Growth need strength and context satisfactions as moderators of the relations of the job characteristics model

Growth need strength and context satisfactions as moderators of the relations of the job characteristics model

Robert B. Tiegs

Empirical investigations of the job characteristics model (JCM; Hackman & Oldham, 1980) have failed to systematically explore the moderating effects of growth need strength (GNS) and context satisfactions (viz., pay, job security, co-worker and supervision) on the relations among the core job characteristics, critical psychological states, and work outcomes. Previous studies also are criticized the use of subgroup analytic techniques, low statistical power resulting from small sample sizes (i.e, often less than 200) and/or samples consisting of individuals of relatively homogeneous jobs/occupations. As an attempt to address these deficiencies in the literature, this study examined the moderating effects of GNS and each of the four context satisfactions using a large sample (N = 6405) of employees from a variety of jobs and occupations. Overall, the results of univariate and multivariate hierarchical moderated multiple regression analyses suggest that none of the five individual difference factors appeared to be viable moderators of any of the relations among job characteristics, psychological states, and three work outcomes (viz., growth satisfaction, overall job satisfaction, and internal motivation). Also, there was no supportive evidence for potential joint moderating effects between GNS and each context satisfaction on the relations of the JCM.

During the past decade, the job characteristics model of work motivation (JCM; Hackman & Oldham 1976; 1980) has been a dominant theoretical framework for understanding an employee’s reactions to the core dimensions of the job (Fried & Ferris, 1987; Kulik, Oldham, & Hackman, 1987). However, this model and associated research has not been without criticisms (for more detailed review and discussion of the issues that have been raised, see Fried & Ferris, 1987, and Roberts & Glick, 1981). The purpose of the present study is to address one of these issues: namely, the extent to which the linkages among job characteristics, psychological states, and work outcomes are moderated by growth need strength (GNS) and work context satisfactions.

JCM and Prior Research

Figure 1 depicts the relations among the constructs of the JCM. In brief, the JCM (Hackman & Oldham, 1976, 1980) posits that a high level of internal (i.e., self-generated) motivation is dependent on the presence of three critical psychological states: namely, experienced meaningfulness, experienced responsibility, and knowledge of results. Although of lesser importance, other work-related outcomes influenced by the psychological states include overall job satisfaction and growth satisfaction (i.e., satisfaction with opportunities for self-enhancement on the job). The development of each of the psychological states is fostered by one or more core characteristics of the job. Specifically, it is proposed that experienced meaningfulness arises from the compensatory relationship among skill variety, task identity, and task significance. Autonomy and job feedback are the antecedents of experienced responsibility and knowledge of results, respectively. Following Hackman and Lawler (1971), Hackman and Oldham (1976; 1980) stipulate that it is the perceptions of the core job characteristics that are directly antecedent to the critical psychological states, rather than the objective job properties. However, convergence between the objective properties of the job and perception of those properties is expected because the objective properties are specified as influencing one’s perception of job dimensions (for a more detailed discussion of this issue, see Fried & Ferris, 1987, and Taber & Taylor, 1990). Finally, the most recent version of the JCM (Hackman & Oldham, 1980) postulates that individuals’ reactions to job characteristics and to psychological states are moderated by the strength of their needs for personal growth and accomplishment at work (i.e., GNS), and satisfaction with certain contextual aspects of their work environment (viz., pay, job security, co-worker, and supervision). Oldham and his colleagues (Hackman & Oldham, 1980; Oldham, Hackman, & Pearce, 1976) further suggest that GNS and each context satisfaction may jointly moderate the relations of the JCM.

The linkages among job characteristics, psychological states, and work outcomes as specified by the JCM have been generally supported (for a recent review see Fried & Ferris, 1987). However, contrary to the JCM, a number of literature reviews (Bottger & Chew, 1986; Fried & Ferris, 1987; Graen, Scandura, & Graen, 1986; Kulik et al., 1987, O’Brien, 1982; Roberts & Glick, 1981) have concluded that empirical support for moderating effects of GNS and context satisfactions have been weak and inconsistent. These findings have led some to advocate either eliminating GNS and context satisfactions from further consideration as moderators (e.g., Bottger & Chew, 1986; Kulik et al., 1987) or reformulation of the JCM (Graen et al., 1986).

It is important to note that these recommendations are based almost exclusively on results of studies that only examined the moderating influence of GNS and context satisfactions on job characteristics-outcome relations. In other words, very little is known about whether the focal individual difference factors moderate the relations of the JCM as proposed in the model. Specifically, investigation of GNS and context satisfactions as moderators of the job characteristics-psychological states relations and the psychological states-outcome relations has been limited to only a few studies in the case of the GNS (Arnold & House, 1980; Hackman & Oldham, 1976; Hogan & Martell, 1987) and none in the case of context satisfactions (Fried & Ferris, 1987). Also, the proposed joint moderating effect of GNS and each context satisfaction (Oldham et al., 1976) has not been extended to the relations among individual job characteristics, psychological states, and work outcomes. That is, Oldham et al., which appears to be the only study to examine GNS and each context satisfaction as joint moderators, restricted its attention to the relations between the overall motivating potential of the job |as assessed by the summary multiplicative measure of all five job characteristics (MPS) given in Hackman & Oldham, 1976~ and several work outcomes. In our empirical examination of the hypothesized individual and joint moderating effects, we attempted to overcome several shortcomings (e.g., small samples, restricted range of jobs, inappropriate data analytic techniques) that have been associated with previous research in this area.

