ARTICLE IN PRESS

Social Science & Medicine 60 (2005) 1117–1131

www.elsevier.com/locate/socscimed

Reviewing the effort–reward imbalance model: drawing up the

balance of 45 empirical studies
Natasja van Vegchela,, Jan de Jongea, Hans Bosmab, Wilmar Schaufelia
a
Department of Social and Organizational Psychology, Utrecht University, P.O. Box 80.140, 3508 TC Utrecht, The Netherlands
b
Department of Medical Sociology, Maastricht University, P.O. Box 616, 6200 MD, Maastricht, The Netherlands
Available online 13 August 2004

Abstract

The present paper provides a review of 45 studies on the Effort–Reward Imbalance (ERI) Model published from
1986 to 2003 (inclusive). In 1986, the ERI Model was introduced by Siegrist et al. (Biological and Psychological Factors
in Cardiovascular Disease, Springer, Berlin, 1986, pp. 104–126; Social Science & Medicine 22 (1986) 247). The central
tenet of the ERI Model is that an imbalance between (high) efforts and (low) rewards leads to (sustained) strain
reactions. Besides efforts and rewards, overcommitment (i.e., a personality characteristic) is a crucial aspect of the
model. Essentially, the ERI Model contains three main assumptions, which could be labeled as (1) the extrinsic ERI
hypothesis: high efforts in combination with low rewards increase the risk of poor health, (2) the intrinsic
overcommitment hypothesis: a high level of overcommitment may increase the risk of poor health, and (3) the
interaction hypothesis: employees reporting an extrinsic ERI and a high level of overcommitment have an even higher
risk of poor health. The review showed that the extrinsic ERI hypothesis has gained considerable empirical support.
Results for overcommitment remain inconsistent and the moderating effect of overcommitment on the relation between
ERI and employee health has been scarcely examined. Based on these review results suggestions for future research are
proposed.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: Effort–reward imbalance; ERI model; Work overcommitment; Employee health

Introduction Effort–Reward Imbalance (ERI) Model (Siegrist, 1996;

Siegrist, Siegrist, & Weber, 1986). The ERI Model has
Occupational health researchers have tried to gain its origin in medical sociology and emphasizes both the
more insight into the relationship between work effort and the reward structure of work (Marmot,
characteristics and employee health. With the help of Siegrist, Theorell, & Feeney, 1999). The model is based
work stress models they attempt to reduce the complex upon the premise that work-related benefits depend
reality into comprehensive and parsimonious models, upon a reciprocal relationship between efforts and
highlighting some core elements in order to explain job- rewards at work. Efforts represent job demands and/or
related health. One of the most important models that obligations that are imposed on the employee. Occupa-
has recently guided occupational health research is the tional rewards distributed by the employer (and by
society at large) consist of money, esteem, and job
Corresponding author. Tel.: +31-30-2534196; fax: +31-30- security/career opportunities. More specifically, the ERI
2534718. Model claims that work characterized by both high
E-mail address: [email protected] (N. van Vegchel). efforts and low rewards represents a reciprocity deficit

0277-9536/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2004.06.043
 ARTICLE IN PRESS
1118 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

between ‘‘costs’’ and ‘‘gains’’. This imbalance may cause (AMI). In addition, he found strong independent effects
sustained strain reactions. So, working hard without of ERI and vital exhaustion on AMI (Appels, Siegrist, &
receiving appreciation is an example of a stressful Vos, 1997). Those results suggest that ERI could lead to
imbalance. In addition, it is assumed that this process cardiac events, but also that this relation might be
will be intensified by overcommitment (a personality mediated by vital exhaustion. Implicitly, the ERI Model
characteristic), such that highly overcommitted employ- can also be considered as predictive for psychological
ees will respond with more strain reactions to an ERI, in well-being, as ERI evokes strong negative emotions,
comparison with less overcommitted employees. which are related to impaired well-being (cf. Gaillard &
Over the past years, the ERI Model has gained Wientjes, 1994). Furthermore, it has been argued that
popularity (especially in European research), and the model can be applied to addictive behavior as well.
numerous studies have applied the model to various According to Blum, Cull, Braverman, and Comings
health outcomes. Although the number of ERI studies (1996), prolonged stress leads to dysfunction or disrup-
rose steadily, a detailed review evaluating those studies is tion of the mesolimbic dopamine system, which in turn
to our knowledge still lacking. The present study tries to stimulates addictive behavior. To summarize, an ERI
fill this gap by presenting a review of empirical studies seems to evoke adverse health by stimulating neuro-
testing the ERI Model. Since the ERI Model has biological, psychological and behavioral pathways.
considerably evolved over time and some studies might In general, the idea prevails that people will not
have been designed from a previous/different perspective passively stay in a high-effort–low-reward imbalance
of the model, it is necessary to get more familiar with the situation, but that they will try to cognitively and
background of the model. Hence, we will give a short behaviorally reduce their efforts and/or maximize their
historical overview of the most relevant developments rewards (e.g., cognitive theory of emotion (Lazarus,
that preceded the model in its current form. 1991) and expectancy theory of motivation (Schönpflug
& Batman, 1989)). Hence, an ERI might not influence
Development of the ERI model health over a longer period. But according to Siegrist
(1996), negative effect associated with ERI may not be
In 1986, a sociological framework being the ERI Model consciously appraised, as it is a chronically recurrent
was introduced by Siegrist et al. (1986) to predict and everyday experience (cf. Gaillard & Wientjes, 1994).
explain (the onset of) cardiovascular-related outcomes. Furthermore, Siegrist (1996) identified some specific
The ERI Model claims that the work role is crucial in circumstances under which a high cost/low gain condi-
order to fulfill individual self-regulatory needs. That is, tion is maintained: (1) when there is no alternative
work offers opportunities to acquire self-efficacy (e.g., choice on the labor market, (2) for strategic reasons
successful performance), self-esteem (e.g., recognition) and (e.g., expecting future gains), and (3) when the employee
self-integration (e.g., belonging to a significant group). is characterized by a motivational pattern of excessive
Based on the principle of social exchange (i.e., reciprocity), work-related overcommitment. Overcommitment is seen
the employee invests efforts and expects rewards in return. as a personality characteristic based on the cognitive,
However, in case an imbalance is present between high emotional and motivational elements of Type A
effort and low reward, this taken-for-granted routine is behavior that reflect an exorbitant ambition in combina-
disrupted and the fulfillment of the self-regulatory needs is tion with the need to be approved and esteemed
threatened. According to Siegrist et al. (1986) this (Hanson, Schaufeli, Vrijkotte, Plomp, & Godaert,
imbalance may lead to a state of ‘‘active distress’’ by 2000; Siegrist, 1998). Overcommitment can be defined
evoking strong negative emotions, which in turn activate as the person-specific component, whereas efforts and
two stress axes, i.e., the sympathetic-adrenomedullary and rewards compromise the situation-specific component.
the pituitary–adrenal–cortical system (Henry & Stephens, To measure the key concepts of the ERI Model,
1977). In the long run, sustained activation of the information was gathered from different sources, i.e.,
autonomic nervous system may contribute to the devel- contextual information (such as administrative data and
opment of physical (e.g., cardiovascular) and mental (e.g., objective measures), descriptive and evaluative informa-
depression) diseases (see also Weiner, 1992). tion (through interviews and questionnaires). A combi-
In its premature years, the ERI Model was primarily nation of those sources was mainly used to measure
used to investigate cardiovascular outcomes. It was not effort and reward. Overcommitment was solely assessed
until 1998 that the model was applied to other by a questionnaire (cf. Matschinger, Siegrist, Siegrist, &
psychological and behavioral outcomes as well. Appels Dittmann, 1986). Subsequently, a questionnaire was
can be considered as a pioneer in linking ERI (i.e., high developed to measure all components of the ERI Model,
effort and low reward) to psychological outcomes such i.e., effort, reward and overcommitment. The introduc-
as vital exhaustion (Appels, Siegrist, & Vos, 1997). tion of this so-called ERI Questionnaire (ERI-Q;
Appels’ work (Appels & Schouten, 1991) showed that (Siegrist & Peter, 1996a) led to a predominant use of
vital exhaustion may lead to acute myocardial infarction questionnaires to test the ERI Model.
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1119

