5. Analysis
5.2 Survey
5.2.2 Results
To test the hypothesis it was necessary to create two new variables. These were defined as the score of the difference between the two different packaging rating of Bellavista. With this method we can see if a change in the self-‐image congruence and in the expression of preference has occurred. The new variables were then named as SELFCHANGE and PREFERENCECHANGE and of course they can have either negative value in case of change or they can be zero in the case where no change has occurred.
I’ve started testing the hypothesis that self-‐image congruence and preference were correlated.
To do so I have employed the Person r method, which is used to measure the correlation between two variables; it varies from 0 to 1, the higher it is the more correlated the variables are. (Bryman and Bell, pg 347) Measuring the correlation in this case is not enough, we have also to check if these correlations are statically significant, in other words we have to see if we can trust the results obtained. All of four brands hold a high statistical significance; the level was at 99% (alpha 0,01).
a) Bellavista with the old packaging: 0,534.
b) Ca’ del bosco: 0,623 c) Ferrari: 0,790
d) Bellavista with the new packaging: 0,768
e) The variable about the preference change and the self-‐change where positively correlated with Pearson’s R of 0,626.
After checking the correlation between the two principal variables, the competitors’ scores were not included in the analysis anymore. Their images and the related questions were placed between the two different Bellavista’s packaging; the reason was to give respondents
something to think about and to free their mind. The mental process was then more fluid and thus ensured the correct validity of the results. Another reason for including the two competitors was also to force consumers to concentrate on the questions and thus to give preference scores not in absolute but in relation with other two brands.
The four items about Bellavista were at fist examined in order to analyse the given hypotheses.
Firstly, the variables about the self-‐image congruence were tested with the Wilcoxon signed-‐
rank test; it is a non-‐parametric test that has to be used when it is needed to compare two related sample to assess if the population has expressed a different mean. It was found that they were significant with the p value of 0,05.
Using the same method, the two variables regarding the expression of preference were found not statistically significant with alpha 0,05; this shows us a problem that has to be evaluated.
The Wilcoxon test has evinced that the change in the self-‐image variables is statistically
significant while the change in the preference scores is not, this suggest not accepting the hypothesis and thus ending the analysis of the survey.
From a different perspective, it is worth to continue the research and to test the correlation between the self-‐image change and the preference change. Using again the calculus of the Pearson’s R for the correlations it is found that their value is statistically significant with a p value of 0,01 and their correlation score is 0,626 which means that the two variables are extremely positively correlated and thus a change in the self image congruence goes along with a change in the preference expressed by respondents.
Furthermore, if we take into account the frequency tables of these two variables we can note that the ties are almost equal (66 for the self image, while 70 for the preference) but the difference is given by the tendency of the respondents to have a greater change in the self than in the preference variable. That is maybe due to the fact that when they answer the survey, they first look at the image thinking about the congruence between them and the product user image.
Having 213 valid responses we have found that a change for the self-‐image has occurred for the 69% of them, 27,7% negative change and 41,3% positive change, the ties represent the 31%.
Regarding the preference scores, we have got the same valid responses but the data differs because the changes are the 67,1%, negative 28,6% and positive 39,5%, while the ties are 32,9%.
After the acceptation of all the hypotheses I have conducted further studies trying to understand which variables are linked and influence the self-‐image and the preference. I have employed some linear generalized model to find out that the only variable statically significant was the age. It was found that it influences the self-‐image positively.
The correlation between the age and the variables of self-‐image are all positive except for the value that includes the change between new and old Bellavista’s packaging. The Pearson’s R obtained was -‐0,194 statically significant for the p value of 0,01; this means that when the self-‐image decreases the age of the respondents increase and vice versa.
The hypotheses formulated in the third chapter were then confirmed:
H1: A change in the self-‐image congruence is correlated to the change in the expression of preference.
H2: If the packaging change is positively evaluated, the consumer’s self-‐image congruence score of the new packaging will be higher than for the old packaging. This will result in a greater preference toward the brand.
H3: If the packaging change is negatively evaluated, the consumer’s self-‐image congruence score of the new packaging will be lower than for the old packaging; consequently the preference toward the brand will diminish.