CURRENT RESEARCH IN SOCIAL PSYCHOLOGY
Volume 5, Number 12
Submitted: March 22, 2000
Resubmitted: May 11, 2000
Accepted: May 22, 2000
Publication date: May 22, 2000
Submitted: March 22, 2000
Resubmitted: May 11, 2000
Accepted: May 22, 2000
Publication date: May 22, 2000
THE IMPORTANCE OF THE CRITICAL PSYCHOLOGICAL STATES IN THE JOB CHARACTERISTICS MODEL: A META-ANALYTIC AND STRUCTURAL EQUATIONS MODELING EXAMINATION
Scott J. Behson
Fairleigh Dickinson University
Fairleigh Dickinson University
Erik R. Eddy
The Group for Organizational Effectiveness, Inc.
The Group for Organizational Effectiveness, Inc.
Steven J. Lorenzet
Rider University
Rider University
ABSTRACT
Hackman and Oldham (1976) originally proposed
their Job Characteristics Theory as a three-stage model, in which
a set of core job characteristics impact a number critical
psychological states, which, in turn, influence a set of
affective and motivational outcomes (see Figure 1).
Interestingly, most subsequent research has omitted the critical
psychological states, focusing, instead, on the direct impact of
the core job characteristics on the outcomes (i.e., a two-stage
model). Meta-analytic data from the thirteen studies that have
investigated the full, three-stage Job Characteristics Model was
used as input into a structural equations modeling analysis
(Viswesvaran & Ones, 1995) to examine competing versions of
the Job Characteristics Model and to determine the importance of
the critical psychological states. Results suggest that, while
the two-stage model demonstrates adequate fit to the data,
information on the critical psychological states is important for
both theoretical and practical reasons.
[170]
---------------
[171]
---------------
[171]
Figure 1. Hackman &
Oldham’s (1976) Job Characteristics Model
RESEARCH ON THE JOB CHARACTERISTICS MODEL
Hackman and Oldham’s (1975, 1976, 1980) Job
Characteristics Model (JCM) is one of the most influential
theories ever presented in the field of organizational
psychology. It has served as the basis for scores of studies and
job redesign interventions over the past two decades, and this
research has been extensively reviewed (Fried & Ferris 1987;
Loher, Noe, Moeller & Fitzgerald, 1985; Taber & Taylor,
1990). The majority of research has supported the validity of the
JCM, although critiques and modifications have been offered
(Roberts & Glick, 1981; Salancik & Pfeffer, 1978).
[171]
---------------
[172]
---------------
[172]
Interestingly, an evaluation of the research that
has been conducted on the JCM suggests that few researchers have
tested the model the way in which it was originally proposed.
According to Hackman and Oldham (1976, 1980), the critical
psychological states (CPS) make up the "causal core of the
model" and should fully mediate the effects of the core job
characteristics (CJC) on relevant individual outcomes. Hackman
and Oldham developed the model by identifying psychological
states important for job satisfaction and motivation, and then
worked backwards to identify job characteristics that could
elicit these psychological states. Thus, the model is centered
around the critical psychological states, and "the core job
characteristics were identified to serve the critical
psychological states, not the other way around" (Johns, et
al., 1992, p. 658).
Although much of the earliest research into the
validity of the JCM (e.g., Arnold & House, 1980; Wall, Clegg,
& Jackson, 1978) explicitly examined all of the linkages
within the JCM, most subsequent investigations have omitted the
CPS, and have instead investigated only the direct relationships
between the CJC and a number of outcomes. "One of the most
critical gaps in JCM research involves how infrequently the total
model has been tested . . . the rarity of studies that
incorporate the mediating psychological states is
remarkable" (Johns, et al., 1992, p. 658). Further,
"since few studies have included the CPS, one could question
whether the motivational underpinnings of this theory have been
adequately examined or represented in JCM evaluations" (Renn
& Vandenberg, 1995, p. 280).
The omission of the CPS from JCM investigations
could be warranted if there were theoretical or practical
rationale for this practice. However, "virtually no
empirical evidence has accumulated supporting the practice of
excluding the CPS from tests of the theory. The practice of
excluding the mediating role appears to have occurred without
empirical or theoretical justification" (Renn &
Vandenberg, 1995, p. 280; see also Fried & Ferris, 1987;
Hogan & Martel, 1987).
