and “price tolerance” (β= .082, p=.089) does not achieve significance.
The regression equations are as follows:
Recommendation= 5.149+0.748*SCTI+0.047*HRI Repurchase= 2.703+0.267*SCTI+0.460*HRI Price Tolerance= 4.459+0.344*SCTI+0.082*HRI
The results and regression equations indicate that H3 is partially supported.
Table 4-5-6: Second-Order Regression of Brand Image on Customer Loyalty
Predictor Variables Customer Loyalty
Recommendation Model 1
Repurchase Model 2
Price Tolerance Model 3
Brand Image SCTI .748*** .267*** .344***
HRI .047 .460*** .082
（Constant） 5.149 2.703 4.459
R 2 .594 .395 .152
Adjusted R 2 .592 .393 .148
F 337.774 151.087 41.250
p .000 .000 .000
VIF 1.270 1.270 1.270
D-W 1.946 2.604 2.231
p＜.05*, p＜.01**, p＜.001***
H5: Perceived value has a positive impact on customer loyalty
The four constructs (Social Value, Quality, Price Value, and Emotional Value) of perceived value as independent variables and the three constructs (recommendation, repurchase, and price tolerance) of customer loyalty are subjected to regression analysis for verifying H4.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
92
Table 4-5-7 depicts that constructs of perceived value is partially associated with constructs of customer loyalty. In details, “Social Value” is significantly and positively associated with “recommendation,” (β= .531, p=.000) and “price tolerance” (β= .173, p=.000), but its HRI with “repurchase,” (β= .013, p=.726) does not achieve significance. “Quality” is significantly associated with “recommendation,” (β= .160, p=.000) “repurchase,” (β= .628, p=.000) in a positive sense, but negatively associated with “price tolerance” (β= -.094, p=.006). “Price Value” is overall associated with each construct of customer loyalty in a positive sense. Specifically, it is significantly associated with “recommendation”, (β= .328, p=.000) “repurchase,” (β= .085, p=.022) and “price tolerance” (β= .889, p=.000). “Emotional Value” is overall significantly associated with customer loyalty; while it is associated with “repurchase” (β= .158, p=.000) in a positive sense, it is negatively associated with “recommendation” (β=
-.147, p=.000) and “price tolerance” (β= -.209, p=.000).
The regression equations are as follows:
Recommendation= 3.535+0.531*Social Value+0.16*Quality
+0.328*Price Value-0.147*Emotional Value Repurchase= 1.409+0.013*Social Value+0.628*Quality
+0.085*Price Value+0.158*Emotional Value Price Tolerance= 1.318+0.173*Social Value-0.094*Quality
+0.889*Price Value-0.209*Emotional Value.
By observation of the results and the equations above, H5 is partially supported.
‧
Table 4-5-7: Second-Order Regression of Perceived Value on Customer Loyalty
Predictor Customer Loyalty
Recommendation
VIF 1.620~1.844 1.620~1.844 1.620~1.844
D-W 1.546 2.478 2.186
p＜.05*, p＜.01**,p＜.001 ***
(4) Regression of Customer Satisfaction on Customer Loyalty H6: Customer satisfaction has a positive impact on customer loyalty
To verify H6, the two constructs (product satisfaction and service satisfaction) of customer satisfaction as independent variables and the three constructs (recommendation, repurchase, and price tolerance) of customer loyalty are subjected to regression analysis.
Table 4-5-8 below indicates that customer satisfaction is significantly associated with customer loyalty in each construct in a positive sense. In details, “product satisfaction”
is significantly and positively associated with “recommendation,” (β= .273, p=.000)
“repurchase,” (β= .697, p=.000) and “price tolerance” (β= .152, p=.000). “Service satisfaction” is significantly associated with “recommendation,” (β= .311, p=.000)
“repurchase” (β= .088, p=.000), and “price tolerance” (β= .362, p=.000) in a positive sense as well.