Research Objectives

The lack of a systematic investigation of the individual and joint moderating effects of GNS and context satisfactions suggests that the position of those who argue for a discontinuance of research on GNS (e.g., Bottger & Chew, 1986) and context satisfactions as moderators (e.g., Kulik et al., 1987) may be premature. To examine this possibility, this study had three major objectives.

The primary goal of this study was to assess the extent to which GNS and each of the four context satisfactions moderate the relations between each job characteristic and its associated psychological state and the relations between each psychological state and three work outcomes: overall (general) job satisfaction, growth satisfaction, and internal motivation. With respect to the moderating effects of GNS on the relations between job characteristics and psychological states, Hackman and Oldham (1976; 1980) state that the complexity of a job is, in part, a function of the levels of the five core job characteristics. They define a complex job as one requiring the use of a number of skills, abilities, knowledge, and information. Consequently, individuals with high GNS will perceive such jobs as providing “opportunities for considerable self-direction, learning, and personal accomplishment” (Hackman & Oldham, 1980: 85). Thus, a positive relation is predicted between the core dimensions of the job and cognitive reactions (i.e., psychological states) of high GNS individuals. On the other hand, because low GNS individuals do not seek to satisfy needs for personal growth and accomplishment at work, these individuals are expected to be, at best, indifferent to the degree of complexity in their jobs.

The rationale underlining the moderating effects of context satisfactions on job characteristics-psychological states relations is that dissatisfaction with a particular contextual factor serves to focus the attention of jobholders toward coping with the experienced problem. As a result of this distraction, employees will be less responsive to the motivating properties of their jobs. Thus, it is only when jobholders are relatively satisfied with an extrinsic factor such as pay, job security, and their relationships with their co-workers and supervisors that they will be able to perceive and respond to the motivating potential of their work (Hackman & Oldham, 1980: 86).

With respect to psychological states-outcome relations, Hackman and Oldham (1976; 1980) suggest that it may be that all individuals experience the psychological states when the core dimensions are high, regardless of their level of GNS or context satisfactions. However, high GNS people will respond more strongly to the psychological states (i.e., they will be more satisfied and more motivated) because they value these cognitions more than lower GNS counterparts. Similarly, individuals who are satisfied with the contextual aspects of their work environment should also respond more strongly to the presence of the critical psychological states relative to those individuals whose attention is focused primarily on coping with perceived pay inequities, job insecurity, or poor relationships with their co-workers and superior(s).

The second objective of the present study was to extend the research of Oldham et al. (1976) by investigating the possibility that GNS and each context satisfaction may jointly moderate the relations between job characteristics and psychological states and between psychological states and work outcomes. The theoretical rationales for these joint moderating effects are simply extensions of those provided for the individual moderating effects of GNS and a focal context satisfaction (see above) and the rationale provided by Oldham et al. (397) with respect to the joint moderating effect of these individual difference factors on the relations between MPS and work outcomes. Specifically, it is proposed here that the positive relations between job characteristics and psychological states will be the strongest for high GNS individuals who are not distracted by a particular contextual factor. Alternatively, it may be that all jobholders experience the psychological states when the core job characteristics are high. However, the hypothesized positive relationships between psychological states and work outcomes will be the strongest for high GNS individuals who are satisfied with a particular contextual aspect of their work environment. For both sets of relations (i.e., between job characteristics and psychological states and between psychological states and outcomes), these positive relationships will be attenuated when either GNS or a focal context satisfaction are low. Finally, if both of these individual difference factors are low, then the relationships among job characteristics, psychological states, and work outcomes may become negative because of the extremely poor fit between the needs of the jobholders and the capacity of the job and work environment to satisfy these needs. In such cases, the jobholders may become psychologically overwhelmed by the complexity of the job (Hackman & Oldham, 1980: 86)

Finally, the present study sought to extend the work of Oldham et al. (1976) by investigating the joint moderating effect of GNS and each context satisfaction on the relations between individual job characteristics and a focal work outcome. In so doing, this study also attempted to replicate the findings of those studies that focused on the moderating influence of GNS and context satisfactions on these relations.

In examining the potential moderating effects of GNS and context satisfactions, this study attempted to overcome several methodological and statistical shortcomings that have characterized previous research in this area (see Bottger & Chew, 1986; Champoux, 1981; Graen et al. 1986; O’Brien, 1982; O’Connor, Rudolf, & Peters, 1980). Among the more important of these deficiencies are four in particular. First, in a number of studies (e.g., Abdel-Halim, 1979; Arnold & House, 1980; Oldham, 1976; and for other studies focusing on GNS, see Graen et al., 1986) the moderating effects of GNS and context satisfactions were examined with a relatively small sample of jobholders (i.e, often less than 200), which raises the concern regarding the extent to which the obtained results were subject to sampling error (Stone, 1988). Second, and often associated with the first, is the fact that many studies (e.g., Abdel-Halim, 1979; Arnold & House, 1980; Bottger & Chew, 1986; Hogan & Martell, 1987; Katerberg, Hom, & Hulin, 1979; Oldham, 1976; Oldham et al., 1976; Orpen, 1979; Pokorney, Gilmore, & Beehr, 1980) have examined GNS or context satisfactions as moderators using data provided by incumbents of a restricted range of jobs. The results of these analyses may be suspect in light of Aldag, Barr, and Brief’s (1981) observation that the psychometric integrity (e.g., reliability and dimensionality) of the more common self-report job characteristic measures (in particular, the Job Diagnostic Survey developed by Hackman & Oldham, 1975), appear to suffer when samples consist of relatively homogeneous jobs. Third, across all studies, the moderating effects of context satisfactions have been examined on a limited number of jobs or occupations, in particular white collar jobs such as managers, professionals, technicians, salespersons, and clerical workers (e.g., Abdel-Halim, 1979; Bottger & Chew, 1986; Champoux, 1981; Oldham, 1976; Oldham et al., 1976). Thus whether the findings of these studies generalize to other types of occupations (e.g., skilled trades and blue collar workers) is not known. Finally, some previous studies relied exclusively on subgroup analytic techniques (e.g., Hackman & Oldham, 1976; Oldham et al., 1976; Sims & Szilagyi, 1976) to evaluate moderators. These types of analytic methods have been shown to have a number of drawbacks in comparison to hierarchical moderated multiple regression (see Cohen & Cohen, 1983, and Stone, 1988). As an attempt to avoid the above mentioned limitations, data obtained from a large sample of incumbents (N = 6405) working in a wide variety of jobs were analyzed using hierarchical moderated multiple regression.