The effort scale in the ERI-Q contains six items of

High effort Low reward
which the content varies from physical load, time
pressure, interruptions, responsibility, working overtime
to increasing demands. Siegrist et al. (2004) have
recommended to include the item on physical load only
in those occupational groups where prevalence of
physical workload is part of the typical task profile. Extrinsic Intrinsic Money
Reward in the ERI-Q has been operationalized by
means of 11 items, and is usually measured as a (demands, (critical coping; Esteem
composite measure. Theoretically, a three-factor struc-
ture underlies the concept of reward (i.e., money, esteem obligations) need for control) Status control
and security/career opportunities). Recent studies con-
firm the presence of the three-factorial structure (Siegrist Fig. 1. Original ERI Model (Siegrist, 1996, p. 30).
et al., 2004), and the importance of splitting three types
of rewards (Dragano, Knesebeck, Rödel, & Siegrist,
2003; van Vegchel, de Jonge, Bakker, & Schaufeli, 2002).
The operationalization of overcommitment has consid-
Intrinsic Overcommitment
erably changed over time. Originally, overcommitment
(person)
was operationalized by the scale ‘‘need for control’’, as a (need for control and approval)
more work-related reformulation of the Type A concept
(cf. Matschinger et al., 1986). The need for control scale
contains two latent factors: vigor and immersion. Vigor
Extrinsic
refers to successful coping (by perfectionism and hard High effort Low reward
(situation)
work). Immersion defines a critical state of coping with
demands reflecting frustrated, but continued efforts and demands money
associated negative feelings. Immersion consists of four obligations esteem
subscales: (a) need for approval, (b) competitiveness, (c) security/career opportunities
disproportionate irritability, and (d) inability to with-
Fig. 2. Current ERI Model (Siegrist, 1999, p. 40).
draw from work (Siegrist, 1996). Although some
empirical studies were able to replicate the factorial
structure of immersion (e.g., Peter et al., 1998), other
studies could not replicate the respective factorial of effort invested is dependent upon both extrinsic (i.e.,
structure, and showed that especially the factor ‘‘in- demands and obligations from work) and intrinsic (i.e.,
ability to withdraw from work’’ was essential for the overcommitment) efforts. So, the main assumption of
ERI Model (see Hanson et al., 2000; Niedhammer, the ERI Model was that a mismatch between high
Siegrist, Landre, Goldberg, & Leclerc, 2000). Therefore, extrinsic or intrinsic efforts and low reward may lead to
a shorter version was developed to represent over- an adverse health. Later, as shown in Fig. 2, over-
commitment, mainly consisting of inability to withdraw commitment is seen as an independent concept. Over-
from work (five items) and one item of disproportionate commitment influences the perception of both high
irritability (Siegrist et al., 2004). efforts and low reward, and therefore influences employ-
Despite developments in their operationalization over ee health indirectly. In addition, overcommitment is
time, the concepts effort, reward and overcommitment thought to have a direct effect on employee health as
remained the core components of the ERI Model. well, as being highly overcommitted (i.e., involved with
Graphical representations of the original version and the work all the time) might be exhaustive in the long run.
current version of the ERI Model are shown in Figs. 1 Based upon this line of reasoning, Siegrist (2002) has
and 2, respectively. As can be seen, two concepts have formulated three predictions that can be postulated for
been relabeled: ‘‘intrinsic effort/need for control’’ into the ERI Model. Firstly, the extrinsic ERI hypothesis: an
‘‘overcommitment’’, and ‘‘status control’’ into ‘‘security/ imbalance between (high) extrinsic effort and (low)
career opportunities’’. Reasons for this are not reported reward increases the risk of poor health, over and above
in the literature. the risks associated with each one of the components
The most profound change in Figs. 1 and 2 comprises (i.e., high efforts and low rewards). Secondly, the
the role of overcommitment. According to Fig. 1, intrinsic overcommitment (OVC) hypothesis: a high level
overcommitment is part of effort. Because highly of overcommitment, possibly resulting in continued
overcommitted employees underestimate challenging exaggerated efforts combined with disappointing re-
situations and overestimate their own capability, they wards, may also increase the risk of poor health (even in
tend to invest (too) many efforts. Therefore, the amount the absence of an extrinsic ERI, i.e., a main effect of
 ARTICLE IN PRESS
1120 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

overcommitment). And finally, the interaction hypothesis paper presentations, personal communication, unpub-
(i.e., ERI
OVC): an extrinsic ERI in combination with lished papers, dissertations, etc. were excluded).
a high level of overcommitment leads to the highest risk Based on the criteria mentioned above, 45 studies
of poor health. Therefore, a complete test of the ERI were included in the review. A notable study that has
Model covers all three conditions mentioned (i.e., effort, been excluded on basis of those criteria is a study by
reward and overcommitment). Aust, Peter, and Siegrist (1997). Since it is the only
Due to changes in the concepts and their operationa- intervention study based on the ERI Model, we felt this
lizations, studies may differ in their design of the study. study needed to be briefly mentioned. Unfortunately this
For instance, some studies might have included over- study only tested overcommitment, and no effect
commitment into the effort concept, and studies might measures to test the model as such were included;
differ in the labels they used for several concepts. The therefore, the study was excluded from the review.
three hypotheses (i.e., ERI, OVC, and ERI
OVC) Otherwise, studies that were included were categorized
mentioned above will guide the current review of the according to their outcome variables, such that a
ERI Model, as they are consistent with the most recent distinction had been made between physical health
view of the model. However, we are aware that the outcomes, behavioral outcomes and psychological
development of the ERI Model has its influence on well-being. Physical health outcomes encompass cardi-
the study. Therefore, we will systematically consider the ovascular outcomes and other non-CVD-related physi-
amount of support that is found for each hypothesis, cal outcomes. Behavioral outcomes are represented
taking into account the development through time. by behavioral reactions to strain (such as alcohol con-
sumption). Psychological well-being includes general
(psycho)somatic health symptoms and job-related
psychological well-being.
Studies included in the review

Since the current review might be the first review for

Review results
the ERI Model, the present review tries to encompass as
much empirical studies as possible. Inspection of the
Table 1 summarizes the ERI studies for each
ERI studies showed that most studies used (logistic)
hypothesis per outcome category. Note that the ERI
regression analyses in which variables are examined in a
hypothesis has been divided into three different columns
particular composition (for instance including socio-
(i.e., all efforts included into an ERI index regardless of
demographic variables), which complicates revealing the
the type of effort used, an ERI index containing only
unique contribution of the ERI variables (as would be
extrinsic efforts, an ERI index including at least intrinsic
necessary for a meta-analysis). As many studies would
effort—combined with or without extrinsic efforts) to
have to be excluded for a meta-analysis, a review of a
give a more accurate reflection of the specific ERI
narrative nature was preferred, which also allows to take
hypothesis under study. Due to previous definitions of
methodological strengths and weaknesses into account
the ERI Model an ERI index might encompass both
(see also van der Doef & Maes, 1999).
extrinsic and/or intrinsic efforts especially in earlier
With the help of several databases, such as PsycIN-
studies, whereas the current hypotheses state that only
FO, Sociological abstracts, MEDLINE, and ERIC,
extrinsic efforts should be included into the ERI index.
studies testing the ERI Model were identified. Sequen-
A detailed overview of the studies and their findings with
tially, the reference lists of relevant publications were
respect to the sample that has been used, the design of
screened for additional empirical studies. In addition,
the study, the specific outcome measures under study,
several researchers known for their involvement with the
method used, operationalization of ERI (and over-
ERI Model were contacted and asked whether they
commitment), and whether or not support for each
knew some additional ERI studies, which resulted in a
hypothesis of the ERI Model was found (results) is
few other papers.
presented as an Appendix in the online version of this
Inclusion criteria for this review were: (1) publication
paper.
in the period from 1986 to 2003 (inclusive), as the model
was firstly introduced in 1986, (2) inclusion of at least
two core dimensions of the model, either extrinsic effort Physical health outcomes
and reward, or overcommitment and reward, and (3)
inclusion of an effect measure to test the predictions of Physical health outcomes were subdivided into
the ERI Model. Hence, descriptive/theoretical papers as cardiovascular disease (CVD) outcomes and other
well as psychometric papers on the ERI Model were not outcomes. A total of 25 studies could be identified that
included in the review. In addition, (4) studies should be studied the ERI Model in relation to physical health
peer-reviewed scientific published papers (therefore outcomes. More specifically, 24 studies were aimed at
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1121

