Descriptive statistics showing the composition of sample and proportions were presented in table 9. The sample was predominantly composed of young people suggested by McDonald’s strategy target in terms of age and job. Disproportion of demographic variables according to time conducted by Chi-square tests indicated mostly no problem (p-value>0.05) except for income and job proportions. In addition, summary statistics indicated in general all constructs were above average in table 10.
In terms of correlation coefficients, traditional marketing elements were less correlated with experiential modules, indicating their distinctiveness.
The result of the 2nd-order CFA is shown in table 11 and appendix 1, with 4 categorical control variables, 2 latent control variables. Fit statistics were at their accepTable levels (Hatcher 1994; Byne 2001). Examination of construct validity shows all constructs are significantly converged including first- and second-order
ones(p-value<0.001), which ensures good convergent validity, and that all constructs are correlated but distinct in that the population correlation for all pairs of constructs do not include one under 95% confidence interval , which ensures discriminant validity.
Also, average variance- extracted estimates are all larger than the square of correlation coefficients of corresponding factors, which further confirms discriminant validity.
Reliability indices exhibit acceptable results with VE>0.5, CR>0.7, Cronbach’s alpha>0.7 for most of constructs recommended by Hatcher (1994). The aforesaid
Table 9. Descriptive statistics
Descriptive Statistics
Demographics Level Sample Proportion
Gender Male
Education Junior high schools
Senior high schools
Monthly Income Below 10000
10001~20000
Loyal segment True loyals
Brand switchers
Frequent motives Party with friends
Just dine
Table10. Summary statistics and correlations between constructs (n=313)
Table 11. Comparison of Competive theoretical models (CTM)
Measurement Model 1 Model 2
Explainatory Power(R2)
SMC Brand Awareness/Association 19.92% 22.79%
SMC Perceived Quality 32.75% 36.42%
Incremental Fit Measures
AGFI 0.7924 0.7924 0.7634
Table 12 standardized path coefficient estimates for the structural models under model 1 or 2
Dependent constructs Brand awareness/association Perceived quality Brand affect Brand loyalty
Competing theoretical models Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2
Research Variables
Holistic experiences 0.3616*** 0.1815+ 0.2814*** 0.0999 0.5629*** 0.6833*** 0.1367 0.3411**
Brand awareness/association 0.2886*** 0.2469* 0.1012 -0.0652 -0.0615
Perceived quality 0.2071** 0.0351 0.0387
Brand affect 0.2522 0.2792+ 0.4201*** 0.3534***
Latent Controls
Ad spending 0.2115** 0.1935+ 0.012 0.0001 0.0399 0.0792 0.0310 -0.0259
Price -0.0497 -0.0569 -0.1149+ -0.1238** 0.0494 0.0153 -0.0575 -0.0361
Categorical Controls
Gender -0.0269 -0.0262 0.0347 0.0340 -0.0029 0.0033 -0.0652 -0.0565
Age 0.1392* 0.1279 -0.1197* -0.1258* 0.0401 0.0353 0.0605 -0.0774
Frequent motives 0.0108 0.0081 0.00003 0.0036 0.0125 0.0240 0.095+ 0.1212
True Loyalty 0.0422 0.0344 0.0832 0.0761 0.0135 0.0305 0.2360*** 0.2027***
Split Loyals 0.0629 0.0611 0.0408 0.0409 -0.0059 0.0138 0.1981*** 0.1942*
Note. Dependent constructs are shown on the column; independent constructs are on the row. Significance claims:p-value<0.1 as
+;p-value<0.05 as *;p-value<0.01 as **; p-value<0.001 as ***.
The satisfactory result of measurement model justified the execution of structural models. Fitting respective structural models under integrated models 1 and 2 allowed the comparison of models to select a better model in the experience-based brand equity framework. Non-nested models disabled Chi-square difference test. Therefore,
three-step approach was here to play with results shown in table 11 and 12 (Huh et al.
2009; Rust et al. 1995). The first step compared model fit statistics of different models.
It appeared that integrated model 1 outperformed model 2, indicating model 1 worked better, which was cross-validated in all subsamples. The superiority of model 1 to model 2 ends the model comparison process. However, we show the whole comparison process for severe purpose.
The second step of three-step comparison addresses the SMC and significance of path coefficients. Structural models based on integrated models 1 and 2 were used to test hypotheses. H1, H2, H3, H5, H7, H9.2 were supported under integrated model 1 whereas H3, H5, H7, H4 were supported under integrated model 2, all with 5% level of significance. The significance of integrated model 1 was mostly
cross-validated by subsamples. However, the significance of integrated model 2 was unstable, not cross-validated. Therefore, we learn integrated model 1 outperforms 2 in terms of significance of path coefficients. SMC results show integrated model 2 outperforms 1 in the explanation of brand awareness/associations, perceived quality, and brand loyalty. However, given the unstable result of cross-validation of model 2, the result is questionable. The comparison of parsimony is the last step. It appears that integrated model 1 surpassed model 2 with the result cross-validated.