[172]
---------------
[173]
---------------
[173]
Most importantly, the omission of the CPS from
empirical investigations of the JCM could lead to erroneous
predictions (Fox & Feldman, 1988). For example, the fact that
skill variety has been found to be positively correlated with job
satisfaction could lead practicing managers to believe that
satisfaction can be improved simply by increasing this CJC.
However, according to the JCM, skill variety should only lead to
positive outcomes to the extent that this increase results in a
corresponding increase in experienced meaningfulness of the work.
If an increase in variety does not result in increased feelings
of meaningfulness, it is reasonable to hypothesize that this
would result in a negative or non-significant change in
satisfaction. The increased variety might only reflect more
boring, meaningless things to do. In short, without measuring the
CPS, our understanding of how CJC affect work outcomes can be
incomplete or misleading. Due to the prominence of the JCM, the
lack of data regarding the relationships between the CPS and the
other elements of the JCM can have far-reaching consequences.
Further, this lack of available data has
prevented the major meta-analytic reviews of the JCM from making
definitive statements about the CPS. While Fried and Ferris
(1987) included 76 studies in their meta-analysis of the JCM,
they could find only eight studies that examined the entire JCM
(i.e., including the CPS) and only three that tested the
mediating effects of the CPS. Thus, Fried and Ferris (1987) were
unable to make definitive conclusions as to the validity or
importance of the CPS, although they stated in their qualitative
discussion that there was suggestive evidence that the CPS are
critical to the model. The Loher et al. (1985) meta-analysis did
not address the critical psychological states at all. Rather, it
focused solely on the relationships between the CJC and
satisfaction. Thus, despite over two decades of active research
on the JCM, the there has yet to be a comprehensive statement
made concerning the role of the CPS in the JCM, and there has yet
to be a quantitative review of the JCM examining all the
relationships within the JCM.
[173]
---------------
[174]
---------------
[174]
Recently, however, several researchers have
called for, and conducted research on, the full JCM model, with
particular emphasis on the CPS. In general, these more recent
studies have utilized sophisticated analytic techniques such as
structural equations modeling, as opposed to bivariate
correlation analysis. While the results and conclusions of these
investigations have varied, there is general consensus that (a)
the original JCM represents an adequate, but imperfect model, (b)
the inclusion of the CPS in the investigation of the JCM explains
additioanl variance in the outcome measures, and (c) that the CPS
may represent partial, not complete, mediators of the CJC-outcome
relationships. Due to the renewed interest in examining the CPS,
we feel that there are a sufficient number of studies to warrant
a summary analysis. Thus, the goals of this paper are to: (a)
quantitatively summarize the findings of all existing studies
which have examined the complete JCM, (b) test the adequacy of
the original Hackman and Oldham model against the more commonly
researched two-stage model, and (c) provide evidence to judge the
importance of the CPS to the JCM.
The two competing models tested in this study
are: (1) The original Job Characteristics Model, as proposed by
Hackman and Oldham (1976) and (2) A modified JCM in which the
critical psychological states are omitted. The original model
will be tested to provide a test of the adequacy of the original
model among the studies that have measured the JCM in its
entirety. It is expected that the original model will provide an
adequate fit for the data. The modified model represents the vast
majority of studies that have measured the links between CJC and
outcomes, while omitting the intervening CPS. It is expected that
this model will not be as adequate as the models that encompass
all three stages of the JCM (Renn & Vandenberg, 1995; Hogan
& Martel, 1987). Please note that moderator variables, such
as Growth Need Strength, were not incorporated into the tested
models. This decision is discussed later in the paper.
The present study utilizes both meta-analytic and
structural equation modeling techniques (see Viswesvaran &
Ones, 1995) to provide a comprehensive test of the JCM based on
the collected results of past research. "Another need for
future research is to continue to utilize structural equation
modeling to analyze data already collected. Numerous JCM data
sets have been analyzed with less sophisticated techniques; such
data could be re-analyzed using causal modeling. . . . The
resulting group of analyses, taken as a whole, might then be
subjected to meta-analysis" (Hogan & Martel, 1987; p.
261-2).
[174]
---------------
[175]
---------------
[175]
This approach for studying the JCM seems
appropriate for several reasons (Hogan & Martel, 1987).