Method

Sample and Instrumentation

The hypothesized moderating effects were evaluated using the database of Oldham, Hackman, and Stepina (1979), which consists of 6930 employees working on 876 jobs (ranging from upper managerial positions to unskilled labor) in 56 organizations. (See the technical report of Oldham et al. for a detailed description of this sample.) Respondents missing information on one or more of the items of a measure were excluded from all data analyses, reducing N to 6405.

The self-report scales of the Job Diagnostic Survey (JDS, Hackman & Oldham, 1975) were used to collect information on all 16 variables subjected to data analyses. The reader is referred to Hackman and Oldham (1980, Appendix A) for a listing of the items for each scale, and to Fried and Ferris (1987) and Taber and Taylor (1990) for recent reviews on the psychometric properties of these measures. Following the recommendation of Aldag and Brief (1979), the job choice scale of the JDS was used as the measure of GNS. Table 1 presents descriptive statistics, estimates of internal consistency (Cronbach’s coefficient alpha) and the intercorrelations among the 16 variables.

Because the JCM specifies that it is the perceptions of job characteristics that give rise to the critical psychological states, the measurement of job characteristics with self-report scales is appropriate. Although our database precludes an examination of Hackman and Oldham’s (1975; 1980) proposition that objective properties of the job influence perceptions of job characteristics, two recent reviews of the literature on this issue (Fried & Ferris, 1987; Taber & Taylor, 1990) have similarly concluded that the available evidence suggests that objective job properties are the primary determinants of perceived job characteristics.

Data Analysis

As Table 1 indicates, there were non-trivial amounts of shared variance among the psychological states and among the work outcomes. Therefore, rather than ignoring these intercorrelations, we used a data analytic procedure that allowed explicit acknowledgement of these intercorrelations while examining the hypothesized interactions. Specifically, analysis of the data was conducted in two stages. First, a number of univariate hierarchical moderated multiple regressions (UHMMR) (Cohen & Cohen, 1983; Stone, 1988) were performed. This permitted replication of previous research findings because prior studies, with the notable exception of Hogan and Martell (1987), had each relied on some form of univariate analyses to examine the hypothesized moderators. The UHMMR TABULAR DATA OMITTED analyses also served to screen out from stage 2 analyses those interactions that failed to account for a meaningful amount of explained variance (see Results section) in a given psychological state or work outcome.

In the second stage, the interactions retained from stage 1 were evaluated in multivariate hierarchical moderated multiple regression (MHMMR). In contrast to UHMMR analysis, MHMMR analyses not only allowed for simultaneous estimation of multiple structural equations (one equation per criterion), but also permitted modelling of correlated criteria.

Several steps were take to minimize the adverse affects of multicollinearity on the computation of the partial regression coefficients in both the UHMMR and MHMMR analyses. First, all variables were standardized to have mean of zero and unit variance. Cross-product terms were then formed using the appropriate standardized variables. Next, following the residual-centering procedure as outlined by Lance (1988), each cross-product term was regressed on its constituent elements to obtain standardized residuals. These residualized cross-product terms were then used in both the UHMMR and MHMMR analyses.

As an additional means of minimizing multicollinearity among the predictors in the UHMMR analyses, each GNS-context satisfaction pair was examined in separate regression models. Multicollinearity was reduced further in the MHMMR analyses through the elimination of those interactions that were not supported by the UHMMR analyses.

Results

Univariate Hierarchical Moderated Multiple Regressions

From the three sets of relations among the primary variables of the JCM (i.e., job characteristics, psychological states, and work outcomes), nine basic UHMMR models were constructed. Three of these basic models pertained to the job characteristics-psychological states linkages involving (a) experienced meaningfulness and each of its antecedent job characteristics (skill variety, task significance, and task identity); (b) autonomy and experienced responsibility; and (c) job feedback and knowledge of results. Three other basic models focused on each of the simultaneous relations of the three work outcomes with the three psychological states. The final three models examined the simultaneous relations of each work outcome with the five job characteristics. These last three models were included as an attempt to fully examine the different segments of the JCM.

