Scaling and Testing Multiplicative Combinations in the Expectancy-Value Model of Attitudes

This article examines the multiplicative combination of belief strength by outcome evaluation in the expectancy–value model of attitudes. Because linear transformation of a belief strength measure results in a nonlinear transformation of its product with outcome evaluation, use of unipolar or bipolar scoring must be empirically justified. Also, the claim that the Belief ¥ Evaluation product fails to explain significant variance in attitudes is found to be baseless. In regression analyses, the main effect of belief strength takes account of the outcome’s valence, and the main effect of outcome evaluation incorporates the outcome’s perceived likelihood. Simulated data showed that multiplication adds substantially to the prediction of attitudes only when belief and evaluation measures cover the full range of potential scores. Decisions under uncertainty involve two interdependent considerations: the likelihood that a contemplated action will result in certain outcomes, and the desirability or valence of these outcomes. Theoretically, knowing one or the other is not sufficient. Consider a woman who believes that using oral contraceptives will prevent pregnancy. By itself, this belief has no obvious implication for behavior. We also need to know how she values this possible outcome of using the birth-control pill. All else equal, women who regard preventing pregnancy to be desirable should be more likely to use the pill than women who consider pregnancy prevention to be undesirable. This analysis implies that the subjective values of a behavior’s outcomes are weighted or multiplied by the perceived likelihood that the behavior will produce the outcomes in question. Although the multiplicative combination of subjective probabilities and values is basic to many models of human judgment and decision making, its implications are often poorly understood. In this article, we deal with two related issues: the scaling of measures that enter into a multiplicative index; and empirical tests of multiplicative models, with special focus on the expectancy–value model of attitudes (Fishbein, 1963; Fishbein & Ajzen, 1975). We will try to show that multiplicative