Mediating and moderating effects in job design

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

The dominant paradigm in job design research, obtaining self-reports of job characteristics from job incumbents, seems to have weathered the storm provoked by Salancik and Pfeffer (1978). They argued that research concerning incumbent reactions to self-reported job characteristics was susceptible to priming effects, common method variance, and the influence of extraneous social cues.

A literature review by Fried and Ferris (1987) concludes that the evidence for priming effects is nil. Although common method variance might inflate relationships between self-reported job characteristics and worker responses (Roberts & Glick, 1981), it is now clear that the importance of this possibility has been exaggerated. Wagner and Crampton (1990) conducted a meta-analysis that compared "percept-percept" correlations with "multisource" correlations in various areas of organizational behavior, assuming that higher percept-percept associations would be indicative of method variance. In this analysis, job design emerged as one of the least tainted research areas. Finally, regarding extraneous social cues, all well designed studies in this domain (e.g., Griffin, 1983: Griffin, Bateman, Wayne, & Head, 1987) show strong and consistent effects for objective variations in job design in spite of the manipulation of social information.

The most influential model guiding self-report research on job characteristics has been the Job Characteristics Model (JCM, Hackman & Oldham, 1976, 1980). The JCM attempts to specify those job characteristics that lead to favorable work outcomes such as internal motivation, job satisfaction, and effective work behavior (good performance and low absence and turnover). It is expected that these outcomes will occur when workers experience three critical psychological states -- meaningful work, personal responsibility, and knowledge of results. In turn, five "core" job characteristics (skill variety, task identity, task significance, autonomy, and feedback) are posited to stimulate particular psychological states. Workers who exhibit high growth need strength, adequate knowledge and skill, and satisfaction with job context factors are expected to respond best to rich or high scope jobs. Measures of most JCM variables are provided in Hackman and Oldham's (1975, 1980) Job Diagnostic Survey. Although this instrument has some psychometric limitations, it has proven useful in job design research (Taber & Taylor, 1990).

The JCM has stimulated a large body of fragmented research. This work has been very ably summarized in a qualitative and quantitative review by Fried and Ferris (1987). The major purpose of the present study was to address a number of gaps in JCM research identified by Fried and Ferris, who generally concluded that the "JCM has received modest support" (1987:309).

One of the most critical gaps in JCM research involves how infrequently the total model has been tested. Although many studies have factor analyzed the core characteristics or correlated them with outcome variables, the rarity of studies that incorporate the mediating psychological states is remarkable. Hackman and Oldham (1976:255) have described the psychological states as "the causal core of the model," and the job characteristics were identified to service the states, not the other way around.

Of the previous tests of the complete model, Hackman and Oldham (1976) employed the largest sample, 658 individuals. However, the statistical procedures they used were not always elaborate enough to match the model's complexity. Wall et al. (1978) and Arnold and House (1980) used more elaborate analyses, but these studies both used small samples (47 and 90 respectively), a potential problem when a complex model with many variables is tested. Griffeth (1985) examined all variables in a field experiment with 57 subjects. Finally, Hogan and Martell (1987) used a sample of 208 and structural equations analysis to test competing versions of the JCM. In all of these studies it is unclear whether the data had a factor structure for the core characteristics that corresponded to that prescribed by the JCM. Fried and Ferris (1987) found that about half of the studies that factor analyzed Job Diagnostic Survey data failed to support a five-factor solution. In the present study, the five-factor solution was appropriate, resulting in a clear and fair test of the JCM.

Drawing on Hackman and Oldham (1976), Fried and Ferris (1987) note the essential questions to be answered in incorporating the psychological states into tests of the JCM:

1. Do the states correspond most closely to the core characteristics that the model specifies will influence them?

2. Do the states occupy a formal mediating role between the core characteristics and outcome variables?

3. Are all three states required to maximize the prediction of outcomes?

Besides mediation, the other critical process specified by the JCM is

moderation. Hackman and Oldham are explicit that this moderation is a dual location process, occurring both between the core characteristics and the psychological states and the states and the outcome variables. This contention has almost never been tested. Rather, most researchers have restricted moderator tests to the relationship between job characteristics and outcomes or overall job scope and outcomes. However, the exact location (or relative strength) of such moderator effects is important for theory development. Do inappropriate levels of the moderators preclude the development of favorable psychological states or do they simply preempt favorable outcomes?