Beruflich Dokumente
Kultur Dokumente
To cite this article: Margot Van der Doef & Stan Maes (1999) The Job Demand-Control (-
Support) Model and psychological well-being: A review of 20 years of empirical research,
Work & Stress: An International Journal of Work, Health & Organisations, 13:2, 87-114,
DOI: 10.1080/026783799296084
Taylor & Francis makes every effort to ensure the accuracy of all the information
(the “Content”) contained in the publications on our platform. However, Taylor
& Francis, our agents, and our licensors make no representations or warranties
whatsoever as to the accuracy, completeness, or suitability for any purpose
of the Content. Any opinions and views expressed in this publication are the
opinions and views of the authors, and are not the views of or endorsed by
Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and
Francis shall not be liable for any losses, actions, claims, proceedings, demands,
costs, expenses, damages, and other liabilities whatsoever or howsoever caused
arising directly or indirectly in connection with, in relation to or arising out of the
use of the Content.
This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sub-
licensing, systematic supply, or distribution in any form to anyone is expressly
forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Downloaded by [Laurentian University] at 08:31 01 October 2013
w o r k & s t r e s s , 1999, v o l . 13, n o . 2 87± 114
Keywords : Job Demand-Control model ; Job Demand-Control-Support model ; Job strain ; Psychological
well- being.
The Job Demand-Control (JDC) model (Karasek, 1979) and the Job Demand-Control-Support
(JDCS) model (Johnson, and Hall, 1988) have dominated research on occupational stress in the last
20 years. This detailed narrative review focuses on the JDC(S) model in relation to psychological
well-being. It covers research from 63 samples, published in the period 1979± 1997.In the review a
distinction is drawn between two diVerent hypotheses prevailing in research on the models.
According to the strain hypothesis of the JDC model, employees working in a high-strain job (high
demands± low control) experience the lowest well-being. The buVer hypothesis states that control can
moderate the negative eVects of high demands on well-being. Translating these hypotheses to the
expanded JDCS model, the iso-strain hypothesis predicts the most negative outcomes among
workers in an iso-strain job (high demands± low control± low social support } isolation), whereas the
buVer hypothesis states that social support can moderate the negative impact of high strain on well-
being. Although the literature gives considerable support for the strain and iso-strain hypotheses,
support for the moderating in¯ uence of job control and social support is less consistent. The
conceptualization of demands and control is a key factor in discriminating supportive from non-
supportive studies. Only aspects of job control that correspond to the speci® c demands of a given job
moderate the impact of high demands on well-being. Furthermore, certain subpopulations appear to
be more vulnerable to high (iso)strain, whereas others bene® t more from high control. On the basis
of the results of this review, suggestions for future research and theoretical development are
formulated.
1. Introduction
One of the most in¯ uential models in research on the relationship between work and health
is the Job Demand-Control (JDC) model, also known as the job strain model. The JDC
model identi® es two crucial job aspects in the work situation : job demands and job control
(Karasek, 1979).In the 1980sa social dimension was added to the model (Johnson, and Hall,
1988; Johnson, Hall, and Theorell, 1989), resulting in the Job Demand-Control-Support
(JDCS) model.
In the past 20 years numerous studies have applied the model and its hypotheses to a
broad array of outcomes. Both the model in itself, and the inconsistent results of empirical
tests of the model, have elicited considerable theoretical and methodological criticism
(Kristensen, 1995). Close inspection shows that the empirical tests of the JDC(S) model do
not all examine the same hypothesis. The authors believe that this discrepancy contributes
signi® cantly to the controversy. In this review the various hypotheses that have been
Work & Stress ISSN 0267-8373 print } ISSN 1464-5335 on line ’ 1999 Taylor & Francis Ltd
http :} } www.tandf.co.uk } JNLS } wst.htm http :} } www.taylorandfr ancis.com } JNLS } wst.htm
88 M. Van der Doef and S. Maes
tested are distinguished and the empirical evidence for each hypothesis is discussed
separately.
To date the JDC(S) model has been systematically reviewed only in relation to physical
health outcomes. In their reviews on cardiovascular disease (CVD) end-points, Schnall,
Landsbergis, and Baker (1994) and Kristensen (1995) concluded that epidemiological
studies generally support the model. Recently, Van der Doef, and Maes (1998) suggested
certain re® nements to this conclusion in their review on a broader array of physical health
outcomes, which distinguished various JDC(S) hypotheses. Other authors have discussed
the model from a more sociological, epidemiological or work psychological perspective
(Muntaner, and O’Campo, 1993; Kasl, 1996; Theorell, and Karasek, 1996; de Jonge, and
Kompier, 1997). Here the research on the JDC(S) model and psychological well-being will
be reviewed. As such, this review refers to an area of outcomes which, to the authors ’
knowledge, has not been systematically reviewed so far.
First, the JDC(S) model and its hypotheses will be brie¯ y presented, focusing on two
distinctive perspectives and their formulation of the central hypothesis. Subsequently,
empirical studies applying the model to psychological well-being are reviewed. Fur-
thermore, attention will be given to the outcome under study, diVerentiating general
psychological well-being and job-related well-being. In the discussion, the theoretical and
practical implications of the results are addressed. Finally, issues for further theoretical
development and future research are discussed.
reactions of psychological strain and physical illness are expected in a `high-strain ’ job, that
is the high demands± low control job. On the other hand, following the `learning ’ diagonal,
a second hypothesis states that high demands in combination with high control lead to
increased learning, motivation and development of skills. As this review focuses on
psychological well-being, this learning mechanism will not be elaborated upon further.