For each UHMMR performed, the sequential entry of main effects and interaction terms into a focal hierarchical equation was as follows:

1. a set of main effect terms consisting of either job characteristics or psychological states

2. the main effect term of GNS

3. the main effect term of a particular context satisfaction

4. the GNS x Context Satisfaction cross-product term

5. a set of Job Characteristics (Psychological States) x GNS cross-product terms

6. a set of Job Characteristics (Psychological States) x Context Satisfaction cross-product terms

7. a set of Job Characteristics (Psychological States) x GNS x Context Satisfaction cross-product terms.

Steps 5 and 6 test for the individual moderating effects of GNS and a focal context satisfaction, respectively, on the relations between job characteristics (psychological states) and a focal criterion. Step 7 evaluates the joint moderating effects of GNS and the local context satisfaction on a particular relation among job characteristics, psychological states, and work outcomes. Also note that with two exceptions (viz., the models involving the autonomy-experienced responsibility and job feedback-knowledge of results relations) step 1 and steps 5 through 7 consisted of a set of predictors that were simultaneously entered into a local regression equation. The rationale for this is the lack of a priori theoretical justification for specifying the importance among the predictors of a set. In sum, separate UHMMR analyses were conducted for each GNS-context satisfaction pair, resulting in 36 (9 models X 4 GNS-context satisfaction pairs) analyses.

To control Type I error rates, we followed Cohen and Cohen’s (1983) adaptation and generalization of Fischer’s protected t-test procedure, which requires that statistical evaluation of individual elements of a focal set of predictors is contingent on the statistical significance of the overall F test (based on Model I error, Cohen & Cohen, 1983) for the change in |R.sup.2~ attributed to the set of predictors. Also, to maintain the nominal level of alpha at .05 for the overall F tests, we employed the Bonferroni approach and divided this alpha level by the number of non-redundant, investigationwise overall F tests (i.e, 198). The resulting quotient (.0002) was then used as the statistical criterion to evaluate each set of predictors and, if appropriate, individual predictors within a set. Finally, because Type II error rates are considerably reduced by the large N of this study, it is possible for a given set of predictors to be judged as statistically reliable, but account for a trivial amount of the explained variance in a focal dependent variable. Therefore, each set of predictors was judged as having substantive importance if its entry into the regression equation increased |R.sup.2~ by more than .01 (cf. Cohen, 1977; Cohen & Cohen, 1983).

Table 2 summarizes the results of the 36 UHMMR analyses. Starting with the three-way interactions (Step 7) and proceeding backwards, the results indicated that only 2 of the 36 sets of three-way interactions (Job Characteristics or Psychological States x GNS x Context Satisfaction) were statistically significant (p |is less than~ .0002). Moreover, the median |delta~|R.sup.2~ attributable to the entry of a set of three-way interactions across the 36 models was .000 (rounding to the third decimal), and 83.33% (30) of these |R.sup.2~s were below .001. Even the strongest of these three-way interaction sets (viz., the set of Job Characteristics x GNS x Co-worker Satisfaction) was able to uniquely explain only 0.3% of the variance in overall job satisfaction. Thus, the data did not support the hypothesized joint moderating effect of GNS and each context satisfaction on any of the relations of the JCM.

The individual moderating effects of the four context satisfaction variables on the relations among job characteristics, psychological states, and work outcomes also were extremely weak or non-existent. Specifically, although 11 of the 36 sets of two-way interactions (Step 6) passed the statistical criterion, none of these accounted for more than 0.5% of the variance in a focal dependent variable. In fact, |delta~|R.sup.2~s ranged from .000 to .005 (Md |delta~|R.sup.2~ = .0015) across the 36 sets of two-way interactions involving the four context satisfactions.

In the moderating effects of GNS (Step 5), only the set of Psychological States x GNS interactions in predicting internal motivation failed to reach statistical significance. However, Table 2 indicates that the median |delta~|R.sup.2~ of 3 of the remaining 8 sets of two-way interactions was greater than or equal to .01 |recall that Job Characteristics (Psychological States) x GNS interactions were examined four times with respect to a given relation: one for each GNS-context satisfaction pair~. These three sets included (a) the set of Job Characteristics x GNS interactions on experienced meaningfulness, (b) the set of Psychological States x GNS interactions on growth satisfaction, and (c) the set of Job Characteristics x GNS interactions on growth satisfaction. Thus, of all the two- and three-way sets of interactions examined in stage 1, only these three sets of interactions were further examined in the second stage.

Multivariate Hierarchical Moderated Multiple Regressions

The intercorrelations among all predictors (including main effect terms and interaction terms) and criteria (either psychological states or work outcomes, depending on the relations examined) were used as input to LISREL-PC 7 (Joreskog & Sorbom, 1989) to conduct MHMMR analyses using maximum likelihood estimation. Note that, in addition to the interactions not supported in step 1, all main effects of the context satisfactions were excluded in these analyses. Also eliminated TABULAR DATA OMITTED from these analyses were the main effects of GNS on those criteria whose relations with job characteristics or psychological states were not moderated by GNS. These effects were excluded from the MHMMR analyses because the JCM does not explicitly acknowledge them.

Tetrick and LaRocca’s (1987) analytic strategy for evaluating moderators within a multivariate framework was used (also see Kenny & Judd, 1984). This procedure involves estimating the fit of a series of sequential, nested models in which each model in the series is progressively more restricted (i.e., has fewer parameters to be estimated) than those that precede it. An important distinction between this procedure and UHMMR is the order in which interactions are evaluated. Specifically, in this procedure the model in which all focal main effect and interaction terms are included is estimated first and then re-estimated with a parameter for a given interaction term fixed at zero. Then the fit of these two models to the observed data are compared. If the more restricted model fits more poorly than the less restricted model, then the hypothesized interaction is supported.