First, structural equations modeling is appropriate for testing
competing interpretations of the same model. Second, structural
equations modeling can handle the simultaneous and multiple-stage
nature of the mediated job characteristics model better than
traditional regression analytic techniques. Further, the use of
meta-analytic data also helps us avoid problems such as small
sample size, low power, and homogeneous samples of jobs and
organizations.
In addition, our analysis has been able to avoid
the most common concerns that have been expressed regarding the
use of the procedures as laid out in Viswesvaran and Ones (1995).
First, the use of meta-analytic input could lead to vastly
different sample sizes for each cell in the input matrix. This
does not appear to be a problem for the current analysis because
all values were gathered from meta-analytic samples ranging from
8,016 to 8,964 individual subjects.
Second, some are concerned that widely discrepant
operationalizations could be combined as indicators of the same
latent variable. All of the studies included in the meta-analysis
used the measurement scales from the Job Diagnostic Survey (JDS)
(see Hackman & Oldham, 1975, 1980), obviating this concern.
Finally, some researchers caution that the use of these
procedures could result in a correlation matrix in which there
are missing values. In this analysis, there are no missing values
in the meta-analytic correlation matrix.
METHOD
Relevant studies were gathered through a variety
of sources: (a) a computer-based search of JCM keywords using Psychlit
and Dissertation Abstracts dating back to 1976, (b) a
reference list search of found articles and existing JCM
meta-analyses, and (c) a hand search of five prominent
organizational psychology/management journals (Academy of
Management Journal, Journal of Applied Psychology, Journal of
Management, Organizational Behavior and Human Decision
Processes/Human Performance, and Personnel Psychology),
from 1976 to 1998. The literature search yielded a total of
thirteen independent studies appropriate for inclusion in the
meta-analysis. Inclusion criteria for studies were (a) the study
must contain information regarding the full JCM, including the
CPS, and (b) the study must report correlations between CPS and
CJC and/or outcome measures.
[175]
---------------
[176]
---------------
[176]
Studies were divided among the three authors and
coded independently. To insure reliability, articles were divided
again and re-coded by a different author. Disagreements were
resolved by discussion. Table 1 provides a list of all the
studies included in the meta-analysis, their sample size, sample,
measure used, and whether the study supports the importance of
the CPS in the JCM. Please note that no study that explicitly
examined the CPS found them to be entirely unimportant to the JCM
model.
Table 1. Characteristics
of Studies Included in the Meta-Analysis
| N | Samples | Measures | Support for CPS? | |
| Arnold & House (1980) | 120 | Engineers | JDS | Did Not Test |
| Barnabe & Burns (1994) | 247 | Teachers | JDS | Yes |
| Becherer, Morgan, & Lawrence (1982) | 211 | Sales | JDS | Yes |
| Champoux (1991) | 247 | State Agency | JDS | Partial |
| Fox & Feldman (1988) | 119 | Variety of Jobs | JDS/JDI | Partial |
| Griffeth (1985) | 76 | Work Study | JDS | Did Not Tex |
| Hackman & Oldham (1975) | 658 | Variety of Jobs | JDS | Partial |
| Hogan & Martell (1987) | 208 | NAVY-Variety of Jobs | JDS | Yes |
| Johns, Xie, & Fang (1992) | 300 | Managers | JDS | Yes |
| Kiggundu (1980) | 138 | Financial Company | JDS | Did Not Test |
| Renn & Vandenberg (1995) | 188 | Variety of Jobs | JDS/JDS-R | Yes |
| Tiegs, Tetrick, & Fried (1992) | 6405 | Variety of Jobs | JDS | Did Not Test |
| Wall, Clegg, & Jackson (1978) | 47 | Sales | JDS | Partial |
Studies were coded for three potential moderator
variables: sample type (white collar, blue collar, mixed),
research design (experiment, quasi-experiment, non-experiment),
and instrument used (JDS, JDS-Revised, other). The analyses for
type of sample revealed no consistent pattern of differences.
Analyses were not conducted for the other two variables, due to
the lack of variation among primary studies.
