Mediating and moderating effects in job design

Journal of Management, Dec, 1992 by Gary Johns, Jia Lin Xie, Yongqing Fang

Correspondence Between Psychological States and Specified Job Characteristics

The JCM suggests that specified job characteristics should account for substantial variance in their corresponding psychological states and that, controlling for this, unspecified job characteristics should not account for substantial additional variance. For example, in a regression equation using variety, identity, and significance to successfully predict meaningfulness, the addition of autonomy and feedback should not have a substantial effect. This analysis is shown in the top part of Table 2. For meaningfulness and responsibility, the addition of the unspecified job characteristics to the respective equations results in statistically significant 5 and 6% increments in explained variance. Only knowledge of results seems to correspond uniquely to its specified core characteristic, feedback.

A slightly different perspective is provided in the lower half of Table 2, where standardized regression coefficients for the full five variable equations for each psychological state are given. The model-specified coefficients for each state are bracketed for clarity. Again, the clearest case of model-specified correspondence TABULAR DATA OMITTED is that between feedback and knowledge of results. The most anomalous case is for experienced meaningfulness, where the regression weight for the model-specified variable of task identity is exceeded by those for the unspecified variables of autonomy and feedback.

Mediating Role of Psychological States

The JCM proposes that higher job scope is translated into favorable outcomes via the psychological states. Wall et al. (1978:188) have set down rigorous criteria for testing this mediating effect:

(a) the critical psychological states should account for sizable proportions of variance in each of the dependent |outcome~ variables; (b) the core job dimensions should add little to this when considered in the same analysis; (c) the core job dimensions alone should account for relatively little of the dependent |outcome~ variable variance; and (d) the critical psychological states should add considerably to this when considered in the same analysis.

As Wall et al. explain, these criteria can be examined using three regression equations for each outcome variable, one using the three psychological states as predictors, the second using the five core characteristics, and the third using all eight predictors together. This analysis is shown in Table 3. If the Wall et al. criteria are taken in a comparative rather than absolute sense, there is good evidence that the psychological states mediate the relationship between the core characteristics and both general satisfaction and internal motivation. The weakest evidence for mediation occurs for the self-rating of performance. The most anomalous result is for growth satisfaction. Here, the job characteristics account for substantial variance in the outcome variable and make a noteworthy incremental contribution, even when controlling for the states. In fact, in the eight-variable regression equation with simultaneous entry, the standardized regression weights for skill variety and autonomy exceeded those for knowledge of results and responsibility.