Linear Attitude Models: A Study of Predictive Ability.

Prediction of consumer preference and purchasing behavior has been a topic of deep interest to marketing researchers for many years. In these studies, a subject's actual preference ordering for brands is compared with a predicted preference ordering based on his attitudes, where his attitude is considered as some composite of ratings on a number of specified attributes The studies cited all used the Fishbein model [8, 9], in which attitude is composed of two components, beliefs and values, This article compares the predictions from the Fishbein model with those from a simpler model against a criterion of stated preference. Predictions using the model were correlated with statements of preference in an actual product test. The simple beliefs-only model consistently performed better than the Fishbein model using the particular form of parameter estimation employed in this study, namely, determining importance weights or values from a four-point rating scale and beliefs from semantic differential scales. While the semantic differential is well accepted for determining beliefs, more controversy surrounds the way values should be determined. [ABSTRACT FROM AUTHOR]