Table 1
Number of studies included in the review sorted by hypothesis (i.e., ERI, OVC, and interaction) and outcome category

Outcome category Total ERI hypothesis (effort
reward) OVC Interaction hypothesis
(n ¼ 45) hypothesis (effort
reward
OVC)
Alla Extrinsic effortb Remainingc

Physical health outcomes 25*

CVD incidence 8 8 5 3 5 1
CVD symptoms and risk 17 15 13 2 11 3
factors
Other outcomes 1 1 1 0 1 1
Behavioral outcomes 3
Behavioral outcomes 3 3 3 0 1 0
Psychological well-being 19*
(Psycho)somatic health 16 15 13 3 7 3
symptoms
Job-related well-being 7 6 6 0 2 4
*Note that some studies include several types of outcomes and therefore are counted twice, i.e., the sum of studies in the sub-categories
exceeds the number of studies in the main outcome categories (and in a similar way this means that counting the total amount of
studies exceeds 45).
a
All studies that tested the ERI hypothesis (i.e., effort
reward).
b
Only studies that explicitly tested the ERI hypothesis with extrinsic effort and rewards (i.e., extrinsic effort
reward).
c
Remaining studies that tested ERI hypothesis at least with intrinsic (and possibly extrinsic) efforts and rewards (i.e., intrinsic/
extrinsic effort
reward).

CVD outcomes, whereas one study related to other value of the ERI Model [6–8, 18, 21, 22]. Another
outcomes (i.e., cortisol) (see Table 1). feature of the CVD incidence studies was that they
mainly tested the model within a male population [1, 8,
CVD outcomes 18, 21, 22], which restricts generalizability. Especially, a
Since the ERI Model was originally developed to study by Peter et al. [14] which reported different results
study the relation between work stress and CVD, it is for women (increased CVD risk by overcommitment)
not astonishing that most studies include a CVD and men (increased CVD risk by ERI) casts doubt on
outcome measure. In more detail, Table 1 shows that the possibility of generalizing male studies to women.
eight studies examined CVD incidence (i.e., mortality Another point of consideration is the operationalization
and morbidity) and 17 studies verified CVD symptoms/ of ERI itself. Firstly, some studies used the original
risk factors (such as hypertension and cholesterol). questionnaire (i.e., the ERI-Q), whereas other studies
Studies that used self-reported heart disease outcomes used proxy measures (e.g., [6–8, 14]). Secondly, different
were also included in the ‘CVD symptoms and risk calculations of an ERI index have been used, e.g., a ratio
factors’ section. [6, 7, 14, 18, 21] or a median split [8]. Thirdly, in three
CVD incidence: Several similarities appeared between (earlier) studies the concept of ‘‘effort’’ within those ERI
studies including CVD incidence (all numbers refer to indices represented both extrinsic and intrinsic efforts
the studies as listed in the Appendix)). Foremost, all (which is actually overcommitment) [18, 21, 22]. Despite
studies indicated that an ERI at work (i.e., ERI, a high- variation in the operationalization of the ERI concept,
effort–low-reward situation) was positively associated results were consistent, supporting the robustness of the
with the occurrence of coronary events [1, 6–8, 14, 18, relation between ERI and CVD. Actually, a recent study
21, 22]. Moreover, the method used was either logistic showed a reasonable agreement between original and
regression analyses [1, 8, 14, 18, 21, 22] or a Cox proxy measures within the same population, with
proportional hazard model [6, 7]. The odds ratios correlations between both kinds of measures ranging
ranged from 1.22 to 8.98, meaning that employees in a from .37 to .62 (Fahlen, Peter, & Knutsson, 2004). With
high-effort–low-reward situation had an elevated risk of the exception of two studies [6, 8], all studies included a
CVD incidence over 1 to almost 9 times as high as measurement of overcommitment into their study
employees in a reversed working situation (i.e., low (whether or not into an ERI index). Four studies found
effort and high reward). In addition, even though two that high overcommitment was associated with an
studies used a retrospective design [1, 14], most studies increased risk for CVD events [7, 14, 21, 22]. In other
used a prospective design, supporting the predictive words, the risk of getting a CVD or even of dying due to
 ARTICLE IN PRESS
1122 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

a CVD was 1.18–4.53 times as high for highly over- More specifically, in highly overcommitted employees a
committed employees in comparison with low over- higher blood pressure [10, 17], higher LDL cholesterol
committed employees. Appels et al. (1997) also found an (but only for women) [12], and more restenosis [5] was
effect of overcommitment on AMI status, but this effect found, indirectly increasing the risk of getting a CVD.
diminished when ERI was added to the analyses. The Overall, highly overcommitted employees had an
combination of ERI and overcommitment was only elevated risk ranging from 1.37 to 1.86 for CVD risk
studied by Kuper et al. [7], but they found no evidence factors. Further, three studies examined the interaction
that highly overcommitted people showed a particularly between ERI and overcommitment, but none of the
high risk on cardiovascular heart disease due to ERI studies found a significant effect [4, 24, 25]. ERI was
(i.e., a moderating effect). mainly assessed by ratios (cf. [6, 12, 13, 24, 25]) and a
In summary, ERI studies including CVD incidence as direct division into categories (cf. [2, 9, 11, 15, 16, 19,
an outcome were quite similar with regard to design 23]). Again, precaution should be taken in interpreting
(prospective), method (relative risk models), and popu- the ERI indices, as effort may represent intrinsic effort
lation (male). Operationalization of concepts and ERI (i.e., overcommitment) [2] or a mingling of extrinsic and
indicators differed, but did not seem to affect the results intrinsic efforts [23]. In addition, 13 out of 17 studies
substantially. The ERI hypothesis was supported by all included only male respondents in their samples [5, 9–11,
studies with respect to CVD incidence. In other words, 13, 15–17, 19, 20, 23–25]. As with CVD incidence,
employees who reported to work under high efforts another study by Peter et al. [12] showed an elevated risk
without receiving proper rewards actually turned out to for CVD risk for men in case of ERI, whereas women
be more prone to get a CVD, and even died of a CVD, a had an elevated risk due to high overcommitment.
few years later. Most studies that examined the intrinsic Various study designs were used in assessing CVD
OVC hypothesis found support, showing that highly symptoms: a prospective design [2, 5, 6, 9, 10, 17, 20], an
overcommitted employees are more likely to get a CVD experimental design [15, 16], a 1 day subject-within
and/or die of a CVD. A combined effect of ERI and design [4, 24, 25], and a cross-sectional design [11–13, 19,
overcommitment was examined in only one study, but 23, 26]. Results with respect to cross-sectional studies
did not support the interaction (ERI
OVC) hypoth- showed a similar pattern. That is, all cross-sectional
esis. That is, working in an ERI situation as well as studies found an association between ERI and CVD risk
being highly overcommitted at the same time did not factors, and none of those studies found an overall effect
seem to elevate the risk over and above the separate of overcommitment. Otherwise, no considerable pattern
effects of ERI and OVC. in finding (non-)support for the ERI Model seemed to
CVD symptoms and risk factors: A high-effort–low- be present due to study design.
reward imbalance (ERI) was positively associated with To recapitulate, most studies investigated and sup-
CVD symptoms and/or risk factors in 13 out of 17 ported the ERI hypothesis, meaning that a high-
studies [2, 6, 9, 11–13, 15–17, 19, 20, 23, 25]. Two studies effort–low-reward situation at work was associated with
did not test a combination of high effort and low reward reporting more CVD symptoms and/or risk factors. The
simultaneously [5, 10]. Therefore results of ERI are intrinsic overcommitment hypothesis was evaluated in
unknown. The other two studies did contain a measure the majority of studies, and about half of those studies
of ERI, but did not find an effect of ERI on CVD risk [4, found that highly overcommitted employees reported
24]. Those studies solely deviated in the method used to more CVD symptoms. The interaction between ERI and
examine the relation between ERI and CVD risk, i.e., overcommitment with regard to CVD symptoms was
multilevel analyses and general linear models, whereas investigated in just three studies, and none of those three
most other studies used logistic regression analyses. To studies found support for this interaction hypothesis.
put it differently, the statistical method used might have Furthermore, most studies, as with CVD incidence,
influenced the outcomes. Nevertheless, on balance included a male population, which is also inherent to the
spending high efforts at work without receiving a proper CVD outcomes (often being male is one of the selection
(high) reward was related to more CVD symptoms and criteria). With the exception of the cross-sectional
risk factors (such as hypertension and high cholesterol). studies that found consistently support for ERI but
More specifically, employees had an elevated risk for not for overcommitment, results did not show a
CVD symptoms ranging from 1.23 to 6.71 in an ERI particular pattern due to the use of different methods
situation. Remarkably, those two studies, which could and study designs.
not demonstrate an effect of ERI, did show an effect of
overcommitment (on high-frequency heart rate and the Other outcomes
fibrinolytic system, respectively). In addition, four other One study included a physical health outcome that
studies found a positive association between over- could not be classified as a CVD-related outcome.
commitment (i.e., ‘‘immersion’’ and ‘‘vigor’’ which are Hanson et al. [3] studied effort, reward, ERI, over-
dimensions of ‘‘need for control’’) and CVD risk factors. commitment as well as the interaction between ERI and
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1123