Besides the `strain ’ hypothesis, another hypothesis has been examined in research applying
the JDC model to health and well-being. According to this hypothesis control can buVer the
potentially negative eVects of high demands on health and well-being. Karasek himself
examined this interactive eVect of demands and control in his ® rst publication on the model
(Karasek, 1979). In a later publication (Karasek, 1989) he stated that the existence of a
multiplicative interaction term is not the primary issue for the model. In contrast, other
researchers (Ganster, 1989) have maintained the view that the moderating eVect of control
is the main thrust of the model.
If one examines these two points of view in greater detail, it becomes clear that they
Downloaded by [Laurentian University] at 08:31 01 October 2013
`high-strain ’ situation (high demands± low control) experience the lowest levels of
psychological well-being ?
(2) Is there support for the `buVer ’ hypothesis, i.e. an interaction between demands and
control, indicating a buVering eVectof control onthe negative impact of high demands
on psychological well-being ?
Also, rephrasing these questions for the JDCS model : is there support for the `iso-strain ’
hypothesis, such that the lowest level of psychological well-being is experienced by
employees working in an `iso-strain ’ situation (high demands± low control± low support) ?
In addition, is there support for the `buVer ’ hypothesis, i.e. an interaction between
demands, control and support, indicating a buVering eVect of support on the negative
impact of high strain on psychological well-being ?
This review is of a narrative nature, and no meta-analysis has been conducted. Inspection
of the empirical studies on the JDC(S) model and psychological well-being indicated that
the majority of studies use regression analysis, often examining other variables besides the
dimensions of the JDC(S) model (e.g. socio-demographic characteristics, neuroticism). In
most of these studies, the unique contribution of the JDC(S) dimensions cannot be
estimated, a necessity for meta-analysis. As such, it is doubtful whether the studies that
could be included in a meta-analysis would be representative of the whole body of research.
Furthermore, given that four hypotheses would have been examined for various outcomes,
using a limited number of studies, a meta-analysis would probably have given limited and
non-representative results. Speci® c comparisons (e.g. male versus female samples, cross-
sectional versus longitudinal studies) would have been impossible due to the small number
of studies in each category.
A detailed narrative review also has certain advantages. First, it is possible to include
every empirical study, giving a more representative view of the whole body of research.
Second, the methodological strengths and weaknesses of the studies can be taken into
account. Last, but not least, a more inductive approach can be used in identifying the crucial
diVerences between supportive and non-supportive studies.
Studies on the Job Demand-Control(-Support) model were identi® ed by means of a
systematic search of several databases, including PSYCLIT, MEDLINE and SOCIOFILE.
In addition, the reference lists of relevant publications were screened for additional
empirical studies.
This review includes studies that met the following criteria : (1) publication in the period
1979 to 1997; (2) reference to the Job Demand-Control(-Support) model ; (3) inclusion of
at least the two core dimensions of the model, job demands and job control ; (4)
examination of the combined eVect of the demands and control (and support) dimensions ;
(5) examination of subjects in a work setting ; and (6) examination of general psychological
well-being or job-related well-being as an outcome variable. As such, studies focusing on
(psycho)somatic health were not included.
On the basis of these criteria, reports on 63 samples were included in this review. The
following criteria were applied in the categorization of studies as examining the `(iso-)-
strain ’ hypothesis and } or the `buVer ’ hypothesis. Studies examining the `buVer ’
hypothesis were de® ned as those studies that explicitly included interaction eVects between
demands and control for the JDC model, or interaction eVects involving demands, control
and support for the JDCS model, in addition to the main eVects of these constructs. To
examine this hypothesis, most studies used moderated regression techniques. A proper test
The Job Demand-Control(-Support ) Model 91
of the three-way interaction in the JDCS model using this technique would include all
possible two-way interactions as well. Studies examining the `(iso)strain ’ hypothesis were
de® ned as those comparing the high (iso)strain group to a reference group (e.g. the low
(iso)strain group) ; those examining product terms or ratios of demands and control (and
support) as predictors without taking main eVects into account; and those reporting on
additive eVects of demands and control (and support).
3. Review results
The JDC(S) studies on outcomes re¯ ecting employees ’ psychological well-being are
discussed. A distinction is made between (1) general measures of psychological well-being,
such as anxiety and depression, and (2) job-related well-being measures, further speci® ed in
job satisfaction, burnout, and job-related psychological well-being. Table 1 gives a
description of the studies and their ® ndings with respect to the hypotheses. A summary is
Downloaded by [Laurentian University] at 08:31 01 October 2013
presented in table 2. In this table the ratio of supportive studies to the total number of studies
is given for each hypothesis and outcome category.
3.1.1. The strain hypothesis of the Job Demand-Control model : Relatively few studies used a
non-linear approach to examine the strain hypothesis. In these studies the high-strain group
was de® ned and contrasted with the low-strain group or with all non-high-strain groups (6,
9, 30, 31, 33, 34, 58, 63). Most studies report tests on linear additive eVects of demands and
control (3, 4, 7, 8, 9, 10, 12, 13, 14, 17, 19, 23, 24, 25, 27, 29, 30, 31, 32, 33, 35, 36, 37, 40,
41, 42, 44, 45, 46, 47, 49, 51, 52, 57, 58, 59, 60, 62). Karasek (1979) also examined the strain
hypothesis by examining the discrepancy between demands and control in relation to
various outcomes (24).
Downloaded by [Laurentian University] at 08:31 01 October 2013
92
94
Table 1 (cont.)
Results
satisfaction
[continued overleaf
Downloaded by [Laurentian University] at 08:31 01 October 2013
Table 1 (cont.)