Under integrated model 1, results of the path analysis indicated brand
awareness/associations, perceived quality, and consumers’ affect toward the brand worked as complete mediators in experience-based brand equity framework, with results cross -validated. Therefore, appropriate strategic experiences are to be arranged to enhance corresponding mediating variables and further to enhance brand loyalty.
Further analysis under integrated model 1 was performed to identify the most effective route of holistic experiences on the increase of brand loyalty as shown in table 13. The result indicated the largest effect lies in the route in which experiences impact affect toward the brand, and then further brand loyalty. McDonald’s holistic experiences
retain customers much better than those appealing to customers’ cognition (Ajzen 2001). As experience hierarchy outperformed the others, we knew that relationship between attitudes and behaviors were better predicted if customers were in the direct experiences (Ajzen 2001). Therefore, the conative dimension of brand loyalty had higher predictive validity of true behaviors in the future. Standard learning hierarchy, in which holistic experiences impacted customers’ cognition, followed by affect, and in turn brand loyalty, was viable but of smaller effect, which was mostly cross-validated.
Low effect of standard learning hierarchy indicated few customers were high-involved when making decisions under experience contexts (Solomon 2009). The effect of holistic experiences on perceived quality, then on consumers’ affect toward the brand, and last on brand loyalty ranked second. The effect of holistic experiences on brand awareness/associations, then on perceived quality, then on consumers’ affect toward the brand, and last on brand loyalty ranked third. In terms of low involvement
hierarchy, that holistic experience had no direct effect on brand loyalty and/or had no indirect effect on brand loyalty through brand awareness/associations or perceived quality justified no effect of low involvement hierarchy. Therefore, behavioral learning was not effective to stimulate loyalty under experience context. Consumers came to consume not on a habitual basis in McDonald’s. Under integrated model 1, consumers’
cognitive and affective dimensions all worked well as complete mediators in the experience-based brand equity creation framework, supporting the claim of C-E that cognitive and affective dimensions are complete mediators. In addition, this kind of holistic processing indicated customers thought experiences as self-expressive or sensory-pleasing (Solomon 2009). In addition, to be quantitative, 87.56 % of holistic experience appealed to experiential hierarchy to attain its goal whereas only 12.44% of holistic experience appealed to standard hierarchy or C-E route to attain its goal. Our claim that C-E theory alone is deficient for explaining the relationship among holistic
Table13. Comparisons among indirect effects under the integrated model 1
Sequence of variables Hierarchy-of-effect route Hierarchy type Effect size
Experience→ Brand affect→ Brand loyalty Affect →Behavior Experiential 0.2365
Experience →Brand awareness/ associations
→Perceived quality →Brand affect →Brand loyalty
Cognition →Affect →Behavior Standard learning 0.0091
Experience→ Perceived quality →Brand affect
→Brand loyalty
Cognition →Affect→ Behavior Standard learning 0.0245
Table14. Comparisons among direct and indirect effects under the integrated model 2
Sequence of variables Hierarchy-of-effect route Hierarchy type Effect size
Experience→Brand affect→Brand loyalty Affect→Behavior Experiential 0.2415
Experience→Brand loyalty Behavior Low involvement 0.3411
experiences, Aaker’s brand equity dimensions, and brand affect is supported.
Under integrated model 2, the result of path analysis indicated that experiential hierarchy and low involvement hierarchy worked in the experience-based brand equity framework in table 14. The effects were all large. However, the model was questioned because cognitive dimensions failed to work as mediators and cross-validated results were unstable. Also, it was nonviable to attain the effectiveness by path defined by E-C theories. Thus, integrated model 2 was not appropriate for the examination of
experience-based brand equity framework. However, the result of the model further indicated the possibility that experiential hierarchy worked better, which bypassed the cognition toward the brand of consumers to influence brand loyalty. This result was cross-validated. In model 2, it is still deficient to explain experience-based brand equity framework only by E-C intervening mechanism in that experience hierarchy works but E-C fails if an integrated model is analyzed.
We also empirically tested hypotheses by using hierarchical linear modeling (HLM), in which there were no level two predictors under the integrated model 1 to reflect the sampling structure in which different customers in the same store may produce dependent responses. The HLM result was shown in appendix 2.