overcommitment in relation to cortisol by means of Psychological well-being

physical measurements during 1 day (1 day within-
subject design). The variables were assessed with a In this section ERI studies which measure psycholo-
Dutch version of the original ERI questionnaire, and gical well-being as an outcome category are discussed. A
ERI was operationalized with the help of a ratio. distinction has been made between (1) (psycho)somatic
Multilevel analyses were used to calculate the relations health symptoms and (2) job-related well-being, includ-
between the variables from 77 health care and office ing burnout (or emotional exhaustion), job satisfaction,
employees, but no significant effects were found. So, in and work motivation (cf. the classification of affective
this study cortisol did not seem to be influenced by a well-being by for instance (Warr (1987) or Ryff and
high-effort–low-reward situation, by high overcommit- Keyes (1995)). As can be seen from Table 1, 16 studies
ment, or the combination of ERI and overcommitment. could be identified measuring (psycho)somatic health
symptoms, and seven studies examined job-related well-
Behavioral outcomes being.

Three ERI studies including behavioral outcomes (Psycho)somatic health symptoms

were identified (see Table 1 and Appendix). Firstly, in In this review, (psycho)somatic health symptoms are
addition to hypertension (see ‘‘CVD symptoms and risk generally defined as physical symptoms of which the
factors’’ section), Peter and Siegrist [11] studied the deeper cause is assumed to be psychological of nature.
cross-sectional relation between ERI and (registered) Therefore, the (psycho)somatic health symptoms vary
sickness absence in male middle managers. Single from mental symptoms (e.g., general mental health and
components (frequent interruptions as a high-effort depression) to physical symptoms (such as self-reported
indicator, forced job change as a low-reward indicator, general physical health status and musculoskeletal
and overcommitment) as well as an ERI index symptoms).
composed of three categories were studied. Results Studies with regard to (psycho)somatic health symp-
showed that forced job change (which indicates low toms were quite similar with respect to statistical
reward in their paper) was significantly and positively analyses (logistic regression analyses), measurement
related to sickness absence, whereas working in a high- instrument (proxy questionnaires) and design (cross-
effort–low-reward situation was not significantly related sectional). In terms of design, the studies guided by
to a higher sickness absence. Kuper [7], Stansfeld [41, 42], and Ostry [37] were notable
Secondly, in another–cross-sectional—study by Peter exceptions, since they used longitudinal prospective
et al. among male middle managers [26] smoking was designs. Most studies found that work characterized by
studied in relation to ERI, but not in relation to ERI was positively associated with impaired employee
overcommitment. Effort and reward, as well as an ERI well-being. More specifically, employees working in a
index with three categories (i.e., neither high effort nor high-effort–low-reward situation had an elevated risk of
low reward, either high effort or low reward, both high 1.44–18.55 for (psycho)somatic symptoms. Two studies
effort and low reward) were studied with logistic that were unable to demonstrate this association
regression analysis and revealed that ERI was associated between ERI and (psycho)somatic symptoms were the
with an elevated risk of 4.34 for smoking. studies in charge of Rothenbacher [40] and Joksimovic
Finally, in a cross-sectional study among employees [33] which included the dyspepsia symptom and specific
of metal-working companies Puls et al. [27] found that musculoskeletal disorders, respectively. Possibly, ERI
ERI was related to an increased alcohol consumption. might be stronger related to more general health
ERI was measured by its original questionnaire and outcomes, whereas it is more difficult to find a relation
operationalized as a ratio. Statistical analyses were between ERI and a specific health outcome, such as
conducted by means of path analyses. Overcommitment dyspepsia and specific musculoskeletal disorders. Re-
however was not examined in this study. markable in this respect is that Dragano et al. [30] did
In summary, the ERI hypothesis was supported in find an association between ERI and specific musku-
two out of three studies [26, 27]. More specifically, loskeletal outcomes, although it should be noted that the
working in a high-effort–low-reward job was related to different types of rewards (i.e., esteem, security and
more smoking and an increased alcohol assumption, but promotion) had been measured separately. As such a
not to a higher (registered) sickness absence rate. The similar level of specificity might have been achieved
overcommitment hypothesis was only examined with between rewards and outcomes. ERI was either calcu-
regard to sickness absence, but no relation was found lated by ratios (cf. [7, 30, 33, 35, 38–40, 43]) or by
between overcommitment and sickness absence [11]. The categories [32, 37, 41, 42, 44, 45]. The study by Stansfeld
ERI
OVC interaction hypothesis to test the simulta- et al. [41, 42] slightly deviated as they used intrinsic
neous influence of ERI and OVC on the behavioral effort (i.e., overcommitment) instead of extrinsic effort
outcomes was not examined. to calculate ERI. As main effects of intrinsic effort were
 ARTICLE IN PRESS
1124 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

not studied by them, the effects of OVC itself remain was the case (i.e., reward measured by the original
unknown. Some studies did not include a measure of questionnaire). Results showed that five studies found
overcommitment [38, 39, 45], whereas studies that did an association between ERI and poor job-related well-
measure overcommitment reported different findings. being [28, 31, 32, 44, 45]. Moreover, all those studies
Six studies found a positive relation between over- showed that employees with high effort and low reward
commitment and impaired well-being [7, 30, 33, 35, 40, had an elevated risk for (emotional) exhaustion, whereas
43]. In other words, employees characterized by high some studies also found that ERI was associated with
overcommitment were 1.92–5.92 times as likely to suffer more depersonalization [28] and less job satisfaction [32,
from various (psycho)somatic symptoms (ranging from 44] (in general ORs ranged from 5.49 to 37.37). It should
musculoskeletal disorders to depression), as compared be mentioned that the study by De Jonge and Hamers
to their less overcommitted employees. However, [31] was partially supportive, lending support for ERI in
Calnan et al. [29] found no effect of high OVC on relation to (more) emotional exhaustion, but not for
general health. Since their study was the only study (less) job satisfaction and work motivation. Further-
using linear (instead of logistic) regression analyses, this more, except for one study [45] all studies examined
might be an effect of the statistical method that was overcommitment. Firstly, main effects of overcommit-
used. Furthermore, three studies examined the interac- ment were reported by Calnan et al. [29] for job
tion between ERI and OVC. De Jonge et al. [32] found satisfaction, and by Bakker et al. [28] for emotional
that risks of ERI were aggravated for highly OVC exhaustion and depersonalization. Secondly, the remain-
employees (i.e., ORs ranged from 3.57 to 6.50), whereas ing studies calculated an interaction between ERI and
Kuper et al. [7] and Van Vegchel et al. [44] found no (high) overcommitment resulting in elevated risks
evidence for a moderating effect of OVC on the relation ranging from 5.44 to 20.81 for job-related well-being
between ERI and (psycho)somatic symptoms. [28, 31, 32, 44]. Results varied across studies. Bakker
In short, with regard to the hypotheses of the ERI et al. [28] found an interaction between ERI and
Model, most studies consistently reported a positive high overcommitment for emotional exhaustion and
relation between ERI at work and (psycho)somatic personal accomplishment. De Jonge et al. [32] found
health symptoms. However, it should be mentioned that that risks of more emotional exhaustion and less job
it seemed to be difficult to find effects of ERI on (very) satisfaction due to ERI were higher in overcommitted
specific health outcomes, unless the specific rewards employees, whereas De Jonge and Hamers [31] found no
have analyzed separately. This was however not the case effect of overcommitment on the ERI–outcome associa-
for overcommitment: general as well as specific health tion, and Van Vegchel et al. [44] found no or even
outcomes often showed an elevated risk due to high reversed effects of overcommitment regarding that
overcommitment. Finally, results with respect to the association.
interaction hypothesis (ERI
OVC) were inconsistent. To recapitulate, all well-being studies seemed to fit a
Out of three studies that examined the moderating effect certain profile such that most studies were cross-
of overcommitment on (psycho)somatic health, only one sectional, included a human service population, used
study found an increased risk for psychosomatic health proxy ERI measures, and examined (the moderating
symptoms and physical health for highly overcommitted effect of) overcommitment. Most studies showed that
employees in a high-effort–low-reward situation, as working during ERI was related to poor job-related
opposed to low overcommitted employees in a similar well-being, especially to higher (emotional) exhaustion.
situation. Two studies examined the intrinsic OVC hypothesis and
found that high overcommitment was related to less job
Job-related well-being satisfaction, and more emotional exhaustion and deper-
Burnout, viz. (emotional) exhaustion which is the sonalization. Four out of five studies investigated the
main component of burnout [28, 31, 32, 36, 44, 45], as moderating effect of overcommitment on the relation
well as job satisfaction [29, 31, 32, 36, 44] were most between ERI and job-related well-being, but results were
extensively studied among the ERI studies including job- not consistent. Half of the studies were supportive, while
related well-being (see Appendix). Another well-being the other half were not.
outcome, i.e., work motivation, was included in one
study [31]. Typically, all studies examining job-related
well-being were cross-sectional, included employees Discussion
from human service organizations (e.g., employees from
medical practices and health care employees), and used The present review balanced the results of 45
proxy ERI measures. Although it should be noted that empirical studies published from 1986 up to and
Bakker et al. [28] used the original questionnaire for including 2003 with respect to the ERI Model. Taking
effort and overcommitment (but a proxy measure for the development of the ERI Model into account, the
reward), while for Lewig and Dollard [36] the opposite model was evaluated on the basis of evidence for three
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1125