96
Results
37 Marshall, Barnett, 326 m social workers and CS SR Psychological distress 6 1 Support strain hypothesis
Baruch, and Pleck, licensed practical nurses Well-being 6 2 dependent on C measure :
1991 for distress decision
authority signi® cant, for
well-being skill discretion
signi® cant
38 McLaney, and Hurrell, 765 Canadian nurses CS SR Job satisfaction 6 2 Additive eVects for nearly all
1988; Hurrell, and combinations of D and C,
McLaney, 1989 except decision control
39 Melamed, Kushnir, 267 m social workers CS SR Job satisfaction 2 2 2 2
and Meir, 1991 Burnout 1 2 1 2
40 Moyle, 1995 143 employees from CS SR Mental health 1 2
diverse worksettings Job satisfaction 1 1
41 Mullarkey, Jackson, 72 operators CS SR Job-related anxiety 2 6 No additive eVects of D and
The Job Demand-Control(-Support ) Model
Table 1 (cont.)
98
Results
Table 1. (cont.)
100
Results
Table 2. Ratio of supportive studies to the total number of studies on each hypothesis of the
JDC(S) model per outcome category.
JDC model JDCS model Total
number
Outcome Strain BuVer Iso-strain BuVer of studies
General psychological well-being 28} 41 15} 31 9} 19 2} 5 43
Job-related well-being
Job satisfaction 18} 30 10} 23 8} 14 2} 6 31
Burnout 3} 4 0} 4 1} 1 0} 2 4
Job-related psychological well-being 7} 8 1} 2 1} 2 1} 1 8
Studies using the non-linear approach almost unanimously found higher psychological
distress in the high-strain jobs (6, 9, 31, 33, 34, 58, 63). The sole exception is LaCroix and
Downloaded by [Laurentian University] at 08:31 01 October 2013
Haynes’s (1987)study of tension in female oYce workers (30). The only apparent diVerence
between this non-supportive study and the seven supportive studies is the type of sample.
The single non-supportive study is the only study with a strictly female white collar sample.
Studies examining the linear additive eVects of demands and control generated more
discrepant results. In 25 samples (partial) support for additive eVects were found (3, 4, 7, 9,
12, 17, 19, 23 (near-signi® cant), 24, 25, 29, 31, 32, 33, 35, 37, 40, 46, 49 (near-signi® cant),
52, 57, 58, 59, 60, 62), whereas 13 samples did not show additive eVects (8, 10, 13, 14, 27,
30, 36, 41, 42, 44, 45, 47, 51). In some studies evidence for the strain hypothesis was not
entirely consistent across the various outcomes examined (29, 32, 49, 57), across the various
statistical analyses performed (32, 33), or across the various concepts of demands and} or
control measured (4, 37, 52, 59, 62).
A comparison of the supportive and non-supportive studies in terms of design, sample
characteristics, conceptualization and measurement of the JDC dimensions, and outcomes,
indicated that the most notable diVerences were in the design, the sample, and the
measurement of JDC dimensions. Longitudinal studies were generally non-supportive ;
only two of the nine studies (4, 8, 10, 14, 23, 27, 42, 44, 47) showed additive eVects of
(change in) demands and control (4, 23). Among the cross-sectional studies, the non-
supportive studies generally used (predominantly) female samples (13, 30, 36, 44, 45, 51,
52), whereas the supportive studies more often used male or mixed samples in their study
(3, 4, 7, 9, 12, 17, 19, 25, 29, 31, 32, 33, 35, 40, 46, 49, 52, 58, 59, 60, 62). In those instances
where personnel managers ’ ratings were gathered (58), or scores on demands and control
were attributed to individuals on the basis of their job title (42), no support for the strain
hypothesis was found. Last but not least, the lack of an eVect of high strain in some of the
studies could have resulted from inclusion of other strong predictors in the analysis (e.g. job
satisfaction (13), anticipation of job loss (36)), or from the conceptual overlap between the
demands measured and the psychological outcome (14, 27, 44, 51). High correlations
between these predictors and the outcome can result in a non-signi® cant contribution of job
control to the variance in the outcome.
3.1.2. The buVer hypothesis of the Job Demand-Control model : Of the 31 studies examining
the buVer hypothesis of the JDC model, 15 instances showed (partial) evidence for a
buVering eVect of control on the relationship between demands and well-being.
In eight samples control moderated the relationship between high demands and
psychological distress (24, 25, 32, 37, 41, 45 (near-signi® cant), 52, 60). In half of these studies
the moderating eVect of control was found for some, but not for all outcomes (24, 25, 32,
102 M. Van der Doef and S. Maes
37). However, no speci® c psychological outcome showed consistent support for the buVer
hypothesis across studies. In three studies more than one conceptualization of demands
and} or control was examined, yielding diVerential results (41, 52, 60). Whereas timing
control moderated the impact of technological abstractness but not of technological
uncertainty in a sample of operators (41), method control did not have any moderating
eVect. In another study decision latitude conceptualized in terms of perceived routinization
and substantive complexity moderated the impact of number of work hours, whereas
perceived job control had no moderating eVects (52). Furthermore, in this study high
demands conceptualized in terms of time pressure and quantitative workload were not
moderated by any type of job control. A third study, comparing a broad measure of
decision latitude with a more focused measure of job control (decision authority), found
signi® cant moderating eVects only for the more speci® c measure of control (60).
