1982 Volume 22 Issue 1 Pages 37-48
Based on the attitude theory which was developed by Rosenberg (1956) and Fishbein (1963), multi-attribute attitude models have been widely used to consumers' attitudes toward brands. In the process of its application to consumer behavior, various models have been proposed. These models, however, include a number of significant research questions, such as conceptualization, measurement, and analysis.
The purpose of the present study is to test the construct and predictive validity of three linear-compensatory models; namely, Fishbein model, P-type Adequacy-Importance model, and S-type Adequacy-Importance model. Each model is formulated as: Aj=Σni=1aibij. Where, Aj: attitude toward brand j. n: number of attributes. ai: evaluative aspect of attribute i (Fishbein), importance of attribute i (P-type·S-type). bij: strength of belief that brand j possesses attribute i (Fishbein·P-type), satisfaction of brand j with respect to attribute i (S-type). Scaling: bipolar (Fishbein), unipolar (P-type·S-type).
The results are as follows:
In Exp. I, we applied ANOVA paradigm (Bettman et al., 1975a) to investigate the underlying multiplying assumptions, and found that 74.6% of subjects in the Fishbein task satisfied the assumption of the bipolar multiplying. This indicates that the Fishbein model is most valid for the construct. Many heterogeneous combination patterns were found in the P-type model, and also the multiplying assumption of S-type model failed to be supported.
In Exp. II, we have found that the predictive power becomes more prominent in the following order: S-type model>Fishbein model>P-tyrpe model. The ai component in the Fishbein model significantly increased the predictive power, which remarkably decreased in the P-type model by the cross-sectional analysis. The S-type model showed a high predictive power without multiplying ai component.
In conclusion, the Fishbein model is most valid at the point of two aspects. We cannot support the application of the P-type model. The conceptual criticism for the S-type model by Cohen et al. (1972) was valid.