‧
The regression equations are as follows:
Recommendation= 2.982+0.273*Product Satisfaction +0.311*Service Satisfaction
Repurchase= 0.078+0.697*Product Satisfaction+0.088*Service Satisfaction Price Tolerance= 0.710+0.152*Product Satisfaction
+0.362*Service Satisfaction
From the above analysis, H6 is supported.
Table 4-5-8: Second-Order Regression of Customer Satisfaction on Customer Loyalty
Predictor Customer Loyalty
Recommendation
(5) Multiple Regression of Brand Image, Perceived Value, and Customer Satisfaction on Customer Loyalty
To examine the whole research structure by testing the HRI of each variable with multiple regressions, the independent variables are drawn from two construct of brand
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
95
image, the four constructs of perceived value, and the two constructs of customer satisfaction, and the dependent variable is drawn from the three constructs of customer loyalty.
The results of analysis are shown in Table 4-5-9. In terms of the construct of brand image, “SCTI” is significantly and positively associated with “recommendation”
(β= .690, p=.000) and “repurchase” (β= .464, p=.000) ; while it is associated with
“price tolerance” (β= .015, p=.790), it does not achieve significance. “HRI” is only significantly and positively associated with “repurchase” (β= .111, p=.022), but its HRI with “recommendation” (β= .003, p=.945) and “price tolerance” (β= -.027, p=.551) does not achieve significance. "Social Value,” one of the constructs of perceived value, is significantly associated with “repurchase” (β= -.502, p=.000) and “price tolerance”
(β= .228, p=.001), but not significantly associated with “recommendation” (β= .024, p=.700). “Quality” is significantly and positively associated with “recommendation”
(β= .247, p=.000) and “repurchase” (β= .363, p=.000), but significantly associated with
“price tolerance” in a negative sense (β= -.110, p=.019). “Price Value” is significantly and positively associated with “recommendation,” (β= .442, p=.000) “repurchase,”
(β= .135, p=.001) and “price tolerance” (β= .945, p=.000). “Emotional Value” is only significantly associated with “price tolerance” (β= -.212, p=.000) but in a negative sense. It is also negatively associated with “recommendation” (β= -.064, p=.146) and
“repurchase” (β= -.073, p=.144). As for the constructs of customer satisfaction,
“product satisfaction” is positively associated with “repurchase” (β= .449, p=.000) and
“price tolerance” (β= .048, p=.496) ; whereas it is associated with “recommendation”
(β= -.155, p=.022) in a negative sense. “Service satisfaction” is associated with constructs of customer loyalty in a negative sense. Specifically, it is associated with
‧
tolerance” (β= -.155, p=.000).The regression equation are as follows:
Recommendation= 4.245+0.69 “SCTI”+0.03 “HRI”
+0.024 “Social Value”+0.247 “Quality”
+0.442 “Price Value”-0.064 “Emotional Value”
-0.155 “Product Satisfaction”
-0.243 “Service Satisfaction”
Repurchase= -4.099+0.464 “SCTI”+0.111 “HRI”
-0.502 “Social Value”+0.363 “Quality”
+0.135 “Price Value”-0.073 “Emotional Value”
+0.449 “Product Satisfaction”
-0.008 “Service Satisfaction”
Price Tolerance= 2.751+0.015 “SCTI”-0.027 “HRI”
+0.228 “Social Value”-0.11 “Quality”
+0.945 “Price Value”-0.212 “Emotional Value”
+0.048 “Product Satisfaction”
-0.155 “Service Satisfaction”
Table 4-5-9: Second-Order Multiple Regression of Brand Image and Perceived Value on Customer Loyalty
Predictor Variables Customer Loyalty
Recommendation
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
97
Emotional Value -.064 -.073 -.212***
Customer Product Satisfaction -.155* .449*** .048
Satisfaction Service Satisfaction -.243*** -.008 -.155***
（Constant） 4.245 -4.099 2.751
R 2 .750 .680 .726
Adjusted R 2 .746 .678 .721
F 171.313 120.991 151.001
p .000 .000 .000
VIF 2.169~8.314 2.169~8.314 2.169~8.314
D-W 1.342 2.216 2.241
p＜.05*, p＜.01**,p＜.001 ***