Seven recent studies showed support for the buVering eVect of control, but only for
employee subgroups de® ned in terms of personality characteristics, type of organization, or
Downloaded by [Laurentian University] at 08:31 01 October 2013
hierarchical position within the organization (4, 14, 28, 29, 46, 47, 62). In three studies the
interactive eVect of demands and control was dependent on locus of control (14, 46, 47). In
a study of male driver examiners, demands and control had additive eVects on mental health
for persons with an internal locus of control, whereas a moderating eVect of control was
found in persons with an external locus of control (46). Similar results were found in a
longitudinal study of college graduates (47). Only for those with an external locus of
control did the interaction between demands and control contribute signi® cantly to the
prediction of level of anxiety. Quite opposite results were found in a longitudinal study of
psychological well-being in British accountants (14). The eVects of increasing levels of
demands upon the well-being of employees with an internal locus of control (internals) and
with high job control was weaker than that for subjects with another status on these
variables. In other words, it was especially in the case of internals that high job control
seemed to exert its buVering eVects. In other studies the moderating eVect of control
depended on private self-consciousness or type A behaviour (28, 29). These studies
indicated that for employees with a high level of private self-consciousness, and for type B
individuals, control moderated the impact of high demands. In a study of employees in the
manufacturing and services industries, the moderating eVect of control was only found for
the manufacturing industry employees (4). In a ® nal study, the moderating eVect was
present only for employees situated at a lower hierarchical level within the organization
(62). In a study of bank managers and clerks, control moderated the relationship between
role con¯ ict and distress for the bank clerks, but not for the bank managers.
The remaining 16 studies failed to support the buVer hypothesis of the JDC model (3, 8,
9, 10, 12, 17, 18, 19, 23, 27, 31, 35, 49, 51, 57, 58). Furthermore, in one study a longitudinal
analysis failed to con® rm its cross-sectional ® ndings in support of the buVer hypothesis (4).
Examination of the supportive and non-supportive studies in more detail indicated that
support is independent of the characteristics of the sample (heterogeneity, sample size,
occupational group, gender), and the speci® c outcome under study. With one exception
(47) support is restricted to the cross-sectional studies. Moreover, further examination of
the measurement of demands and control suggested that the non-supportive studies more
often used a broad conceptualization of either demands or control (e.g. 8, 17, 23).
Supportive studies on the contrary use more focused measures of job demands (e.g.
quantitative overload, monitoring demands, role con¯ ict) and job control (e.g. in¯ uenceon
amount and pace of work, decision authority) (4, 28, 41, 60, 62). Important in this respect
seems to be the extent to which the speci® c type of control corresponds to the speci® c type
of demand, in other words whether control is indeed applicable to the demands experienced.
The Job Demand-Control(-Support ) Model 103
Furthermore, the results suggest that speci® c subgroups of employees (in terms of
personality characteristics, type of organization or hierarchical position within the
organization) can be identi® ed who experience more bene® cial eVects from high control.
3.1.3. The iso-strain hypothesis of the Job Demand-Control-Support model : The one study that
examined the iso-strain hypothesis in a non-linear way was not supportive of the
hypothesis. Iso-strain was not associated with an elevated risk for psychological distress
among white collar workers (6).
As was the case with the strain hypothesis, the iso-strain hypothesis was in most studies
examined in terms of additive eVects of demands, control, and support on psychological
distress (3, 4, 8, 9, 12, 13, 19, 23, 24, 27, 32, 34, 35, 40, 42, 52, 58, 59). Support for the iso-
strain was found in eight cross-sectional studies (9, 19, 24, 32, 35, 52, 58, 59) and in one
prospective study (23). However, an equal number of studies failed to ® nd such linear
additive eVects (3, 4, 8, 12, 13, 27, 34, 40, 42).
Downloaded by [Laurentian University] at 08:31 01 October 2013
3.1.5. Summary : Summarizing the above review of studies applying the JDC(S) model
to general psychological well-being, the distinction between the various hypotheses appears
to be worthwhile. Generally speaking, the strain and iso-strain hypotheses yielded more
consistent support than the buVer hypotheses of the JDC model and the JDCS model.
Comparison of studies supportive and non-supportive of the various hypotheses indicated
diVerences in design, sample characteristics, and measurement and conceptualization of the
JDCS dimensions. Whereas cross-sectional studies suggest that working in high strain and
iso-strain situations is associated with psychological distress, there is almost no support in
longitudinal studies. Furthermore, the results indicate that although self-report measures of
the work dimensions are related to distress, non-self-report ratings are not. Finally,
especially with respect to the strain hypothesis, gender diVerences are evident, as there is
considerably less support for the negative impact of high strain in the case of female
employees.
104 M. Van der Doef and S. Maes
With respect to the buVering eVect of control on the impact of high demands, support
is mainly to be found in cross-sectional studies. Furthermore, the inconsistency in support
for this eVect appears to depend on how demands and control are conceptualized. Two
things seem to be important. First, in the supportive studies demands and control are
conceptualized at a comparable level of speci® city. Second, the speci® c aspect(s) of control
more closely match the speci® c demand(s). In other words, the control measure more
adequately represented the amount of control the employee can exert over the demands he
or she experiences. Furthermore, the results suggest that certain subpopulations bene® t
more than others from the moderating eVect of high control.
In the few studies examining the moderating eVect of social support on the impact of
high strain, supportive results are restricted to male samples.
Within this outcome category a distinction is made between job satisfaction, burnout,
and job-related psychological well-being.