The decrease in overall fit attained by a more restricted model relative to the fit obtained by a less restricted model can be statistically evaluated by conducting chi-square difference tests (Bentler & Bonnett, 1980). However, as was the case with the UHMMR analyses, the large sample size of the present study makes these inferential tests extremely powerful and thus can lead to the rejection of the more restricted model in favor of the less restricted model even though there are trivial differences between the two models. Bentler and Bonnett (1980) and others (e.g., Marsh, Balla, & McDonald, 1988) have recommended in such instances that researchers rely on non-statistical indices of overall fit of the target model to the observed data to evaluate a particular model as well as to make model comparisons. Here, the following incremental indices of fit were calculated: The Tucker-Lewis Index (TLI, Tucker & Lewis, 1973), the normed-fit-index (NFI, Bentler & Bonnett, 1980) and the parsimonious-fit-index (PFI, James, Muli ak, & Brett, 1982). The TLI and NFI assess the degree of fit achieved by a target model relative to the null model (which specifies no relationships among any of the variables). For both indices, values greater than .9 are typically used to indicate good fitting target models. The PFI (= the ratio of degrees of freedom of the target model to the degrees of freedom of the null model times NFI) is an adjustment to the NFI in that it penalizes a less restricted model that achieves a good fit by wasting degrees of freedom (James et al., 1982). In addition to these incremental indices of fit, the following stand-alone indices (Marsh et al., 1988) provided by LISREL were also used: the goodness-of-fit index (GFI), the adjusted goodness-of-fit index (AGFI), and the root-mean-squared-residual (RMSR). Supplementing these overall indices of fit were the total coefficient of determination and the coefficient of determination of each criterion. In general, support for a hypothesized effect was established if the more restricted model in which this effect was fixed at zero resulted in a sizeable increase in RMSR and noticeable decreases in the other overall indices of fit relative to the less restricted model. Consistent with the UHMMR analyses, additional support for a hypothesized interaction was a drop in |R.sup.2~ for the focal criterion by more than .01 in the more restricted model as compared to the corresponding value in the less restricted model.

Because the order in which the hypothesized effects were examined was similar for each of the three sets of interactions retained from step 1, it will be illustrated in detail only for the sequential, nested models used to evaluate the moderating effects of GNS on the relations between experienced meaningfulness and its associated job characteristics. The least restricted model (M1) of that sequence involved regressing (a) experienced meaningfulness on skill variety, task significance, task identity, GNS, and the three Job Characteristics x GNS interactions; (b) experienced responsibility on autonomy; and (c) knowledge of results on job feedback. All three regression equations were estimated simultaneously. Also freed to be estimated in this model were the parameters of the PSI matrix, including the diagonal elements (which were variances of the disturbance terms) and off-diagonal elements (which were correlations among the disturbance terms). Freeing the off-diagonal elements of the PSI matrix represented an attempt to account for the shared variance among the psychological states by acknowledging the possibility that these criteria are a function of each other or that all three criteria are related to one or more omitted (nonmeasured) exogenous variables (Hayduk, 1987). These omitted exogenous variables may include, for example, other organizational constructs not incorporated in the JCM (e.g., role stressors and leadership style) or the possibility of a measurement artifact arising from the fact that the three variables were measured using self-report scales (cf. Williams, Cote, & Buckley, 1989).

Three more restricted models (M2A, M2B, and M2C) were then estimated to evaluate each of the three interactions terms individually. In each of these models, the parameter representing the effect of a particular Job Characteristic x GNS interaction term on experienced meaningfulness was fixed at zero, whereas the parameters of the other interaction terms of that set were freed (note that though each of these models are nested within M1, they are not nested within each other). To evaluate all three interaction terms simultaneously, another model (M3) was estimated in which the parameter of each interaction term in the set was fixed at zero. The remaining progressively more restricted models were a model (M4) with the same specifications as M3, except that the main effect of GNS on the focal criterion was fixed at zero; a model (M5) with same specifications as M4, except all off-diagonals of the PSI matrix were fixed at zero; and the null model (M6).

Table 3 summarizes the results of the sequential, nested models that addressed the moderating effects of GNS on the relations between experienced meaningfulness and its three antecedent job characteristics (skill variety, task significance, and task identity). A comparison of the least restricted model (M1) with M2A shows that |R.sup.2~ for experienced meaningfulness decreased by only .01 in M2A, which just meets our cut-off point. This comparison also reveals that the goodness of fit indices for M2A either decreased slightly (GFI and NFI) or actually increased (AGFI, TLI, and PFI) relative to their counterparts in M1. Overall, then, these comparisons can be interpreted as failing to support the moderating effect of GNS on the skill variety-experienced meaningfulness relation. Similar comparisons of M1 with M2B and M2C clearly indicate no support for the hypothesized Task Significance x GNS and Task Identity x GNS interactions, respectively. The hypothesized moderating effects of GNS on relations between experienced meaningfulness and its three antecedent job characteristics were further disconfirmed by the fact that the indices of fit changed little, relative to their counterparts in M1, when the parameters of all three interactions were simultaneously fixed at zero (i.e., M3). To complete the picture, the results appeared to favor rejection of the main effect of GNS on experienced meaningfulness (compare M3 with M4), but support the assumption that the three psychological states are either a function of each other or related to one or more omitted variables (compared M4 with M5). In sum, the best fitting model of the eight evaluated was M4, which depicted the hypothesized direct relations between the job characteristics and psychological states, correlated disturbance terms, and fixed at zero parameters for the main effect of GNS and all three interaction terms.