[176]
---------------
[177]
---------------
[177]
The meta-analytic correlations between each of
the elements are displayed in Table 2. Each of the effect sizes
were based upon between nine and thirteen independent samples and
upon between 8,016 and 8,964 participants. The mean sample size
of each of the studies included in the meta-analysis was 690 and
the median sample size was 208. Effect sizes were not corrected
for unreliability at this stage of the analysis. This correlation
matrix was transformed into a covariance matrix using the
standard deviations calculated by Oldham, Hackman, and Stepina
(1979), which are based on 6,930 respondents from 876 different
jobs in 56 organizations and were previously used to represent
population parameters by Arnold and House (1980), Fried and
Ferris (1987) and Hackman and Oldham (1980). The reader should
note that the standard deviations used in this analysis are based
on normative data, and were not meta-analytically derived from
the included studies.
Table 2. Meta-Analytic
Correlations and Mean Reliabilities
| SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | |
| 1. Skill Variety | 1.57 | .70 | ||||||||||
| 2. Task Significance | 1.25 | .41 | .59 | |||||||||
| 3. Task Identity | 1.44 | .22 | .20 | .65 | ||||||||
| 4. Autonomy | 1.39 | .43 | .32 | .32 | .67 | |||||||
| 5. Feedback | 1.34 | .35 | .34 | .26 | .39 | .71 | ||||||
| 6. Experienced Meaningfulness | 1.14 | .46 | .45 | .24 | .42 | .38 | .75 | |||||
| 7. Experienced Responsibility | 0.96 | .34 | .33 | .27 | .39 | .34 | .59 | .71 | ||||
| 8. Knowledge of Results | 1.14 | .16 | .23 | .28 | .29 | .49 | .40 | .34 | .72 | |||
| 9. Satisfaction | 1.07 | .35 | .29 | .22 | .42 | .36 | .65 | .49 | .42 | .80 | ||
| 10. Growth | 1.15 | .50 | .38 | .26 | .54 | .44 | .65 | .51 | .40 | .69 | .81 | |
| 11. Internal Satisfaction | 0.77 | .35 | .33 | .17 | .30 | .42 | .57 | .59 | .25 | .43 | .50 | .69 |
Note. Mean
reliabilities are reported on the diagonal.
Note. All 95% confidence intervals did not include zero.
Note. Standard deviations from Oldham, Hackman & Stepina (1979)
Note. All 95% confidence intervals did not include zero.
Note. Standard deviations from Oldham, Hackman & Stepina (1979)
[177]
---------------
[178]
---------------
[178]
Next, the procedures outlined by Viswesvaran and
Ones (1995) for using meta-analysis to create a covariance matrix
to be used as input to a structural equations analysis were
employed. The seven-step process is shown in Table 3. Similar
procedures have been employed by Carson, Carson, and Rowe (1993),
Horn, Caranikas-Walker, Prussia and Griffeth, (1992), and Premack
and Hunter (1988), among others. Our meta-analysis is consistent
with these procedures, except that (a) a LISREL 8.0 analysis was
performed instead of traditional path analysis and (b) the
correlations used in the analysis were not corrected for
attenuation due to unreliability. This decision will be discussed
later in the paper.
Table 3. Steps for
Combining Psychometric Meta-Analysis
and Structural Equations Modeling
Measurement
Model
|
|
| Causal Model |
|
Note. Adapted
from framework presented by Viswesvaran and Ones (1995).
[178]
---------------
[179]
---------------
[179]
For the LISREL 8.0 analyses, the parameter
estimates were based on a sample covariance matrix and a maximum
likelihood solution. The median sample size, 208, was used in
this stage of the analysis because the X2 statistic is
biased against large sample sizes (Jaccard & Wan, 1996).
The fit of the data to the model was assessed
using several indices, including: the X2 statistic, the
Goodness of Fit Index (GFI), the Root Mean Square Error of
Approximation (RMSEA), and the Comparative Fit Index (CFI). The X2 statistic, and
the GFI are indices of absolute fit which measure how far the
model deviates from a model of perfect fit. The CFI is an index
of comparative fit that measures how far a model deviates from a
model of good fit. The RMSEA is a test of parsimony that takes
the number of paths into account when determining fit. Model
adequacy is also assessed by examining the amount of variance
explained in the outcome measures and the ratio of predicted to
significant paths.
The GFI, CFI, and RMSEA statistics are useful for
assessing the fit of the individual models; however, they cannot
be used to compare across models. The X2 statistic can
be used to compare the relative fit of competing models, but only
if these models are nested within each other. However, the two
models being compared in this study are not nested. Therefore,
two commonly used statistical indices, the Akaike Information
Criterion (AIC) and the CIAC (an extension of the AIC, which more
strongly penalizes models for lack of parsimony), were used to
compare these two non-nested models on a common metric. These
statistics are seen as most appropriate when comparing two
non-nested models (see Lin & Dayton, 1997).