3.2.1. Job satisfaction : In 31 samples the JDC(S) model was examined in relation to job
satisfaction (table 1: 1, 2, 3, 5, 9, 12, 13, 15, 16, 19, 20, 22, 23, 24, 26, 29, 31, 32, 38, 39, 40,
41, 45, 48, 50, 53, 55, 57, 60, 61, 62). Twenty-three studies reported on the buVer hypothesis
of the JDC model (3, 5, 9, 12, 15, 16, 19, 20, 23, 24, 29, 31, 32, 38, 39, 41, 45, 48, 53, 57,
60, 61, 62), whereas only six studies examined the buVer hypothesis of the expanded JDCS
model (12, 15, 24, 32, 39, 48). With one exception (16) all studies report on the strain
hypothesis. The iso-strain hypothesis was examined in 14 samples (1, 3, 5, 9, 12, 13, 15, 19,
23, 24, 32, 39, 40, 48).
The studies of job satisfaction were all cross-sectional, and measurement of the
dimensions of the JDC(S) model was mostly based on self-report. Only in a few instances
were objective data gathered from other sources (16, 20).
3.2.1.1. The strain hypothesis of the Job Demand-Control model : As already mentioned, all
but one of the studies (16) reported on the strain hypothesis. The two studies examining the
strain hypothesis in a non-linear way are both supportive (26, 31). Furthermore, in two
samples higher ratios of demands and control, indicating high job strain, were associated
with job dissatisfaction (2, 50).
About one-half of the studies examining linear additive eVects of demands and control
are partially or fully supportive (1, 3, 5, 9, 15, 19, 20, 24, 29, 31, 32, 38, 40, 50, 60, 61). In
one study support was restricted to the females within the sample (19). In other studies
support depended on the measure of demands (20) or control (38, 60). A supervisor rating
of demands gave non-supportive results (20). In two studies support for the strain
hypothesis was not found for decision authority, but was found for other types of control
(schedule control, control over the physical environment, resource control, and skill
discretion) (38, 60).
Examination of the supportive and non-supportive studies in further detail showed
diVerences in sample characteristics, and in type of analysis. Supportive studies generally
examined somewhat larger samples, including mainly health care personnel, blue collar
workers and samples from the general population. The non-supportive studies are
characterized by generally smaller sample sizes, and more often include female and white-
collar populations. Furthermore, the studies examining the strain hypothesis in non-linear
The Job Demand-Control(-Support ) Model 105
analyses, or focusing on the discrepancy between demands and control, are generally
supportive.
3.2.1.2. The buVer hypothesis of the Job Demand-Control model : In ten studies full or partial
support for the buVering eVect of control on the demands± job satisfaction relationship was
found (16, 20, 24, 29, 32, 41, 45, 48, 53, 60). In some studies the ® ndings depended on the
type of statistical analysis : support was found in linear analyses, but not in non-linear
analyses (32), or when a special interaction term was used rather than the conventional
cross-product term (53). Furthermore, in a number of studies support depended on the way
demands (16, 20, 41) or control (60) was measured. Non-self-report ratings of demands
showed less consistent moderation by control than did self-reported demands. In one study
a more speci® c measure of job control moderated the impact of demands, whereas the
broad concept of decision latitude failed to do so (60). Finally, in one study the moderating
impact of control was stronger for a subpopulation within the sample, namely for the type
Downloaded by [Laurentian University] at 08:31 01 October 2013
Bs (29).
Three studies observed interactive eVects between demands and control, but discounted
them on the basis of limited additional contribution to the explanation of job dissatisfaction
(5), or due to their discordant form (19, 57). Another 10 studies failed to support the buVer
hypothesis of the JDC model (3, 9, 12, 15, 23, 31, 38, 39, 61, 62).
Comparison of the supportive and non-supportive studies indicated that con® rmation of
the buVer hypothesis seems to be independent of sample characteristics (heterogeneity,
sample size, occupational group, gender). In line with the general pattern observed in
studies testing the buVer hypothesis for general psychological well-being, the studies that
supported this hypothesis for job satisfaction tended to use more focused measures of
demands and control. Furthermore, supportive studies are characterized by better
correspondence between the control measure and the speci® c type of demands measured
(e.g. 16, 29, 60).
3.2.1.3. The iso-strain hypothesis of the Job Demand-Control-Support model : The iso-strain
hypothesis was examined in 14 samples (1, 3, 5, 9, 12, 13, 15, 19, 23, 24, 32, 39, 40, 48).
Partial or full support for linear additive eVects of demands, control and support on job
satisfaction was found in seven studies (1, 3, 9, 19, 24, 32, 40). Furthermore, a canonical
analysis suggested additive eVects in the study by Beehr, and Drexler (1986) as well (5). No
such linear additive eVects were found in six studies (12, 13, 15, 23, 39, 48).
Further examination of the studies suggested that supportive and non-supportive studies
diVer in the composition of their sample. Supportive studies mainly included blue collar
workers, or general population samples consisting of male or mixed samples. Non-
supportive studies mostly examined female or predominantly female samples.
3.2.1.4. The buVer hypothesis of the Job Demand-Control-Support model : Of the six studies
examining the buVer hypothesis of the JDCS model (12, 15, 24, 32, 39, 48) only two studies
yielded support for this hypothesis. Social support moderated the impact of high strain on
job satisfaction in two samples of male employees (24, 32). The only characteristic
diVerentiating the supportive from the non-supportive studies is the composition of the
sample. Whereas the two supportive studies both used male samples, the other studies used
female or mixed samples.
3.2.1.5. Summary : The results of studies testing the JDC(S) model with respect to job
satisfaction are generally in line with those for general psychological well-being. The strain
106 M. Van der Doef and S. Maes
hypothesis and the iso-strain hypothesis received fairly consistent support, especially in
male or mixed samples, suggesting gender diVerences in the impact of high (iso)strain. As
mentioned in connection with the studies on general psychological well-being, the
measurement of demands and control seems to be essential to the moderating eVect of
control. Supportive studies tend to conceptualize demands and control at a comparable
level of speci® city, and the control that is measured concerns control over the speci® c
demands that the employee experiences. Furthermore, whereas self-reported demands and
control are consistently associated with job satisfaction, non-self-report ratings are not.
