Expectancy Value Models of Attitude Measurement: A Measurement Problem

Expectancy value models of attitude have become increasingly popular in recent years and have inspired a large and growing body of research. The models to date are largely based on interpretations of Rosenberg's [13] or Fishbein's [6] model of attitudes. Although the literature has discussed different scaling procedures, most researchers employ fiveto sevenpoint Likert or bipolar type scales for measurement of the independent variables "belief" (Bj)and "desirability" (flj).' The differing techniques and opinions of the various researchers in this field are thoroughly discussed and reviewed by Wilkie and Pessemier [ 17]. Their review reveals that tbe marketing literature investigating attitudes has become quite sophisticated and refined. However, there is an important measurement problem in this area that has only been alluded to [2, 3] and which needs specific attention. This •is not to detract from previous work but simply an attempt to improve future efforts. Researchers in marketing have examined both additive [4] and multiplicative [1, 15] functional forms for the independent variables in the expectancy value models. Both Fishbein and Rosenberg specify the combinatorial rule as multiplicative [5]. An important scaling consideration in tests of multiplicative models is presented here. Advocacy of a multiplicative composition rule implies the desirability of that rule relative to alternative rules. But multiplication of scales, in order to be logically meaningful, requires the existence of a true rational zero point on both the measures entering into