RESULTS
First, the original JCM model (Model 1) was
tested (see Table 4). The fit indices for this model were: X2 (25) = 124.25,
p < .05, GFI = .91, RMSEA = .14, and CFI = .89. The CFI
and GFI indicate acceptable levels of model fit, while the RMSEA
and the X2
value are less supportive of good model fit. However, X2 is influenced
by sample size, and the RMSEA index penalizes models for lack of
parsimony. Therefore, these findings are not unexpected.
[179]
---------------
[180]
---------------
[180]
Table 4. Results of Tests
of Goodness of Fit for the Various Models
Statistic
|
c2 | df | Ratio of explained paths | RMSEA | GFI | CFI | Explained variance in DV |
Model AIC
|
Model CIAC |
Rules of
Thumb for "Good Fit"
|
ns | - | - | <.08 | >.90 | >.90 | - | - | - |
1. Original Job
Characteristics Model
|
124.25* | 25 | 12/14 | .14 | .91 | .89 |
.42
sat.
.42 growth .38 mot. |
294.48 | 446.29 |
2. Normally Tested JCM
(excluding CPS)
|
12.09* | 3 | 7/15 | .16 | .99 | .98 |
.37
sat
.43 growth .32 mot |
80.09 | 227.56 |
Note. * indicates
result was statistically significant at p < .05
Figure 2 shows the estimates of the structural
coefficients for Model 1. Standardized estimates appear on each
path. Twelve of the fourteen paths in this model were
statistically significant, and the variables in the model were
able to account for approximately 42% of the variance in
satisfaction, 42% of the variance in growth satisfaction, and 38%
of the variance in motivation.
Figure 2. SEM of the
Original JCM
[180]
---------------
[181]
---------------
[181]
Next, the two-stage model normally tested in the
literature was explored (Model 2). The results of the goodness of
fit indices were: X2
(3) = 12.09, p < .05, GFI = .99, RMSEA = .16, and CFI =
.98. All of these values, except for the RMSEA, indicate good
model fit. Seven of the fifteen paths were statistically
significant in this model (see Figure 3). The model was able to
account for approximately 37% of the variance in satisfaction,
43% of the variance in growth satisfaction, and 32% of the
variance in motivation.
Figure 3. SEM of the JCM
Normally Tested in the Literature
[181]
---------------
[182]
---------------
[182]
In short, the original JCM can be seen to (a)
explain more variance in the dependent variables, and (b) have a
greater percentage of statistically significant causal pathways
than the abridged version of the JCM. The two-stage JCM, however,
attained greater model fit, as indicated by the GFI, CFI and
chi-squared indices. Neither model showed an acceptable level of
parsimony according to the RMSEA index.
Finally, in order to compare the models with a
common metric, the AIC and CIAC statistics were used. When
comparing two or more models, the model of best fit is the one
with the lowest values (Lin & Dayton, 1997). Both the AIC and
the CAIC indicate that the normally tested two-stage model
demonstrates superior fit (see Table 4).
DISCUSSION
The quantitative results of this analysis suggest
that the two-stage model normally tested in the literature may
provide a better fit to the available data than the three-stage
model originally proposed by Hackman and Oldham (1976). However,
adequate comparison among competing models requires more than
comparing fit ratios. The reasonableness of values contained in a
model and a model’s correspondence with relevant theory are
equally, if not more, important. Thus, while the two stage model
may result in more adequate model fit, a closer examination of
the two models support, rather than refute, the contention that
the CPS are indeed critical to the JCM.
Several path coefficients in Model 2 run counter
to well-established theory regarding the design of work. In
particular, eight of the nine paths between skill variety, task
significance, and task identity and the three outcome variables
are not statistically significant (see Figure 2). In comparing
these path coefficients with those of Model 1, the importance of
the CPS to the JCM becomes clear. In Model 1, both skill variety
and task significance demonstrate statistically significantly
positive indirect relationships with the outcome variables, as
mediated by experienced meaningfulness. These relationships
provide evidence that, while skill variety and task significance
may not be directly related to job affect and motivation, they
can be important in eliciting experienced meaningfulness of the
work. It is this psychological state, however, that is crucial
for the beneficial outcomes of job redesign. Thus, the comparison
between the path coefficients in these two competing models
accentuates the importance of the CPS to job redesign. The
non-significant paths in Model 2 provide evidence that increasing
job characteristics may have little or no impact if the employee
does not experience the CPS. This underscores the importance of
the CPS as the "causal core of the model" (Hackman
& Oldham, 1976, p. 255).