3.2.2. Burnout: Only four cross-sectional studies (31, 39, 43, 56) have examined the
JDC(S) model in the context of burnout. All four studies reported on the strain hypothesis
of the JDC model. This hypothesis tended to ® nd support, with only one study showing
no additive eVects of demands and control on burnout (43). Whereas most supportive
Downloaded by [Laurentian University] at 08:31 01 October 2013
studies focused on health care personnel, the non-supportive study included an occupational
group that has not previously been shown to suVer from burnout, namely workers at a
construction company. All four studies tested for moderating eVects of control and were
unanimous in their refutation of this hypothesis.
The one study that examined the iso-strain hypothesis was supportive (39), reporting
linear additive eVects of demands, control and support on burnout among female social
workers. No buVering eVect of support on the impact of high strain was found in the two
studies examining this hypothesis (39, 43).
3.2.2.1. Summary : Although only a limited number of studies have examined burnout,
their consistent ® ndings suggest a number of conclusions. Support for the (iso)strain
hypothesis is fairly consistent, indicating a relationship between high (iso)strain work and
burnout. However, it seems unlikely that the negative impact of high demands on burnout
can be moderated by high levels of control and social support.
3.2.3. Job-related psychological well -being : Eight studies examined indicators of job-related
psychological well-being, such as self-reported occupational stress and worries, and job-
related mood (11, 21, 24, 26, 41, 54, 55, 61).
The strain hypothesis was examined in all eight studies, with generally supportive results.
Three studies showed linear additive eVects of demands and control (21, 26, 61). In addition,
four studies showed associations between working in a high-strain job and job-related
distress (11, 24, 54, 55). In one of these studies (54), however, an equal proportion of
workers in high-strain jobs and active jobs reported medium to high levels of work stress,
suggesting that work stress was solely the result of high demands.
One of the two studies on the moderating eVect of control (41, 61) showed partial
support for this hypothesis (41). Only two studies investigated the social support dimension
(21, 24). Whereas one study using a large heterogeneous sample showed additive eVects of
demands, control, and support on job-related mood, with support moderating the impact
of high strain (24), the other study that used a small sample of car mechanics, showed no
in¯ uence of social support on job-related worries (21).
lower job-related psychological well-being. The few studies on the moderating eVects of
control and social support yielded inconsistent results.
4. Discussion
In this review a distinction has been drawn between two hypotheses tested in research on
the Job Demand-Control (JDC) model and the Job Demand-Control-Support (JDCS)
model. For both the JDC and the JDCS model a distinction was made between the strain
or iso-strain hypothesis and the buVer hypothesis. The results of this review suggest that the
distinction between these hypotheses is worthwhile. Table 2 gives the ratio of supportive
studies to the total number of studies for each hypothesis and outcome category.
The strain hypothesis of the JDC model, which states that the most negative
psychological well-being is found in employees working in high demands± low control
Downloaded by [Laurentian University] at 08:31 01 October 2013
mis® t due to a highly active coping style and low job control showed the most prominent
increase in burnout due to job demands. Need for control on the contrary did not act as a
moderator in the JDC model.
The supportive and non-supportive studies on the moderating eVect of control in this
review diVer in two respects. First, almost none of the longitudinal studies yielded support
for the buVer hypothesis. Second, close examination of the conceptualization and
measurement of demands and control indicated that many of the supportive studies
measured demands and control at a comparable level of speci® city and achieved a closer
match in their conceptualization of demands and control. Studies that measured a speci® c
demand (e.g. time pressure) in combination with a speci® c corresponding aspect of control
(e.g. decision authority over pace and method) were particularly likely to show a
moderating eVect of control (Kushnir, and Melamed, 1991; Kivimaki, and Lindstro$ m,
1995; Wall et al., 1996). The non-supportive studies often used a broader conceptualization
of demands (e.g. responsibility pressure, quality concern, role con¯ ict and work overload
(Bromet, Dew, Parkinson, and Schulberg, 1988)), or job control (e.g. decision latitude). As
such, the job control reported by the employees in these studies did not adequately re¯ ect
the control that they could exert over the demands experienced (Kivimaki, and Lindstro$ m,
1995; Kasl, 1996).
In the case of social support, the other moderator within the JDCS model, the need for
a match between the stressor and the type of support is well documented. Social support
research suggests that it is only when social support matches the stressor which is
experienced that support can buVer potentially negative eVects of the stressor (Terry,
Neilsen, and Perchard, 1993). The question of whether the moderating eVect of social
support in a work context depends on the source or type of support cannot be answered
here, due to the limited number of studies on the JDCS model. It seems worthwhile to focus
future research on the potential buVering eVects of certain components of job control and
of diVerent forms of social support. BuVering eVects are to be expected when job control
and social support `match ’ the demands of the job.
This review has identi® ed at least four methodological problems that should be
overcome in future research. First, in a number of non-supportive studies there was a
conceptual overlap between the measure of demands and the psychological outcome under
study (Kawakami, Haratani, and Araki, 1992; Reisine, and Fi® eld, 1995).
Second, susceptibility to self-report bias or common method bias is high in most studies
in this review. Self-report measures were typically used to estimate the JDC(S) dimensions,
as well as employee well-being. This can lead to an overestimation of the strength of the
The Job Demand-Control(-Support ) Model 109
association between work characteristics and well-being, as such in¯ ating the apparent
support for the (iso)strain hypothesis. As Wall et al. (1996)have argued, this is not a problem
for the tests of the moderating eVect of control and support, as there is no reason to suspect
that this bias would lead to an overestimation of the interactive eVects; on the contrary, the
main eVects would be overestimated at the cost of underestimation of interaction eVects.