In the sequential, nested models that evaluated the moderating effect of GNS on the relations between the psychological states and growth satisfaction, the least restricted model involved regressing growth satisfaction on each psychological state, GNS, and the three Psychological States x GNS interactions terms. Also, overall job satisfaction and internal motivation were both regressed on all three psychological states and all elements of the PSI matrix were freed to be estimated. Inspection of Table 4 suggests that none of the three models where the parameter of a given interaction term was fixed at zero (M2A to M2C) resulted in noticeable decreases in the indices of fit relative to their counterparts in the least restricted TABULAR DATA OMITTED TABULAR DATA OMITTED model. Also supporting this conclusion is the fact that the coefficient of determination for growth satisfaction decreased by at most .005 in any of these three more restricted models. Moreover, in the model where all three parameters of the interaction terms were fixed at zero simultaneously (M3), the indices of fit deviated little from their counterparts in the least restricted model. Comparison of M3 with M4, and M4 with M5 suggests the elimination of the main effect of GNS on growth satisfaction, but retaining of the correlated disturbance terms. Thus the best fitting model of this sequence was M4, which depicted the hypothesized relations between each psychological states and each outcome, correlated disturbance terms, and fixed at zero parameters for the main effect of GNS on growth satisfaction and all interaction terms.

In the series of nested models that examined the moderating effects of GNS on the relations between the five job characteristics and growth satisfaction, the least restricted model involved regressing growth satisfaction on each of the five job characteristics, GNS, and each of the five Job Characteristic x GNS interactions. Also freed were the parameters relating each job characteristic to overall job satisfaction and internal motivation as well as all elements of the PSI matrix. As Table 5 shows, when compared to the least restricted model (M1), the indices of fit and the coefficient of determination for growth satisfaction of the five models where each interaction was fixed at zero (M2A to M2E) either did not change (e.g., RMSR) or were not noticeably reduced. In fact, for some of the five models, several of the indices (e.g., TLI, NFI, PFI) actually increased. Moreover, because the model where all five interactions were fixed at zero simultaneously also failed to appreciably change the overall fit indices, it appears that none of the five Job Characteristics x GNS interaction were substantively important. Consistent with conclusions reached in the models examining psychological states-outcome relations, the results presented in Table 5 also suggest elimination of the main effect of GNS on growth satisfaction (compare M3 with M4), but retaining of the correlated disturbance terms (compare M4 with M5). Based on these comparisons, the best fitting model of the sequence was M4, which specified each job characteristic being related to each outcome, correlated disturbance terms, and fixed at zero parameters for the main effect of GNS on growth satisfaction and all interaction terms.

Discussion

The impetus of the present study was to provide empirical information concerning the validity of the hypothesized moderating effects of GNS and each context satisfaction, which heretofore have been either inadequately examined or ignored in the literature on the JCM. We found that when evaluated in terms of statistical and practical significance, the data of this study generally did not support TABULAR DATA OMITTED either the individual moderating effects of GNS and context satisfactions or the joint moderating effect of GNS and each context satisfaction on the relations among job characteristics, psychological states, and motivational and affective outcomes. In contrast to previous studies, the results of the present study can not be discounted on the basis of the low statistical power or extreme homogeneity of jobs/occupations in the sample.

Our findings clearly cast doubt upon the proposition that a person’s reaction to a particular job characteristic or psychological state is jointly moderated by his or her desire for challenging work and satisfaction with a focal contextual aspect of the work environment. These findings are consistent with those reported by Oldham et al. (1976), which found that GNS and each of the four context satisfactions did not jointly moderate the relation between the overall motivating potential of a job (MPS) and internal motivation (neither overall job satisfaction nor growth satisfaction were examined). However, it is important to point out that Oldham et al. also reported finding positive support for the joint moderating effects of these individual difference factors on MPS-performance relations. This finding suggests that future research may wish to replicate the present study by including performance indices as relevant outcome variables.

To the authors’ knowledge, the present study appears not only to be the first to examine the moderating effects of context satisfactions on job characteristics-psychological states and psychological states-outcome relations, it also seems to be the only large-sample based study to examine the moderating effects of these variables on relations between individual job characteristics and outcome variables. Nevertheless, the trivially small increments in |R.sup.2~ attributed to Job Characteristics x Context Satisfactions interactions found in this study are similar in magnitude to those found by relatively large-sample based studies (Bottger & Chew, 1986; Champoux, 1981) that focused on the relations between an index of overall job complexity (either MPS or an unweighted composite of the five characteristics) and work outcomes. In sum, our results not only reinforce the conclusion of Kulik et al. (1987) that “context satisfaction appears not to moderate the job characteristics-outcome relationship” (287), but also question whether any of the four context satisfactions moderate the job characteristics-psychological states and psychological states-outcomes linkages.