[182]
---------------
[183]
---------------
[183]
Our results also lead to several other
interesting observations. For instance, in both of the competing
models, autonomy is the CJC with the strongest relationships with
outcome variables. This finding is consistent with several recent
streams of research into work motivation, including Ajzen’s
(1991) Theory of Planned Behavior and Deci and Ryan’s
Cognitive Evaluation Theory (e.g., Deci & Ryan, 1991), which
stress the importance of autonomy and self-determination.
Further, recent practitioner-oriented research on organizational
development and change has established that allowing personal
control is a key to successful change in employee attitudes,
behaviors, and value orientation (e.g., Parker, Wall &
Jackson, 1997).
In addition, it should be noted that neither
model tested in this study demonstrated exceptional fit to the
data. It was certainly expected that the JCM, in either form,
would not be particularly parsimonious. However, this study does
provide some suggestions for avenues of future research. In
particular, research aimed at trimming the model and balancing
parsimony and variance explanation concerns is clearly warranted.
Again, autonomy is seen as a particularly crucial construct for
this purpose.
The limitations of the present study also warrant
discussion. First, the meta-analytic data was derived from only
13 primary studies, and some have argued that this relatively low
k could lead to unstable meta-analytic results (Oswald &
Johnson, 1998). However, this number of primary studies is not
uncommonly low, given recent publications (e.g., Donovan &
Radosevich, 1998). Further, our data was derived from a large
number of subjects (n varied from 8,016 to 8,964) across a wide
variety of occupations and job settings. Thus, one can be
reasonably confident in the external validity of our results.
Another potential criticism of this research is
that a large proportion of our sample was derived from one
primary study (Tiegs, et al., 1992). To address this concern, we
ran our analyses both with and without this study included in our
sample, and found no significant differences. In fact, in
comparing the two resultant correlation matrices, only one of the
fifty-five pairs of correlations differed by more than .05.
[183]
---------------
[184]
---------------
[184]
The combined use of meta-analysis and SEM is a
relatively new analytic strategy. While this technique is
promising, its validity hinges on a number of statistical
assumptions. Efforts were made to address some commonly voiced
concerns regarding this technique. However, more psychometric and
simulation-based research regarding the limits and potential
drawbacks of this approach is clearly needed.
In addition, we did not include any information
on moderating variables, such as Growth Need Strength (GNS), in
our analysis. This decision was made for several reasons,
including: (a) the fact that few of the studies selected for our
meta-analysis included information on GNS, (b) Tiegs, Tetrick
& Fried (1992) offer compelling evidence that GNS is not, in
fact, a significant moderator of the relationships in the model,
(c) that the analysis of the GNS moderator in the manner
originally proposed by Hackman and Oldham (moderation at two
stages) is troublesome and would either require the addition of
14 additional paths to Model 1 or the splitting of continuous
variables into categorical ones (Jaccard & Wan, 1996), and
(d) the effects of moderators are tangential to the specific
purpose of the present paper.
Finally, the correlations used as input to the
structural equations analysis were not corrected for
unreliability at either the meta-analytic stage or the SEM stage,
although techniques for such corrections are commonly employed.
There were two reasons for this decision. First, research on the
JCM and the JDS have long acknowledged that common method
variance and multicollinearity serve to inflate the correlations
among the JCM constructs (Roberts & Glick, 1981; Taber &
Taylor, 1990). While unreliability serves to attenuate
correlations, correcting for this attenuating effect while
ignoring the factors which serve to artificially inflate variable
correlations would result in biased correlations which overstate
the strength of the relationships among the JCM variables.
Second, when the analyses were conducted using corrected
correlations as input, several statistical problems were
encountered. In particular, the inflated correlations led to
suppressor effects among the independent variables in Model 2
(the abridged model). This led to several statistically troubling
results, including a standardized path coefficient greater than
1.0 (1.41 between autonomy and satisfaction) and negative causal
paths between variables whose zero-order correlations are
positive.