Furthermore, the majority of studies used cross-sectional designs, yielding no in-
formation on the causality of the relations between work characteristics and well-being.
The lack of evidence of a negative eVect of high (iso)strain and a moderating eVect of
control in most of the longitudinal studies is worrying. More longitudinal studies, in which
the methodological problems mentioned above are overcome, are necessary. These studies
should include measures of the JDC(S) characteristics at each follow-up, so that it can be
determined whether the lack of predictive power in prior studies is attributable to changes
in the work characteristics over time.
Last but not least, the results of studies using an imputation method (Muntaner, Tien,
Downloaded by [Laurentian University] at 08:31 01 October 2013
Eaton, and Garrison, 1991),or personnel managers ’ ratings of work characteristics (Dwyer,
and Ganster, 1991; Stansfeld, North, White, and Marmot, 1995) are generally non-
supportive of the JDC(S) hypotheses. The lack of support in these studies suggests that the
way in which the individual experiences work characteristics is crucial to their eVects. A
challenge for the future will be to develop a self-report questionnaire on work characteristics
that provides a more objective report of the work environment by the individual employee.
As Frese, & Zapf (1988) indicate, measures of job characteristics can be placed on a
continuum from more objective to more subjective, depending on the cognitive and
emotional processing that the items require. Crucial in this matter is the wording of the
items.
Another issue in the measurement of the JDCS dimensions refers to the conceptualization
of demands. Karasek, and Theorell (1990) indicate that for most workers work load is the
central component of job demands. However, this may not apply to all occupational
groups. For health care personnel and teachers, stressors related to interactions with patients
and students could constitute equally or more important job demands. Monitoring
demands and equipment problems could be the central job demands for manufacturing
employees. Depending on the speci® c demands of a job, an employee may need speci® c
corresponding types of control and social support to cope with these demands. This
suggests that occupation-speci® c measurement of demands, control, and support could
improve the explanatory and predictive power of the JDC(S) model (Kasl, 1996).
Summarizing, future research should address theoretical questions concerning the
JDC(S) model, as well as identify the work characteristic(s) that are crucial to the creation
of `healthy work situations ’.
First of all, from a theoretical as well as a practical point of view there is a need to further
clarify the nature of the eVects of demands, control and support. In future tests of the
JDC(S) model it would be worthwhile to examine the various hypotheses concurrently, so
that additive and interactive eVects of demands, control, and social support can be
examined. Additionally, it seems important to examine whether the relations are of a linear
or a non-linear nature. The possibility of threshold eVects and curvilinear eVects has been
raised (Warr, 1990, 1994; Landsbergis, Schnall, Deitz, Friedman, and Pickering, 1992;
Schnall et al., 1994).
Second, the measurement of work characteristics should be improved. The assessment of
job demands should not be restricted to, for instance, the time pressure component, but
should instead give a more comprehensive view of job demands. Bearing in mind the need
for a match between stressors and resources, those aspects of control and social support
110 M. Van der Doef and S. Maes
should be assessed that enable the individual to eVectively cope with the speci® c job
demands. Furthermore, it is important to assess the objective work environment of the
individual employee. In other words, measures of demands, control and support should
re¯ ect not the satisfaction of the employee with the job characteristics, but the employee’s
work situation. Considering this, it would be worthwhile to develop more occupation-
speci® c measurements for speci® c occupational groups.
Third, longitudinal studies are needed to establish the causality of the relationships
between job demands, control, and support and psychological well-being. Fourth, the issue
of vulnerability (or resistance) to unfavourable work characteristics merits further
exploration. A number of studies have identi® ed populations that are particularly
vulnerable (or resistant) to negative work conditions. As such, the exclusive focus on the
characteristics of the work environment, which the JDC(S) model advocates, would be
only part of the total picture. A closer look at the role of individual characteristics (e.g.
coping style, locus of control) in the relationship between work environment and employee
Downloaded by [Laurentian University] at 08:31 01 October 2013
Acknowledgements
The authors thank Laura Sweeney, MA, for reviewing the English, and the three
anonymous reviewers for their valuable comments.
References
Al f r e d s s o n , L., Sp e t z , C. L., & Th e o r e l l , T. (1985). Type of occupation and near-future
hospitalization for myocardial infarction and some other diagnoses. International Journal of
Epidemiology , 14, 378± 388.
Am i c k , B. C., & Ce l e n t a n o , D. D. (1991).Structural determinants of the psychosocial environment :
introducing technology in the work stress framework. Ergonomics, 34, 625± 646.
Ar n e t z , B. B., An d r e a s s o n , S., St r a n d b e r g , M., En e r o t h , P., & Ka l l n e r , A. (1988). Comparison
between surgeons and general practitioners with respect to cardiovascular and psychosocial
risk factors among physicians. Scandinavian Journal of Work , Environment and Health , 14,
118± 124.
Ba k e r , E., Is r a e l , B., & Sc h u r m a n , S. (1996). Roleof control and support in occupational stress : an
integrated model. Social Science and Medicine , 43, 1145± 1159.
Ba r n e t t , R. C., & Br e n n a n , R. T. (1995). The relationship between job experiences and
psychological distress : a structural equation approach. Journal of Organizational Behavior , 16,
259± 276.
Ba r n e t t , R. C., & Br e n n a n , R. T. (1997). Change in job conditions, change in psychological
distress, and gender : a longitudinal study of dual-earner couples. Journal of Organizational
Behavior , 18, 253± 274.
