An examination of the relative predictive accuracy of three versions of the Vroom model of motivational force using within-subject and across-subject analysis

Three valence versions of the model of work motivation described in 1964 by Victor Vroom in Work and Motivation were examined with respect to their relative abilities to predict work effort. The examination was carried out under both the across-subjects and within-subject theoretical assumptions. Previous across-subjects studies had failed to indicate diffrential predictive abilities among the versions of the model with (1) positive and negative valences, (2) only positive valences, and {3) no valences. While previous prediction comparisons had used only across-subjects analysis, this study utilized both across-subjects and within-subject analysis for model comparisons.

The sample consisted of 31 employees of a small Virginia manufacturer. Nine model predictions each for overall job effort and effort on each subject's most import.ant task were obtained by the across-subjects analysis. Only one of these 18 predictions was statistically significant. In contrast, six of nine within-subject task effort rank predictions demonstrated statistical significance. The patterns of these findings were generally compatible with the results of previous research. Neither the across-subjects analysis nor the within-subject analysis indicated statistically significant differences among the abilities of the various valence versions of the model to predict effort.

The results of the cross-subjects analysis, combined with the findings of earlier studies, yielded the tentative conclusion that the across-subjects theoretical assumption may not be appropriate for tests of the ability of the Vroom force model to predict effort. Three conclusions were tentatively indicated by the results of the within-subject analysis. First, the Vroom force model is a significant, though moderate, predictor of effort. Second, the use of the within-subject theoretical assumption appears appropriate for testing the ability of the Vroom force model to predict effort. Third, the use of valence weights has no effect on the ability of the model to predict effort.