hypotheses (cf. Siegrist, 2002). Firstly, the (extrinsic) associated with poor employee health. To do justice to
ERI hypothesis which assumes that employees with high the development of the ERI Model, in addition to a
efforts and low rewards at work will have a higher risk ‘‘total’’ confirmative rate of studies supporting the ERI
of poor health. Secondly, the OVC hypothesis which hypothesis, the ERI hypothesis was split up into
maintains that employees with a high level of over- ‘‘extrinsic effort’’ representing the current version of
commitment will also have a higher risk of poor health the model (i.e., high extrinsic effort–low reward) and
(independently of an ERI situation). And, a combina- ‘‘remaining’’ to reflect a previous version of the model
tion of both hypotheses, the interaction (ERI
OVC) (i.e., high intrinsic and/or extrinsic effort–low reward)
hypothesis maintains that employees with both ERI and (see also Figs. 2 and 1, respectively). Eight studies
a high level of overcommitment will have an even higher included intrinsic effort (with or without extrinsic effort)
risk of poor health. A distinction was made between into their ERI index, and also supported the ERI
physical health outcomes (mainly CVD-related out- hypothesis. More specifically, based on outcome cate-
comes), behavioral outcomes (i.e., sickness absence, gory, the review revealed the following findings. For
smoking and alcohol consumption) and psychological physical health outcomes (mainly cardiovascular), the
well-being (i.e., (psycho)somatic health symptoms and combination of high effort and low reward not only
job-related well-being). Table 2 shows the confirmative proved to have explanatory power for the manifestation
rate of the ERI studies for each hypothesis per outcome of CVD (i.e., CVD incidence, 100% support), but also
category. The confirmative rate contains a percentage to be able to identify employees at risk at an earlier stage
that consists of the number of supportive studies divided of CVD development (i.e., CVD symptoms/risk factors,
by the total number of studies that examined the 87% support). The only physiological outcome not
concerning hypothesis. (directly) related to CVD, being cortisol, did not support
As can be seen from Table 2, the ERI hypothesis was the ERI hypothesis. With regard to behavioral out-
most intensively studied (i.e., 48 out of 52 studies; note comes, ERI was associated with increased smoking and
that some studies include several types of outcomes and alcohol consumption but not to an increased sickness
therefore are counted twice, i.e., total amount of studies absence (i.e., 2 out of 3 studies showed support).
exceeds 45) and in general the majority of those studies Psychological well-being seemed to be related to ERI
supported the respective hypothesis, even after extensive as well, supporting the use of the ERI Model beyond
confounder adjustment. In other words, in most studies cardiovascular outcomes. Negative effects of a high-
a combination of high effort and low reward was effort–low-reward situation were found for both

Table 2
Confirmative rate of studies with regard to the ERI hypothesis, the OVC hypothesis and the interaction (ERI
OVC) hypothesis per
outcome category (a study is regarded as confirmative if at least half of the outcome measures in a certain study show support for the
concerning hypothesis based on the ‘‘+’’ in the Appendix)

Outcome category Total ERI hypothesis (effort
reward) OVC Interaction hypothesis
(n=45) hypothesis (effort
reward
OVC)
Alla Extrinsic effortb Remainingc

Physical health outcomes 25*

CVD incidence 8 100% (8/8) 100% (5/5) 100% (3/3) 80% (4/5) 0% (0/1)
CVD symptoms and risk 17 87% (13/15) 85% (11/13) 100% (2/2) 45% (5/11) 0% (0/3)
factors
Other outcomes 1 0% (0/1) 0% (0/1) 0% (0/0) 0% (0/1) 0% (0/1)
Behavioral outcomes 3
Behavioral outcomes 3 67% (2/3) 67% (2/3) 0% (0/0) 0% (0/1) 0% (0/0)
Psychological well-being 19*
(Psycho)somatic health 16 87% (13/15) 85% (11/13) 100% (3/3) 86% (6/7) 33% (1/3)
symptoms
Job-related well-being 7 83% (5/6) 83% (5/6) 0% (0/0) 100% (2/2) 50% (2/4)
*
Note that some studies include several types of outcomes and therefore are counted twice, i.e., the sum of studies in the sub-categories
exceeds the number of studies in the main outcome categories (and in a similar way this means that counting the total amount of
studies exceeds 45).
a
All studies that tested the ERI hypothesis (i.e., effort
reward).
b
Only studies that explicitly tested the ERI hypothesis with extrinsic effort and rewards (i.e., extrinsic effort
reward).
c
Remaining studies that tested ERI hypothesis at least with intrinsic (and possibly extrinsic) efforts and rewards (i.e., intrinsic/
extrinsic effort
reward).
 ARTICLE IN PRESS
1126 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