Be e h r , T. A., & Dr e x l e r , J. A. (1986). Social support, autonomy, and hierarchical level as
moderators of the role characteristics-outcome relationship. Journal of Occupational Behaviour ,
7, 207± 214.
The Job Demand-Control(-Support ) Model 111
Cl e g g , C., Wa l l , T., & Ke m p , N. (1987). Women on the assembly line : a comparison of main and
interactive explanations of job satisfaction, absence and mental health. Journal of Occupational
Psychology , 60, 273± 287.
Da n i e l s , K., & Gu p p y , A. (1994). Occupational stress, social support, job control, and psychological
well-being. Human Relations , 12, 1523± 1544.
d e Jo n g e , J., & Ko m p i e r , M. A. J. (1997). A critical examination of the Demand-Control-Support
model from a work psychological perspective. International Journal of Stress Management , 4,
235± 258.
d e Jo n g e , J., & La n d e w e e r d , J. A. (1993). Toetsing van de job demand-control benadering bij
werknemers in de gezondheidszorg . Een secundaire data-analyse [Test of the Job Demand-
Control approach among health care workers : a secondary analysis], Gedrag en Organisatie , 21,
79± 92.
d e Ri j k , A. E., Le Bl a n c , P. M., Sc h a u f e l i , W. B., & d e Jo n g e , J. (1998). Active coping and need
for control as moderators of the job demand-control model : eVects on burnout. Journal of
Occupational and Organizational Psychology , 71, 1± 18.
De r o g a t i s , L. R. (1983). SCL-90-R: Administration , Scoring & Procedures Manual -II, 2nd edn.
Baltimore : Clinical Psychometric Research.
Dw y e r , D. J., & Ga n s t e r , D. C. (1991). The eVects of job demands and control on employee
attendance and satisfaction. Journal of Organizational Behavior , 7, 595± 608.
El o v a i n i o , M., & Ki v i m a k i , M. (1996).Occupational stresses, goal clarity, control, and strain among
nurses in the Finnish health care system. Research in Nursing and Health , 19, 517± 524.
El s a s s , P. M., & Ve i g a , J. F. (1997). Job control and job strain : a test of three models. Journal of
Occupational Health Psychology , 2, 195± 211.
Fl e t c h e r , B. C., & Jo n e s , F. (1993). A refutation of Karasek’s demand-discretion model of
occupational stress with a range of dependent measures. Journal of Organizational Behavior , 14,
319± 330.
Fo x , M. L., Dw y e r , D. J., & Ga n s t e r , D. C. (1993). EVects of stressful job demands and control on
physiological and attitudinal outcomes in a hospital setting. Academy of Management Journal , 36,
289± 318.
Fr e s e , M., & Za p f , D. (1988). Methodological issues in the study of work stress : objective vs
subjective measurement of work stress and the question of longitudinal studies. In C. L.
Cooper and R. Payne (Eds.), Causes, Coping and Consequences of Stress at Work (pp. 375± 411).
Chichester: Wiley.
Ga n s t e r , D. C. (1989). Worker control and well-being : a review of research in the workplace. In
S. L. Sauter, J. J. Hurrell, & C. L. Cooper (Eds.), Job Control and Worker Health (pp. 3-24).
Chichester: Wiley.
Ge l l e r , P. A., & Ho b f o l l , S. E. (1994).Gender diVerences in job stress, tedium and social support in
the workplace. Journal of Social and Personal Relationships , 11, 555± 572.
Go l d b e r g , D. P. (1978). Manual of the General Health Questionnaire . Windsor : NFER.
Ha l l , E. M. (1991). Gender, work control, and stress : a theoretical discussion and an empirical test.
In J. V. Johnston, & G. Johansson (Eds.), The Psychosocial Work Environment. Work
Organization , Democratization and Health (pp. 89± 108). Amityville, NY : Baywood.
112 M. Van der Doef and S. Maes
Re i s i n e , S., & Fi f i e l d , J. (1995). Family work demands, employment demands and depressive
symptoms in women with rheumatoid arthritis. Women and Health , 22, 25± 45.
Ro x b u r g h , S. (1996). Gender diVerences in work and well-being : eVects of exposure and
vulnerability. Journal of Health and Social Behavior , 37, 265± 277.
Ro x b u r g h , S. (1997). The eVect of children on the mental health of women in the paid labor force.
Journal of Family Issues, 18, 270± 289.
Sa u t e r , S. L. (1989). Moderating eVects of job control on health complaints in oYce work. In S. L.
Sauter, J. J. Hurrell, & C. L. Cooper (Eds.), Job Control and Worker Health (pp. 91-96).
Chichester: Wiley.
Sc h e c h t e r , J., Gr e e n , L. W., Ol s e n , L., Kr u s e , K., & Ca r g o , M. (1997). Application of Karasek’s
Demand } Control Model in a Canadian occupational setting including shift workers during a
period of reorganization and downsizing. American Journal of Health Promotion, 11, 394± 399.
Sc h m i d t , S. H., Me i j m a n , T. F., Sc h o l t e n , A., v a n Oe l , C. J., & Oo r t -Ma r b u r g e r , D. (1993).
Factors contributing to job satisfaction following rehabilitation for musculoskeletal impair-
ments. Journal of Occupational Rehabilitation , 3, 213± 222.
Sc h n a l l , P. L., La n d s b e r g i s , P. A., & Ba k e r , D. (1994).Job strain and cardiovascular disease. Annual
Downloaded by [Laurentian University] at 08:31 01 October 2013