[184]
---------------
[185]
---------------
[185]
Two potential causes of these supressor effects
are that the average reliabilities calculated from the primary
studies were consistently lower than acceptable standards for
scale reliability (along the diagonal in Table 2), and that
multicollinearity may exist among the variables in the model. Our
findings are consistent with Roberts and Glick’s (1981) and
Taber and Taylor’s (1990) conclusions that the JDS is a
useful, albeit limited, instrument, but that additional and
alternate measures and methodologies are required in order to
advance the field of job redesign. Thus, due to statistical
anomalies and our desire to remain conservative in our analyses,
no corrections for attenuation were made.
In sum, the central finding of the present
analysis is that, while the abridged two-stage model demonstrates
adequate fit, JCM researchers need to pay more attention to the
CPS. The results of our meta-analysis support recent contentions
that "researchers and practitioners who are interested in
the impact of jobs on employees might consider measuring
psychological states more often than is commonly done"
(Johns, et al., 1992, p. 672). Thus, this paper contributes
quantitative evidence to support those who have criticized how
research has commonly been conducted on the JCM (see Fried &
Ferris, 1987; Fox & Feldman, 1988; Hogan & Martel, 1987;
Renn & Vandenberg, 1995).
Failure to incorporate CPS into the JCM could
lead to unexpected results and misdirected organizational
interventions. This classic theory is quite complex and rich, and
has implications for many of the workplace change initiatives
(e.g., JIT, TQM, MBO) in use in organizations today. Even though
the two-stage model represents a more parsimonious model,
important information may be lost if the CPS are not included.
REFERENCES
Ajzen, I. (1991). 'The theory of planned
behavior." Organizational Behavior and Human Decision
Processes, 43:179-211.
Arnold, H.J. & House, R.J. (1980).
"Methodological and substantive extensions to the job
characteristics model of motivation." Organizational
Behavior and Human Decision Processes, 25:161-183.
[185]
---------------
[186]
---------------
[186]
Barnabe, C. & Burns, M. (1994).
"Teachers’ job characteristics and motivation."
Educational Research, 36:171-185.
Becherer, R.C., Morgan, F.W. & Richard, L.M.
(1982). "The job characteristics of industrial salespeople:
Relationship to motivation and satisfaction." Journal of
Marketing, 46:125-135.
Brief, A.P. & Aldag, R.J. (1975).
"Employee reactions to job characteristics: A constructive
replication." Journal of Applied Psychology,
60:182-186.
Carson, P.P., Carson, K.D. & Roe, C.W.
(1993). "Social power bases: A meta-analytic examination of
interrelationships and outcomes." Journal of Applied
Social Psychology, 23:1,150-1,169.
Champoux, J.E. (1991)."A multivariate test
of the job characteristics theory of work motivation." Journal
of Organizational Behavior, 12:431-446.
Deci, E.L. & Ryan, R.M. (1991).
"Intrinsic motivation and self-determination in human
behavior." In Steers, R.M. & Porter, L.W. (Eds.) Motivation
and Work Behavior, 5th Edition. New York:
McGraw-Hill, Inc., pp. 44-58.
Donovan, J.J., & Radosevich, D.J. (1998).
"The moderating role of goal commitment on the goal
difficulty-performance relationship: A meta-analytic review and
critical reanalysis." Journal of Applied Psychology,
83:308-315.
Evans, M.G., Kiggundu, M.N. & House, R.J.
(1979). "A partial test and extension of the job
characteristics model of motivation." Organizational
Behavior and Human Performance, 24:354-381.
Fox, S. & Feldman, G. (1988). "Attention
state and critical psychological states as mediators between job
dimensions and job outcomes." Human Relations,
41:229-245.
[186]
---------------
[187]
---------------
[187]
Fried, Y. & Ferris, G.R. (1987). "The
validity of the job characteristics model: A review and
meta-analysis." Personnel Psychology, 40:287-322.
Griffeth, R.W. (1985). "Moderation of the
effects of job enrichment by participation: A longitudinal field
experiment." Organizational Behavior and Human Decision
Processes, 35:73-93.
Hackman J.R. & Oldham, G.R. (1976).
"Motivation through the design of work: 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 Publishing Company
Inc.
Hackman J.R. & Oldham, G.R. (1975).
"Development of the job diagnostic survey." Journal
of Applied Psychology, 60:159-170.