psychosomatic health symptoms (i.e., 13 out of 15 To recapitulate, the ERI Model seems to be a fruitful
studies) and job-related well-being (5 out of 6). model in order to crystallize the relation between work
Remarkably, ERI seemed less predictive for more characteristics and employee health. In general, the ERI
specific (psycho)somatic health outcomes and especially hypothesis has been intensively examined and most
the job-related outcome exhaustion seemed to be of studies support the notion that a combination of high
importance. This latter finding might lend support for effort and low reward induces an impaired employee
Appel’s notion (1997) that ERI is mediated by vital health. The OVC hypothesis was studied in about half of
exhaustion in explaining CVD. the studies, of which most studies found evidence that
The OVC hypothesis, stating that high overcommit- high-overcommitted employees had an impaired health
ment leads to impaired health and well-being, was compared to their less overcommitted counterparts.
examined in about half of the studies (i.e., 27 out of 52, Even though it should be noted that results varied
see Table 2). Review results varied depending on considerably among outcome categories. The interaction
outcome category. For physical/cardiovascular health hypothesis between ERI and OVC had been hardly
outcomes, most studies found that high overcommit- incorporated into empirical research, which precludes
ment was predictive of a future CVD incident (i.e., 4 out strong conclusions with respect to the overall support of
of 5), and roughly one-half of the studies identified CVD this hypothesis.
symptoms/risk factors based on high overcommitment In a way, it is not surprising that the ERI hypothesis
(i.e., 5 out of 11). Notable is the study by Joksimovic et was studied most often. After all, a high-effort–low-
al. (1999), which showed that overcommitment was also reward imbalance lies at the heart of the ERI Model.
useful in predicting the further course of disease in those Furthermore, from the beginning an ERI was assumed
who already suffered from a manifest CVD. The sole to affect employee well-being (i.e., the ERI hypothesis),
study examining overcommitment in relation to a whereas the role of overcommitment (and its hypoth-
behavioral outcome (i.e., sickness absence) was not eses) evolved over time. Originally, overcommitment
supportive. With respect to psychological well-being, was considered to be part of the effort concept (i.e.,
most studies (i.e., 6 out of 7) showed a relation between intrinsic effort), while in later versions overcommitment
high overcommitment and (psycho)somatic health was assumed to be an independent concept. The erratic
symptoms (varying from particular musculoskeletal role of overcommitment shines through the (empirical)
pains to depression), and two studies that investigated literature as many different roles have been dedicated to
overcommitment in relation to job-related well-being overcommitment. For instance, overcommitment could
both supported the OVC hypothesis as well. be seen as a moderator influencing the relation between
Finally, Table 2 shows the confirmative rate of studies ERI and employee health (e.g., de Jonge, Bosma, Peter,
investigating the interaction (ERI
OVC) hypothesis. & Siegrist, 2000). Overcommitment could also have a
Compared to the other hypotheses, the interaction direct effect on employee health (with or without
hypothesis was scarcely examined (i.e., 12 out of 52 reward). Other possible roles of overcommitment are
studies). Therefore, strong conclusions with regard to that overcommitment directly influences ERI (without
the moderating role of overcommitment on the relation influencing employee health as such). And, as Appels et
between ERI and employee health cannot be made. A al. (1997) assume in their study, overcommitment could
difference did exist between studies with regard to the be a result of ERI, i.e., overcommitment as an outcome.
outcome category. Whereas studies including physical/ All in all, the development of the ERI Model may have
cardiovascular and behavioral outcomes did hardly or contributed to the fact that more attention has been
not at all include a test of the moderating effect, studies devoted to the ERI hypothesis, as compared to the
with regard to psychological well-being (especially with overcommitment part. However, in order to evaluate the
regard to job-related well-being) studied the moderating full ERI Model, future research should contain a test of
role of overcommitment more intensively. None of the effort, reward as well as overcommitment. A good
studies including physical health outcomes supported starting point might be to test the specific hypotheses as
the interaction hypothesis (i.e., 1 study for CVD suggested by Siegrist (2002), being the ERI, the OVC,
incidence, 3 studies for CVD risk, and 1 study for and the ERI
OVC hypothesis.
cortisol). One out of three studies including (psycho)so- Although the ERI hypothesis was most extensively
matic health symptoms showed that risks of an ERI studied, there are still many ways to operationalize the
work situation were aggravated for employees charac- co-occurrence of high effort and low reward. Firstly,
terized by high overcommitment. And exactly half of the depending on the theoretical interpretation of the ERI
job-related well-being studies (i.e., 2 out of 4) supported Model alternative formulations can be postulated. For
the hypothesis that risks of poor well-being (i.e., instance, do high effort and low reward separately have
exhaustion, less personal accomplishment and job an effect on employee health, or should they occur
dissatisfaction) due to ERI were higher in highly together to have a possible influence on employee health,
overcommitted employees. and if so, should there be an interaction effect (i.e., effect
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1127

over and above the sum of the separate effects of effort and working overtime). Several dimensions might be
and reward)? In line with the most recent hypotheses an captured by effort, for instance physical and psycholo-
interaction effect has been mentioned by Siegrist (2002), gical demands. Possibly, in case of some specific
between effort and reward. occupations it might be desirable to split up specific
Secondly, looking in detail at the calculation of the demands. In a similar vein, the concept of reward has
ERI index showed that studies differed considerably. been operationalized in various ways, mostly as one
Even though most studies used ratios or categories for global reward construct including several specific
high/low effort and high/low reward, several ways are rewards. A disadvantage of one global reward indicator
possible to calculate a ratio (continuous or dichoto- (and using different specific rewards to construct this
mized, with or without logarithmic transformation) and indicator) might be that it is not clear whether specific
to handle the resulting quotient. For instance, the rewards can have different effects. Hence, a clear
quotient has been used to make a division between definition of which specific rewards should be included
two groups (i.e., ratio41 as risk group vs. ratiop1 as and analyzing those specific rewards separately might be
non-risk group), three groups (i.e., low, intermediate, a fruitful extension of the ERI Model (Dragano et al.,
high), or four groups. In a similar vein, internal 2003; van Vegchel et al., 2002). Moreover, the present
differences exist for the categories. For example, results review showed that support was found for musculoske-
(continuous or dichotomized) have been divided into letal disorders when specific rewards were included in
two (e.g., cut off point and median split), three (e.g., the analyses (Dragano et al., 2003), whereas no effect
neither high efforts nor low rewards, either high efforts was found when a global reward construct was used
or low rewards, both high efforts and low rewards) or (Rothenbacher, Peter, Bode, Adler, & Brenner, 1998)
four (e.g., low efforts–high rewards, low effort–low (note that both studies used the original questionnaire).
rewards, high efforts–high rewards, high efforts–low The statistical method often involved logistic regres-
rewards) categories. In earlier publications it was sion analyses for both naturally dichotomous and
common to use three categories, but lately the use of (dichotomized) continuous variables. Although analyz-
four categories has been intensified since research has ing a dichotomous outcome variable (such as CVD-
shown the importance of distinguishing the conditions incidence) requires the use of logistic regression ana-
high effort–high reward from low effort–low reward (cf. lyses, there is a risk in dichotomizing continuous
de Jonge et al., 2000). Another definition for ERI, variables. First of all, by dichotomizing continuous
primarily used in the earlier version of the ERI studies, variables information is lost and variance might be
is the co-manifestation of at least one high effort and thrown away. Secondly, the cut off point determines the
one reward indicator. Although this definition might division among categories, making the division depen-
have been formulated for explorational reasons, diffi- dent on the value of the cut off point, which could give a
culties resulting from this are the many possible wrong estimation. The same could be argued for the
compositions of different efforts and rewards enlarging calculation of the ERI index. Most studies use categories
possible chance capitalization. In general, it should be or ratios of which the quotient is divided into categories.
mentioned that the use of cut off points for variables is Therefore, in case of continuous variables the use of
rather arbitrary to create an ERI (or stressful work) continuous analyses (e.g., hierarchical regression ana-
situation, as neither natural nor clinically based thresh- lyses) as well as a continuous ERI index (e.g., a
olds are available thus far (de Jonge et al., 2000). multiplicative interaction term) might be advisable in
Moreover, it is no doubt that the use of different ERI future research.
indices complicates the comparison between studies. ERI studies including behavioral and job-related
Thirdly, on a more general level of measurement the outcomes were rather scarce. In addition, to our
review showed that approximately one-half of the ERI knowledge only one intervention study with respect to
studies used the original questionnaire, whereas the the ERI Model exists (Aust et al., 1997). Intervention
other half used proxy measures. On the one hand, the studies are important to gain more insight into the
use of the original questionnaire simplifies comparison practical value of the ERI Model, and are therefore
across studies and relates as closely as possible to the highly recommended (cf. Kompier, Geurts, Gründe-
original concepts as meant by the model (cf. Beehr, mann, Vink, & Smulders, 1998).
Glaser, Canali, & Wallwey, 2001; Schnall, Landsbergis, In general, study designs were either prospective (i.e.,
& Baker, 1994). On the other hand, studies using longitudinal) or cross-sectional, and the proportion of
original as well as proxy measures found support for the longitudinal vs. cross-sectional studies differed by out-
ERI Model, indicating an effect of ERI regardless of the come measures. Studies examining CVD incidence were
measure being used, strengthening the robustness of the mainly prospective, strengthening the predictive value of
model. In addition, the concept of effort in the original the ERI Model. On the other hand, studies including
questionnaire is composed of a variety of items of which CVD symptoms/risk factors also used cross-sectional
the content differs (such as physical load, time pressure designs. Remarkably, all those cross-sectional studies
 ARTICLE IN PRESS
1128 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