It should be emphasized, however, that this conclusion is limited to operationalizing selected features of the work context in terms of satisfaction with those features. As Ferris and Gilmore (1984) have asserted, this method of operationalization may not be the optimal way to evaluate the moderating effects of aspects of the work context. Indeed, from the perspective that human beings are limited information processors, an anonymous reviewer argued that Hackman and Oldham (1980: 89) were incorrect to assume that one’s assessments of context satisfactions are independent of the three work outcomes examined here. Paralleling the proposition of James and James (1989) that a general factor underlies perceptual measures of the work environment, the reviewer’s argument suggests the possibility that these affective measures may be linked through a common higher-order factor representing an individual’s overall affective orientation to his or her work (see also Hogan & Martell, 1987). The magnitudes of the correlations among the affective variables in the present study seem to support this argument. On the other hand, these intercorrelations may simply be attributable to common method variance. It is unfortunate, but the limitation of our database precluded partitioning this shared variance into that which is due to a methodological artifact and that which is the result of a theoretically meaningful higher-order latent factor. Resolution of this issue thus requires further study. Also, future researchers may benefit from Ferris and Gilmore’s (1984) recommendation to examine the moderating effects of work context using measures other than contextual satisfactions, such as scales measuring dimensions of psychological climate.

The present study is only one of a few that have examined GNS as a moderator of the relations between job characteristics and psychological states and between psychological states and work outcomes. In comparison to these other studies, our study possessed several methodological and analytical strengths. In particular, the analytic procedures used by Hackman and Oldham (1976) and Hogan and Martell (1986) have been severely critiqued by Arnold and House (1980) and Evans (1991), respectively. Moreover, though Arnold and House (1980) used hierarchical moderated multiple regression procedure, they did not take into account the intercorrelations among the criteria. Also, the fact that sample size in that study was less than 100 leaves open the possibility that the obtained findings were simply the result of sampling fluctuations.

With respect to GNS as a moderator of job characteristics-outcome relations, the results of this study are consistent with the conclusion reached by Graen et al. (1986) in their narrative review of this literature. Although two recent meta-analyses (Loher, Noe, Moeller, & Fitzgerald, 1985; Spector, 1985) suggested that GNS may moderate zero-order correlations between job characteristics and psychological or affective outcomes, these findings were disputed by a more recent meta-analysis by Fried and Ferris (1987). The difference between the conclusions reached by the two earlier meta-analytic studies and that of Fried and Ferris may lie in the fact that the latter study was based on a larger sample of studies and controlled for at least one additional statistical artifact. However, because the data used in the present study were part of the database analyzed in Fried and Ferris, we cannot cite their study as supporting our results.

Finally, the presence of correlated disturbance terms among the psychological states and among the work outcomes suggests that greater attention should be paid to the relations among the psychological states and among the work outcomes. These findings indicate that the focal criteria may be related to each other because the criteria are a function of one another (e.g., internal motivation may be a cause of overall job satisfaction and growth satisfaction) and/or the criteria are related to a common omitted variable or variables (see Hayduk, 1987). With respect to the latter interpretation, the omitted variable may be a theoretically meaningful construct excluded from the JCM (e.g., role stress and leadership behavior), a higher-order factor (e.g., an overall affective orientation to one’s job may underlie the work outcomes examined here), and/or a methodological artifact attributable to the exclusive use of self-report measures. Because a fair competitive evaluation of these alternative interpretations is beyond the capabilities of our database, future researchers may wish to further explore this issue.

In conclusion, as damaging to the JCM as these results seem to be, our confidence in them requires collaborative support from studies using more sophisticated research designs (e.g., longitudinal field studies and field experiments) and multiple operationalizations of the focal constructs. Future research should also include other relevant work outcomes, in particular indices of performance. In fact, the meta-analytic results of Fried and Ferris (1987) indicated that there was sufficient residual (i.e., unaccounted for) variance in the corrected zero-order correlations of both job characteristics and psychological states with performance outcomes to justify the search for moderators of these relations.

References

Abdel-Halim, A. A. 1979. Individual and interpersonal moderators of employee reactions to job characteristics: A reexamination. Personnel Psychology, 32: 121-137.

Aldag, R. J., Barr, S. H., & Brief, A. P. 1981. Measurement of perceived task characteristics. Psychological Bulletin, 90:415-431.

Aldag, R. J., & Brief, A. P. 1979. Examination of a measure of higher-order need strength. Human Relations, 32:705-718.

Arnold, H. J., & House, R. J. 1980. Methodological and substantive extensions to the job characteristics model of motivation. Organizational Behavior and Human Performance, 25: 161-183.

Bentler, P. M., & Bonnett, D. G. 1980. Significance tests and goodness-of-fit in the analysis of covariance structures. Psychological Bulletin, 88: 588-606.

Bottger, P. C., & Chew, I. K-H. 1986. The job characteristics model and growth satisfaction: Main effects of assimilation of work experience and context satisfaction. Human Relations, 39: 575-594.

Champoux, J. E. 1981. The moderating effect of work context satisfactions on the curvilinear relationship between job scope and affective response. Human Relations, 34: 503-515.

Cohen, J. 1977. Statistical power analysis for the behavioral sciences (Rev. ed.). New York: Academic Press.

Cohen, J., & Cohen, P. 1983. Applied multiple regression/correlation analysis for the behavior sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Associates.

Evans, M. G. 1991. The problem of analyzing multiplicative composites: Interactions revisited. American Psychologist, 46: 6-15.

Ferris, G. R., & Gilmore, D. C. 1984. The moderating role of work context in job design research: A test of competing models. Academy of Management Journal, 27: 885-892.

Fried, Y., & Ferris, G. R. 1987. The validity of the job characteristics model: A review and meta-analysis. Personnel Psychology, 40: 287-332.

