In order to determine whether the questions measured the expected items, we conducted a factor analysis including all the questions measuring consumer evaluation. The results of the test suggested extraction of two factors with eigenvalues above 1.00, which explained 70.13%
of the total variance. We expected that the test would extract four components pertaining to brand reputation, corporate ability, brand trust and brand attitude respectively. However, the test only extracted two components, which indicate that the total variance is better explained by two than by four factors. (See table 4.2)
Component
1
Component 2
Popular 0.703
Liked 0.626
Well known 0.793
High quality 0.810
Innovative 0.557
Leading company 0.873
Trust 0.809
Rely 0.815
Sincere 0.827
Safe 0.866
Dislike 0.925
Negative impression 0.935
Bad 0.938
Table 4.2: Factor analysis of the dependent variables
On the other hand, Singh (1991) explains that if there are substantial and significant differences in antecedents and consequences of the focal constructs, then one can claim nonredundancy among these. We argue that brand reputation, corporate ability, brand trust and brand attitude are strong pre-defined terms within brand management, and that they have different antecedents and consequences. In other words, we claim that there are theoretical justifications to view the four constructs as logically different conceptualizations. E.g. there are different underlying mechanisms behind brand reputation and brand trust. A consumer that perceives a brand’s reputation as good does not necessarily trust the brand. Thus, one cannot uncritically compute brand reputation, corporate ability and brand trust into one mutual variable.
Moreover, we conducted an additional factor analysis where we included the extraction of four factors. Additionally, we suppressed all absolute values under 0.6. Thus, we were able to
observe if there existed any differences which could cause different loadings. When examining the scree plot diagram, we observed that the third and fourth component explained 6.294 % and 6.021 % of the total variance. The eigenvalue scores were below 1.00 (0.818 and 0.783 respectively). By using a scree plot analysis of the eigenvalues, we observed a drop in eigenvalue between the fourth and fifth factor (see figure 6). This was consistent with the expected factor structure. The same approach was utilized by Nysveen et al. (2005) where factors with lower eigenvalues than 1.00 were included. According to Kaiser (1960, p. 143) cited in Rust el al.. (2004), the 1.00 eigenvalue cutoff is typically employed in marketing.
However, the author argues that this is just one of many possible cutoff criteria. Kaiser (1960) further states that the most important viewpoint for choosing the number of factors depends on the “psychological meaningfulness”. This means that the cutoff should be chosen such that the results are substantively meaningful. The eigenvalue of the fourth factor in our analysis was 0.783 while the eigenvalue of the fifth factor was 0.515. Considering our study design, it therefore seems meaningful to choose an eigenvalue cutoff that is located between 0.515 and 0.783. Additionally, the third and fourth factor would increase the total explained variance by 12.3%.
Figure 6: Scree Plot analysis of Eigenvalues
When examining the results illustrated in table 4.3, we noticed a pattern where the different questions loaded on the expected component. Popular, liked and well-known loaded on component 3. Trust, rely, sincere and safe loaded on component 1. Dislike, negative impression and bad loaded on component 2.
From these results, we argue that the questions in fact measure the four focal constructs respectively. Due to the suppression of absolute values below 0.6 we did not observe any loadings for innovative and leading company. This indicated that the questions might have been unfavorable in explaining corporate ability.
Although there were no distinct differences in factor loadings, we noticed that the questions loaded on the expected factor when extracting 4 factors. The minor differences between the items could be due to the differences in pre-existing attitudes toward the brands. Regarding the fictitious brands, consumers can have experience difficulties in evaluating the different questions. When consumers have no pre-existing attitudes it might be hard to separate between e.g. brand reputation and brand trust items. This can have affected our data and might thus work as a potential explanation for why the initial factor analysis only extracted two components.
Component
1
Component 2
Component 3
Component 4
Popular 0.782
Liked 0.668
Well known 0.809
High quality
Innovative 0.897
Leading company
Trust 0.904
Rely 0.915
Sincere 0.768
Safe 0.831
Dislike 0.913
Negative impression 0.926
Bad 0.927
Table 4.3: Factor analysis of the dependent variables with extraction of 4 factors
In connection to the pretest, we conducted a factor analysis for selected real brands in order to validate the Grohmann`s (2009) scale (see chapter 3.1.4). Here we found that the traits were reliable in terms of measuring the brand gender. It is also important to conduct a factor analysis for the fictitious brands we created in our main study. This will indicate whether the manipulation of brand gender was sufficient.
Component
1
Component 2
Adventurous 0.634
Aggressive 0.858
Brave 0.693
Daring 0.678
Dominant 0.801
Sturdy 0.650
Expresses Tender Feelings 0.549
Fragile 0.542
Graceful 0.835
Sensitive 0.853
Sweet 0.811
Tender 0.843
Table 4.4: Factor analysis for XB masculine and XB feminine
From table 4.4 we observe that the MBP traits (Adventurous, Aggressive, Brave, Daring, Dominant and Sturdy) load on the same component, while the FBP traits (Expresses tender feelings, Fragile, Graceful, Sensitive, Sweet and Tender) load on the other component. This indicates that the gender manipulation for the fictitious brands was sufficient. The results of the factor analysis conducted on both real and fictitious brands contribute to the validation of Grohmann`s (2009) scale.