Hogan E.L. & Martell, D.A. (1987). "A
confirmatory structural equations analysis of the job
characteristics model." Organizational Behavior and
Human Decision Processes, 39:242-263.
Horn, P.W., Caranikas-Walker, F., Prussia, G.E.
& Griffeth, R.E. (1992). "A meta-analytical structural
equations analysis of a model of employee turnover."
Journal of Applied Psychology, 77:890-909.
Jaccard, J. & Wan, C.K. LISREL Approaches
to Interaction Effects in Multiple Regression. Thousand
Oaks, CA: Sage Publications.
Johns, G., Xie, J.L. & Fang, Y. (1992).
"Mediating and moderating effects in job design."
Journnal of Management, 18:657-676.
Kiggundu, M.N. (1980). "An empirical test of
the theory of job design using multiple job ratings." Human
Relations, 33:339-351.
Lin, T.H & Dayton, C.M. (1997). "Model
selection information criteria for non-nested latent class
models." Journal of Educational and Behavioral
Statistics, 22:249-264.
[187]
---------------
[188]
---------------
[188]
Loher, B.T., Noe, R.A., Moeller, N.L. &
Fitzgerald, M.P. (1985). "A meta-analysis of the relation of
job characteristics and job satisfaction." Journal of
Applied Psychology, 70:280-289.
Oldham, G.R., Hackman, J.R. & Stepina, L.P.
(1979). "Norms for the Job Diagnostic Survey." JSAS
Catalog of Selected Documents in Psychology (Ms. No. 1819),
9, 14.
Oldham, G.R. & Rotchford, N.L. (1983).
"Relationships between office characteristics and employee
reactions." Administrative Science Quarterly,
28:542-556.
Oswald, F.L. & Johnson, J.W. (1998). "On
the robustness, bias, and stability of statistics from
meta-analysis or correlation coefficients: Some initial monte
carlo findings." Journal of Applied Psychology,
83:164-178.
Premack, S.L. & Hunter, J.E. (1988).
"Individual unionization decisions." Psychological
Bulletin, 103:223-234.
Renn, R.W. & Vandenberg, R.J. (1995).
"The critical psychological states: An underrepresented
component in job characteristics model research."
Journal of Management, 21:279-303.
Roberts, K.H. & Glick, W. (19810). "The
job characteristics approach to task design: A critical
review." Journal of Applied Psychology, 66:193-217.
Salancik, G.R. & Pfeffer, J. (1978). "A
social information processing approach to job attitudes and task
design." Administrative Science Quarterly,
23:224-253.
Taber, T.D. & Taylor, E. (1990). "A
review and evaluation of the psychometric properties of the job
diagnostic survey." Personnel Psychology,
43:467-500.
Tiegs, R.B., Tetrick, L.E. & Fried, Y.
(1992). "Growth need strength and context satisfaction and
moderators of the relations of the Job Characteristics
Model." Journal of Management, 18:575-593.
[188]
---------------
[189]
---------------
[189]
Viswesvaran, C. & Ones, D.S. (1995).
"Theory testing: Combining psychometric meta-analysis and
structural equations modeling." Personnel Psychology,
48:865-885.
Wall, T.D., Clegg, C.W. & Jackson, P.R.
(1978). "An evaulation of the job characteristics
model." Journal of Occupational Psychology,
51:183-196.
AUTHOR BIOGRAPHIES
Dr. Scott J. Behson is an Assistant Professor of
Management at Fairleigh Dickinson University, where he teaches,
conducts research, and provides consulting services in
organizational change, organizational behavior and human resource
management. Scott is also a member of the Center for Human
Resource Management (CHRMS) (www.chrms.org)
at FDU. Email: Behson@mailbox.fdu.edu,
Website: www.scottbehson.homestead.com
Dr. Erik Eddy is a Project Director with The
Group for Organizational Effectiveness. His interests include
continuous and organizational learning, informal methods of
knowledge acquisition, and organizational privacy. E-mail: Erik.Eddy@groupOE.com
Dr. Steven Lorenzet is a recent graduate of the
University at Albany, SUNY and will be beginning an appointment
in the fall at Rider University as an Assistant Professor of
Human Resource Management. His interests include training and
development, employee and organizational learning, teams, and
multiple levels of analysis. E-mail: sloren1@worldnet.att.net
[189]
---------------
---------------
No comments:
Post a Comment