Variance-component types for level 2 errors were compound symmetry for perceived quality and brand loyalty, and variance components for brand awareness/association and brand affect due to better fit. All the constructs were computed as factor-based scores in which responses of items for the first order factors were averaged to obtain scores of the first order factors whereas scores of the first order factors were averaged to obtain the score of the second order factor. SEM and HLM results had the same result under 5% level of significance. Brand awareness/associations, perceived quality, consumers’ affect toward the brand all worked as complete mediators, corresponding to the C-E proposition that cognition, affect should be included when studying relationship between experience-response. Also, further analysis indicated that
experiential hierarchy ranked first in terms of the effectiveness of experience strategies.
Standard learning was of smaller effect. Significant path coefficients under SEM were also significant under HLM. However, the very aim of our study was to empirically test the theoretical framework. SEM outperformed HLM to take the whole research framework into account simultaneously. Therefore, SEM result was more rigorous and reliable than HLM result. In sum, the same result of SEM and HLM would justify the robustness of our study.
In integrated model 1, we performed a series of Chi-square difference tests to understand the best way of resource allocation under holistic experiences strategy, as shown in table15. The result indicated two possible ways of resource allocation. One way indicates the most important resource was the relate module (standardized
Table15. Series of Chi-square difference tests for strategic resource allocation
Hypotheses for resource
allocation
p-value of χ2
statistics
Meaning for arrangements of strategic
experiece modules
supported/not supported
H01: S=A=T=F=RE 0.00215*** No need for resource allocation Not supported
H02:S=A=T=F 0.0094** Equivalence of modules except Relate Not supported
H03:S=A=T=RE 0.0028** Equivalence of modules except Feel Not supported
H04:S=A=F=RE 0.0012** Equivalence of modules except Think Not supported
H05:S=T=F=RE 0.00256** Equivalence of modules except Act Not supported
H06:A=T=F=RE 0.0598+ Equivalence of modules except Sense Not supported
H07:S=A=T 0.0498* Relate, Feel are different Not supported
H08:S=A=F 0.0081** Relate, Think are different Not supported
H09:S=A=RE 0.0021** Think, Feel are different Not supported
H10:S=T=F 0.0021** Relate, Act are different Not supported
H11:S=T=RE 0.0020** Act, Feel are different Not supported
H12:S=F=RE 0.0008*** Think, Act are different Not supported
H13:A=T=F 0.1564 Relate, Sense are different Supported
H14:A=T=RE 0.0322* Sense, Feel are different Not supported
H15:A=F=RE 0.0246* Sense, Think are different Not supported
H16:T=F=RE 0.3500 Sense, Act are different Supported
loading=0.8769), followed by feel, act, and think modules (standardized
loading=0.7895), and sense module ranked the least (standardized loading=0.6477).
The second way indicates relate, feel, and think modules rank equally first (standardized loading=0.8325), followed by the act module (standardized
loading=0.7344), and sense module ranked the last (standardized loading=0.6589).
It was noteworthy to indicate that when traditional marketing variables such as ad and price were in the model, the standardized pure effects of holistic experiences on brand equity surpassed traditional marketing elements in table 1. Therefore, the
the brand, not traditional elements.
Under model 1, we discussed the effects of control variables in table 12. As for control constructs for traditional marketing elements, they were ineffective in terms of brand equity, representing marketing elements were in their saturated state. Ad
expenditures only had effect on brand awareness/associations, which corresponded to previous literatures. Price had negative effect on perceived quality against literatures.
McDonald’s chose delivering on dollar value as their experience value promise, making it reasonable for a lower price to enhance perceived quality. As for
demographics, gender failed as a control variable. Age succeeded partially. Customers less than 30 years old had higher brand awareness/association compared to those above 30. Now that the strategic target of McDonald’s is those below 30, this result seems reasonable. However, customers less than 30 years old had lower perceived quality toward McDonald’s compared to those above 30. Therefore, it is desirable for the company to boost perceived quality for those below 30. In terms of behavioral variables, loyalty segment worked as powerful controls under brand loyalty,
corresponding to Aaker’s proposition. Compared to brand switchers, true loyals and split loyals all had higher brand loyalty. As for phychographic variables, frequent motives had partially controlled. Compared to others, customers who just came to dine had higher brand loyalty. In addition, an alternative model where cognitive and
affective responses do not influence each other was also considered (Blackwell, Miniard, and Engel 2006). However, the fit was significantly worse than both the model 1 and the measurement model learned from Chi-square difference tests (both p-value<0.001). Thus, it was not considered in the analysis of this study. Under integrated models 1 and 2, the resulting framework was shown on Figures 3 and 4, respectively.
Figure3. Consequent framework for experience-based brand equity creation under model 1.
Note. Dotted line indicated insignificant path coefficients under 5% level of significance.
Figure4. Consequent framework for experience-based brand equity creation under model 1.
0.3411**
Note. Dotted line indicated insignificant path coefficients under 5% level of significance.