Graen, G. B., Scandura, T. A., & Graen, M. 1986. A field experimental test of the moderating effects of growth need strength on productivity. Journal of Applied Psychology, 71: 484-491.

Hackman, J. R., & Lawler, E. E. 1971. Employee reactions to job characteristics. Journal of Applied Psychology Monograph, 55: 259-286.

Hackman, J. R., & Oldham, G. R. 1975. Development of the Job Diagnostic Survey. Journal of Applied Psychology, 60: 159-170.

Hackman, J. R., & Oldham, G. R. 1976. Motivation through the design of work: A test of a theory. Organizational Behavior and Human Performance, 16: 250-279.

Hackman, J. R., & Oldham, G. R. 1980. Work redesign. Reading, MA: Addison-Wesley.

Hayduk, L. A. 1987. Structural equation modeling with LISREL: Essentials and advances. Baltimore, MD: The John Hopkins University Press.

Hogan, E. A., & Martell, D. A. 1987. A confirmatory structural equations analysis of the job characteristics model. Organizational Behavior and Human Decision Processes, 39: 242-263.

James, L. A., & James, L. R. 1989. Integrating work environment perceptions: Explorations into the measurement of meaning. Journal of Applied Psychology, 74: 739-751.

James, L. R., Muli ak, S. A., & Brett, J. M. 1982. Causal analysis: Assumptions, models and data. Beverly Hills, CA: Sage.

Joreskog, K. G., & Sorbom, D. (1989). LISREL 7 user’s reference guide. Mooresville, IN: Scientific Software Inc.

Katerberg, R., Hom, P. W., & Hulin, C. L. 1979. Effects of job complexity on the reactions of part-time employees. Organization Behavior and Human Performance, 24: 317-332.

Kenny, D. A., & Judd, C. M. 1984. Estimating the nonlinear and interactive effects of latent variables. Psychological Bulletin, 96: 201-210.

Kulik, C. T., Oldham, G. R., & Hackman, J. R. 1987. Work design as an approach to person-environment fit. Journal of Vocational Behavior, 31: 278-296.

Lance, C. E. 1988. Residual centering, exploratory and confirmatory moderator analysis, and decomposition of effects in path models containing interactions. Applied Psychological Measurement, 12: 163-175.

Loher, B. T., Noe, R. A., Moeller, N. L., & Fitzgerald, M. P. 1985. A meta-analysis of the relation of job characteristics to job satisfaction. Journal of Applied Psychology, 70: 280-289.

Marsh, H. W., Balla, J. R., & McDonald, R. P. 1988. Goodness-of-fit indexes in confirmatory factor analysis: The effects of sample size. Psychological Bulletin, 103: 391-410.

O’Brien, G. E. 1982. Evaluation of the job characteristics theory of work attitudes and performance. Australian Journal of Psychology, 34: 383-401.

O’Connor, E. J., Rudolf, C. J., & Peters, L. H. 1980. Individual differences and job design reconsidered: Where do we go from here? Academy of Management Review, 5: 249-254.

Oldham, G. R. 1976. Job characteristics and internal motivation: The moderating effect of interpersonal and individual variables. Human Relations, 29: 559-569.

Oldham, G. R., Hackman, J. R., & Pearce, J. L. 1976. Conditions under which employees respond positively to enriched work. Journal of Applied Psychology, 61: 395-403.

Oldham, G. R., Hackman, J. R., & Stepina, L. P. 1979. Norms for the Job Diagnostic Survey. JSAS Catalog of Selected Documents in Psychology, 9, 14 (MS. No. 1819).

Orpen, C. 1979. The effects of job enrichment on employee satisfaction, motivation, involvement, and performance: A field experiment. Human Relations, 32: 189-217.

Pokorney, J. J., Gilmore, D. C. & Beehr, T. A. 1980. Job diagnostic survey dimensions: Moderating effect of growth needs and correspondence with dimensions of job rating form. Organizational Behavior and Human Performance, 26: 222-237.

Roberts K. H., & Glick, W. 1981. The job characteristics approach to task design: A critical review. Journal of Applied Psychology, 66: 193-217.

Sims, H. P., & Szilagyi, A. D. 1976. Job characteristics relationships: Individual and structural moderators. Organizational Behavior and Human Performance, 17: 211-230.

Spector, P. E. 1985. Higher-order need strength as a moderator of the job scope-employee outcome relationship: A meta-analysis. Journal of Occupational Psychology, 58: 119-127.

Stone, E. F. 1988. Moderator variables in research: A review and analysis of conceptual and methodological issues. In K. Rowland & G. R. Ferris (Eds.), Research in personnel and human resources management: 191-229. Greenwich CT: JAI Press.

Taber, T. D., & Taylor, E. 1990. A review and evaluation of the psychometric properties of the job diagnostic survey. Personnel Psychology, 43: 467-500.

Tetrick, L. E., & LaRocca, J. M. 1987. Understanding, prediction, and control as moderators of the relationships between perceived stress, satisfaction, and psychological well-being. Journal of Applied Psychology, 72: 538-543.

Tucker, L. R., & Lewis, C. 1973. The reliability coefficient for maximum likelihood factor analysis. Psychometrika, 38: 1-10.

Williams, L. J., Cote, J. A., & Buckley, M. R. 1989. Lack of method variance in self-reported affect and perceptions at work: Reality or artifact? Journal of Applied Psychology, 74: 462-468.

COPYRIGHT 1992 JAI Press, Inc.

COPYRIGHT 2004 Gale Group