It has been asserted that (I) tests of expectancy-value models require withinpersons analyses and that (2) within-persons analyses yield better predictions of behavioral tendencies than do across-persons analyses. The first assertion is correct; the second is not. Justification for within-persons tests of expectancy-value models must be made on theoretical rather than empirical grounds.
KEY WORDS Correlation Expectancy-value models Methodology Withinpersons analysis Idiographic vs. nomothetic analysisThe class of behavior models known as 'expectancy-value' or 'expectancy-valence' models has been used to study a wide variety of behavioral phenomena (Feather, 1982; Lawler, 1971, pp. 87-92; Mitchell and Biglan, 1971). These models vary in form, but all theorize that the strength or 'force'of an individual's tendency to act in a certain way is an additive function of the strength of his or her expectancies that the act or behavior will be followed by given outcomes multiplied by the desirabilities, values, or valences of those outcomes:where F, the strength or force of the individual's tendency to perform, attitude toward, or preference for, act i E,, =the strength of the individual's expectancy that act i will be followed by outcomej V, the valence or value of outcome j to the individual n = the number of outcomes.Expectancy-value theories assume that individuals choose from among alternative courses of action the one with the greatest force (Vroom, 1964, pp. 18-19). Any expectancy-value model is a model of the behavior of individuals. It theorizes that an individual is most likely to perform the behavior toward which he or she holds the most favorable combination of ), although many if not most continue to use across-persons analyses (). Perhaps because Mitchell's review focused on expectancy-value models of job satisfaction, occupational preference, and effort, withinpersons analyses tend now to be used more often in studies of organizational behaviors, whereas acrosspersons analyses continue to predominate in studies of social-psychological attitudes, health behaviors, and consumer preferences.Several recent studies have not only acknowledged that tests of the expectancy-value model require within-persons analyses, but have also hypothesized that within-persons tests yield better predictions of behavioral tendencies than do across-persons tests (Feldman-Summers and Ashworth, 198 1;Kennedy, Fossum, and White, 1983;Kopelman, 1977; Muchinsky, 1977; Murray and Frazier, 1986). These studies compared the results of within-persons and across-persons analyses performed on the same data sets and found, in support of the hypothesis, that the average (mean or median) within-persons correlation exceeded the average across-persons correlation. Mitchell (1982) concluded that 'the closer the methodological procedures are to the theoretical formulation, the better the prediction.'
Journal of Behavioral Decision Making