supported the ERI assumption, but none of them ERI Model is on the right track to help explain the work
supported the OVC assumption. Possibly, ERI and stress phenomenon, the road is never ending. There is
overcommitment could have different time lagged effects still much to be done as we continuously have to learn
such that working during an ERI has short term effects about the particular relationship between job character-
at least in terms of CVD symptoms, whereas over- istics and employee health and well-being.
commitment is a long term determinant of CVD
symptoms. With regard to behavioral outcomes and
psychological well-being mainly cross-sectional designs Acknowledgement
were used. However, to examine causal patterns
preferably with different time lags between effort, The current study is part of the research program on
reward, overcommitment and health more longitudinal ‘‘Fatigue at Work’’ by NWO, the Dutch Organisation
studies are needed (e.g., see Taris & Kompier, 2003). In for Scientific Research (Grant 580-02-209).
addition, with the help of longitudinal studies the causal
direction of the relation between ERI variables (i.e.,
effort, reward and overcommitment) and employee Appendix A. Supplementary analysis tools
health could be further examined. For instance, does
an ERI cause poor employee health, or do employees A collection of macros that can be applied to sequence
with a poor health report more efforts and less rewards batches of up to 5000 sequences can be obtained from
(i.e., reversed causation), or do they mutually influence our web site (http://igm.ccc.uab.edu/~schroeder/publi-
each other (i.e., reciprocal relation: cf. de Jonge et al., cations/).
2001). In a similar vein, the role of overcommitment
(being a personal characteristic) could be further
examined in relation to ERI and employee health. Appendix B. Supplementary data
And finally, to establish the generalizability of the
ERI Model, research should be expanded to an even The online version of this article contains additional
more diverse array of respondents. The present review supplementary data. Please visit doi:10.1016/j.socscimed.
showed that research including CVD-related outcomes 2004.06.043.
predominantly used male populations, which hampers
generalizing results to a female population. This is
particularly important, given two studies by Peter et al.
References
(1998, 2002) that report different results for men (i.e.,
CVD risk due to ERI) and women (i.e., CVD risk due to Appels, A., & Schouten, E. (1991). Burnout as a risk factor for
OVC). This possibly reflects the multi-faceted nature of coronary heart disease. Behavioral Medicine, 17(2), 53–59.
the ERI Model for men and women. Furthermore, Appels, A., Siegrist, J., & de Vos, J. (1997). ‘Chronic workload’,
studies with regard to job-related well-being all included ‘need for control’ and ‘vital exhaustion’ in patients with
a population from the human service sector. Although myocardial infarction and control: a comparative test of
different occupations have been studied within those cardiovascular risk profiles. Stress Medicine, 13, 117–121.
human service professions, studying different popula- Aust, B., Peter, R., & Siegrist, J. (1997). Stress management in
tions would be recommendable in order to generalize bus drivers: a pilot study based on the model of effort–re-
ward imbalance. International Journal of Stress Manage-
results to non-human service occupations. In addition,
ment, 4, 297–305.
the vast majority of ERI studies have been conducted in
Bakker, A. B., Killmer, C. H., Siegrist, J., & Schaufeli, W. B.
West-European countries. Three notable exceptions are (2000). Effort–reward imbalance and burnout among
a study by Pikhart et al. (2001) in five post-communistic nurses. Journal of Advanced Nursing, 31(4), 884–891.
countries, a Japanese study by Tsutsumi, Kayaba, Beehr, T. A., Glaser, K. M., Canali, K. G., & Wallwey, D. A.
Theorell, and Siegrist (2001), and an Australian study (2001). Back to basis: re-examination of demand-control
by Lewig and Dollard (2003). In order to establish theory of occupational stress. Work and Stress, 15(2),
generalizability and cross-cultural differences, the ERI 115–130.
Model should be examined in different (non-West- Blum, K., Cull, J. G., Braverman, E. R., & Comings, D. E.
European) countries as well. (1996). Reward deficiency syndrome: addictive, impulsive
and compulsive disorders—including alcoholism, attention-
At the end, it should be noted that a review will
deficit disorders, drug abuse and food bingeing—may have
always be limited to papers that have been published.
a common genetic basis. American Scientist, 84, 132–145.
The present review tried to encompass as much studies Bosma, H., Peter, R., Siegrist, J., & Marmot, M. (1998). Two
as possible, being unaware of results of unpublished alternative job stress models and the risk of coronary heart
papers. To draw the balance, in 18 years of empirical disease. American Journal of Public Health, 88, 68–74.
research the ERI Model has proven to be a valuable Calnan, M., Wainwright, D., & Almond, S. (2000). Job strain,
contribution to occupational health research. While the effort–reward imbalance and mental distress: a study of
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1129

occupations in general medical practice. Work and Stress, prevention: the success of a participative and stepwise
14(4), 297–311. approach. Stress Medicine, 14, 155–168.
van der Doef, M., & Maes, S. (1999). The job demand-control Kuper, H., Singh-Manoux, A., Siegrist, J., & Marmot, M.
(-support) model and psychological well-being: a review of (2002). When reciprocity fails: effort–reward imbalance in
20 years of empirical research. Work and Stress, 13, 87–114. relation to coronary heart disease and health functioning
Dragano, N., Von dem Knesebeck, O., Rodel, A., & Siegrist, J. within Whitehall II study. Occupational Environment
(2003). Psychosoziale Arbeitsbelastungen und muskulo- Medicine, 59, 777–784.
skeletale Beschwerden: Bedeutung für die Prävention Larisch, M., Joksimovic, L., von dem Knesebeck, O., Starke,
(Psychosocial work stress and musculoskeletal diseases: D., & Siegrist, J. (2003). Berufliche Gratifikationskrisen und
Significance for prevention). Zeitschrift für Gesundheitswis- depressive Symptome: Eine Querschnittstudie bei Erwerb-
senschaften, 11, 196–207. stätigen im mittleren Erwachsenenalter (Effort–reward
Fahlen, G., Peter, R., & Knutsson, A. (2004s). The effort–re- Imbalance at work and depressive symptoms: A cross-
ward imbalance model of psychosocial stress at the work- sectional investigation of middle-aged employees). Psy-
place—a comparison of exposure measurement using two chotherapie, Psychosomatik, Medizinische Psychologie, 53,
estimation methods. Work and Stress, 18(1), 81–88. 223–228.
Gaillard, A. W. K., & Wientjes, C. J. E. (1994). Mental load Lazarus, R. (1991). Emotion and adaptation. New York: Oxford
and work stress as two types of energy mobilization. Work University Press.
and Stress, 8, 141–152. Lewig, K. A., & Dollard, M. F. (2003). Emotional dissonance,
Hanson, E. K. S., Godaert, G. L. R., Maas, C. J. M., & emotional exhaustion and job satisfaction in call centre
Meijman, T. F. (2001). Vagal cardiac control throughout workers. European Journal of Work and Organizational
the day: the relative importance of effort–reward imbalance Psychology, 12(4), 366–392.
and within-day measurements of mood, demand and Lynch, J., & Krause, N. (1997). Workplace conditions, socio-
satisfaction. Biological Psychology, 56, 23–44. economic status, and the risk of mortality and acute
Hanson, E. K. S., Schaufeli, W., Vrijkotte, T., Plomp, N. H., & myocardial infarction: the kuopio heart disease risk factor
Godaert, G. L. R. (2000). The validity and reliability of the study. American Journal of Public Health, 87(4), 617–623.
Dutch effort–reward imbalance questionnaire. Journal of Lynch, J., Krause, N., Kaplan, G. A., Salonen, R., & Salonen,
Occupational Health Psychology, 5(1), 142–155. J. T. (1997). Workplace demands, economic reward, and
Henry, J. P., & Stephens, P. M. (1977). Stress, health, and the progression of carotid atherosclerosis. Circulation, 96(1),
social environment. Berlin, Germany: Springer. 302–307.
Joksimovic, L., Siegrist, J., Meyer-Hammer, M., Peter, R., Marmot, M., Siegrist, J., Theorell, T., & Feeney, A. (1999).
Franke, B., Klimek, W. J., Heintzen, M. P., & Strauer, B. E. Health and the psychosocial environment at work. In M.
(1999). Overcommitment predicts restenosis after coronary Marmot, & R. G. Wilkinson (Eds.). Social determinants of
angioplasty in cardiac patients. International Journal of health (pp. 105–131). Oxford: Oxford University Press.
Behavioral Medicine, 6(4), 356–369. Matschinger, H., Siegrist, J., Siegrist, K., & Dittmann, K. H.
Joksimovic, L., Starke, D., von dem Knesebeck, O., & Siegrist, (1986). Type A as a coping career—toward a conceptual
J. (2002). Perceived work stress, overcommitment, and self- and methodological redefinition. In T. H. Schmidt, T. M.
reported musculoskeletal pain: a cross-sectional investiga- Dembroski, & G. Blümchen (Eds.). Biological and psycho-
tion. International Journal of Behavioral Medicine, 9(2), logical factors in cardiovascular disease (pp. 104–126).
122–138. Berlin: Springer.
de Jonge, J., Bosma, H., Peter, R., & Siegrist, J. (2000). Job Niedhammer, I., Siegrist, J., Landre, M. F., Goldberg, M., &
strain, effort–reward imbalance and employee well-being: a Leclerc, A. (2000). Etude des qualites psychometriques de la
large-scale cross-sectional study. Social Science & Medicine, version francaise du modele du Desequilibre Efforts/
50, 1317–1327. Recompenses (Psychometric properties of the French
de Jonge, J., Dormann, C., Janssen, P. P. M., Dollard, M. F., version of the Effort–reward Imbalance model). Revue
Landeweerd, J. A., & Nijhuis, F. J. N. (2001). Testing d’epidemiol et de sante publique, 48(5), 419–437.
reciprocal relationships between job characteristics and Ostry, A. S., Kelly, S., Demers, P. A., Mustard, C., &
psychological well-being: a cross-lagged structural equation Hertzman, C. (2003). A comparison between the effort–re-
model. Journal of Occupational and Organizational Psychol- ward imbalance and demand-control models (on-line). BMC
ogy, 74, 29–46. Public Health, retrieved, from the World Wide Web: http://
Kivimäki, M., Leino-Arjas, P., Luukkonen, R., Riihimäki, H., www.biomedcentral.com/1471-2458/10.
Vahtera, J., & Kirjonen, J. (2002). Work stress and risk of Peter, R., Alfredsson, L., Hammar, N., Siegrist, J., Theorell, T.,
cardiovascular mortality: prospective cohort study of & Westerholm, P. (1998). High effort, low reward, and
industrial employees. British Medical Journal, 325(7369), cardiovascular risk factors in employed Swedish men and
857–861. women: baseline results from the WOLF Study. Journal of
von dem Knesebeck, O., & Siegrist, J. (2003). Reported Epidemiology and Community Health, 52, 540–547.
nonreciprocity of social exchange and depressive symptoms: Peter, R., Alfredsson, L., Knutsson, A., Siegrist, J., &
extending the model of effort–reward imbalance beyond Westerholm, P. (1999). Does a stressful psychosocial work
work. Journal of Psychosomatic Research, 55, 209–214. environment mediate the effects of shift work on cardio-
Kompier, M. A. J., Geurts, S. A. E., Gründemann, R. W. M., vascular factors? Scandinavian Journal of Work, Environ-
Vink, P., & Smulders, P. G. W. (1998). Cases in stress ment and Health, 25(4), 376–381.
 ARTICLE IN PRESS
1130 N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131

