5. Results and discussion
5.8. SEM results for Model 3 – Combined research model
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
was found to have positive effect on all aspects of CBE and a direct positive effect on brand loyalty, what is consistent with prior studies (e.g. Algesheimer et al., 2005; Jahn & Kunz, 2012;
Sprott et al., 2009). While a study by Leckie et al. (2016, p.571) found a positive link between self-expressive brand, cognitive processing, and affection the authors indicate that ‘consumers do care about their self-image and how congruent this is with the brand image of the mobile phone service provider.’
The differences in the research findings can be due to the poor model fit or different mobile market characteristics, consumer behavior in a diverse culture setting, or cognitive processes in a different language context. While the nature of those differences requires further attention, the significance of this research’s findings should be considered with caution.
5.8. SEM results for Model 3 – Combined research model
A structural equation modeling method was also used to test the combined research model.
Table 12 presents the model fit obtained from the structural model. Similarly to structural model 1 and 2, GFI and AGFI have not reached the desired .95, thus indicating poor model fit.
Table 12. GOF indices for Model 3.
GOF statistics df p-value CMIN/DF GFI AGFI RMSEA NFI TLI CFI
Results 1358.956 623 <.001 2.181 0.825 0.792 0.059 0.895 0.932 0.940
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
The SEM results are shown in Table 13. The results support the H1bc as there is a positive
relationship between satisfaction and CBE dimensions affection (β=.517, t=7.830) and activation (β=.-.390, t=-3.516). H1a is rejected as there is no link between satisfaction and
cognitive processing. H2a,b,c are supported as there is a positive relationship between trust and all three CBE dimensions, including cognitive processing (β=.310, t=2.775), affection (β=.378, t=5.304) and activation (β=.764, t=6.092), respectively. Commitment has a link with cognitive
processing (β=.152, t=4.699) and activation (β=.386, t=1.283) in support of H3a,c, however H3b
is rejected due to the lack of relationship between commitment and affection (β=.060, t=2.38).
Involvement has a positive relationship with only one dimension of CBE- Activation (β=-.160, t=-2.762). No link has been noted between involvement and cognitive processing or affection,
therefore H4a,b are rejected. Participation resulted in positive association with cognitive processing (β=.171, t=3.061), however it failed to prove any positive link to the other two CBE
dimensions- affection (β=.002, t=.061) and activation (β=.063, t=1.107). Finally, results have revealed positive association between self-expressive brand and affection in support of H6b, with the brand loyalty. Finally, results support H6a,c as there is a positive relationship between self-expressive brand and CBE factors, positive processing (β=.242, t=3.750) and activation (β=-.159, t=-2.341). In terms of the direct relationship of antecedents on brand loyalty, the result link positively satisfaction (β=.262, t=2.878), commitment (β=.502, t=7.835), and trust (β=.234,
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
t=2.226) with brand loyalty, thus supporting H8a,b,c. suggesting that CBE only partially mediates
the relationship between them and brand loyalty. H8e is supported only for the one-tail test (β=.153, t=2.909), indicating involvement might have a positive relationship with brand loyalty.
Hypotheses H8d,f are rejected as involvement (β=-.016, t=-.411) and self-expressive brand (β=-.054, t=-1.137) show no direct relationship with brand loyalty suggesting that CBE
completely mediates the relationship between them and brand loyalty. Regarding the relationship between the CBE dimensions and brand loyalty, results support H7c pointing to the positive influence of activation (β=268, t=5.594), while rejecting H7a,b exposing no such
influence between brand loyalty and cognitive processing (β=-.087, t=-1.868), and affection (β=-.053, t=-.566).
The combined research model portrayed varied results between antecedents, CBE dimensions and brand loyalty, in comparison with separated Model 1 and Model 2. While the links between relationship quality factors and CBE dimension and brand loyalty were by large consistent with results obtained in Model 1 (except this time results showed a negative link between commitment and affection), the level of support was much lower between involvement, participation, self-expressive brand and CBE dimensions and brand loyalty.
‧
After the analysis of the results of the three models, it is reasonable to conclude that some of the antecedents, for instance involvement, participation and self-expressive brand provide better results when placed along similar factors, while relationship quality antecedents tend to confirm its importance regardless of the model. Furthermore, CBE dimensions were found to play the role of mediating variable, at least partially, in each of the models; they have also impacted brand loyalty in positive and negative ways.
Table 13. Structural model results and comparison.
Hyphotesis β S.E. C.R. Conclusion Model 2 Model 1
H1a: Satisfaction → Cognitive processing -0.027 0.087 -0.272 Not supported Not supported
H1b: Satisfaction → Affection 0.517 0.072 7.830 Supported Supported
H1c: Satisfaction → Activation -0.390 0.143 -3.516 Supported Supported
H8a: Satisfaction → Brand loyalty 0.262 0.109 2.878 Supported Supported
H2a: Trust → Cognitive processing 0.310 0.086 2.775 Supported Supported
H2b: Trust → Affection 0.378 0.068 5.304 Supported Supported
H2c: Trust → Activation 0.764 0.142 6.092 Supported Supported
H8b: Trust → Brand loyalty 0.234 0.110 2.226 Supported Supported
H3c: Commitment → Cognitive processing 0.152 0.098 4.699 Supported Supported
H3a: Commitment → Affection 0.060 0.061 2.038 Supported Not supported
H3b: Commitment → Activation 0.386 0.047 1.283 Not supported Supported
H8c: Commitment → Brand loyalty 0.502 0.071 7.835 Supported Supported
H4a: Involvement → Cognitive processing 0.010 0.049 0.179 Not supported Not supported
H4b: Involvement → Affection 0.010 0.038 0.290 Not supported Supported
H4c: Involvement → Activation -0.160 0.078 -2.762 Supported Not supported
H8d: Involvement → Brand loyalty -0.016 0.049 -0.411 Not supported Supported H5a: Participation → Cognitive processing 0.171 0.075 3.061 Supported Supported H5b: Participation → Affection 0.002 0.055 0.061 Not supported Supported H5c: Participation → Activation 0.063 0.111 1.107 Not supported Supported H8e: Participation → Brand loyalty -0.076 0.071 -1.915 Supported* Not supported H6a: Self-expressive brand → Cognitive processing
0.242 0.057 3.750 Supported Supported
H6b: Self-expressive brand → Affection 0.017 0.043 0.431 Not supported Supported H6c: Self-expressive brand → Activation -0.159 0.088 -2.341 Supported Supported H8f: Self-expressive brand → Brand loyalty -0.054 0.057 -1.137 Not supported Supported H7a: Cognitive processing → Brand loyalty -0.087 0.063 -1.868 Not supported Not supported H7b: Affection → Brand loyalty -0.053 0.104 -0.566 Not supported Supported
H7c: Activation → Brand loyalty 0.268 0.044 5.594 Supported Supported
Note: β= standardised regression weights, S.E.= standardised estimates, C.R.=critical ratio, * for a one-tail test
‧
國立 政 治 大 學