Peter, R., GeiXler, H., & Siegrist, J. (1998). Associations of Siegrist, J., & Klein, D. (1990). Occupational stress and
effort–reward imbalance at work and reported symptoms in cardiovascular reactivity in blue-collar workers. Work and
different groups of male and female public transport Stress, 4, 295–304.
workers. Stress Medicine, 14, 175–182. Siegrist, J., Klein, D., & Voigt, K. -H. (1997a). Linking
Peter, R., & Siegrist, J. (1997). Chronic work stress, sickness sociological with physiological data: the model of effort–re-
absence, and hypertension in middle managers: general of ward imbalance at work. Acta Physiologica Scandinavia,
specific sociological explanations? Social Science & Medi- 16(suppl. 640), 112–116.
cine, 45, 1111–1120. Siegrist, J., & Matschinger, H. (1989). Restricted status control
Peter, R., Siegrist, J., Hallqvist, J., Reuterwall, C., & Theorell, and cardiovascular risk. In A. Steptoe, & A. Appels (Eds.).
T., SHEEP Study Group (2002). Psychosocial work Stress, personal control and health (pp. 65–82). Chichester:
environment and myocardial infarction, Improving risk Wiley.
estimation by combining two complementary job stress Siegrist, J., Matschinger, H., Cremer, P., & Seidel, D. (1988).
models in the SHEEP Study. Journal of Epidemiology and Atherogenic risk in men suffering from occupational stress.
Community Health, 56, 294–300. Atherosclerosis, 69, 211–218.
Peter, R., Siegrist, J., Stork, J., Mann, H., & Labrot, B. (1991). Siegrist, J., & Peter, R. (1994). Job stressors and coping
Zigarettenrauchen und pscyhosoziale Arbeitsbelastung bei characteristics in work-related disease: issues of validity.
Beschäftigen desmittleren Managements (Cigarette smoking Work and Stress, 8, 130–140.
and psychosocial work stress in middle managers). Sozial Siegrist, J., & Peter, R. (1996a). Measuring effort–reward
und Präventivmedizin, 36, 315–321. imbalance at work: guidelines. Düsseldorf: Heinrich Heine
Pikhart, H., Bobak, M., Siegrist, J., Pajak, A., Rywik, S., University.
Kyshegyi, J., Gostautas, A., Skodova, Z., & Marmot, M. Siegrist, J., & Peter, R. (1996b). Threat to occupational status
(2001). Psychosocial work characteristics and self rated control and cardiovascular risk. Israel Journal of Medical
health in four post-communist countries. Journal of Sciences, 32, 179–184.
Epidemiology and Community Health, 55, 624–630. Siegrist, J., Peter, R., Cremer, P., & Seidel, D. (1997b). Chronic
Puls, W., Wienold, H., & Blank, T. (1998). Die einwirkung van work stress is associated with lipids and elevated fibrogen in
Gratifikationskrisen am Arbeitsplatz auf den Konsum von middle-aged men. Journal of International Medicine, 242,
Alkohol: Eine schriftliche Befragung in Betrieben der 149–156.
metallverarbeitenden Industrie (The influence of Effort–re- Siegrist, J., Peter, R., Junge, A., Cremer, P., & Seidel, D. (1990).
ward Imbalance in the workplace on the consumption of Low status control, high effort at work and ischemic heart
alcohol: a written survey carried out in metal-working disease: prospective evidence from blue-collar men. Social
companies). Sucht, 44, 183–199. Science & Medicine, 31(10), 1127–1134.
Rothenbacher, D., Peter, R., Bode, G., Adler, G., & Brenner, Siegrist, J., Peter, R., Motz, W., & Strauer, B. E. (1992). The
H. (1998). Dyspepsia in relation to helicobacter pylori role of hypertension, left ventricular hypertrophy and
infection and psychosocial work stress in white collar psychosocial risk in cardiovascular disease: prospective
employees. The American Journal of Gastroenterology, evidence from blue-collar men. European Heart Journal,
93(9), 1443–1449. 13(suppl. D), 89–95.
Ryff, C. D., & Keyes, C. L. M. (1995). The structure of Siegrist, J., Siegrist, K., & Weber, I. (1986). Sociological
psychological well-being revisited. Journal of Personality concepts in the etiology of chronic disease: the case of
and Social Psychology, 69(4), 719–727. ischemic heart disease. Social Science & Medicine, 22,
Schnall, P. L., Landsbergis, P. A., & Baker, D. (1994). Job 247–253.
strain and cardiovascular disease. Annual Review of Public Siegrist, J., Starke, D., Chandola, T., Godin, I., Marmot, M.,
Health, 15, 381–411. Niedhammer, I., & Peter, R. (2004). The measurement of
Schönpflug, W., & Batman, W. (1989). The costs and benefits of effort–reward imbalance at work: European comparisons.
coping. In S. Fisher, & J. Reason (Eds.). Handbook of stress, Social Science & Medicine, 58, 1483–1499.
cognition and health (pp. 699–714). Chichester: Wiley. Stansfeld, S., Bosma, H., Hemingway, H., & Marmot, M.
Siegrist, J. (1996). Adverse health effects of high-effort/low- (1998). Psychosocial work characteristics and social support
reward conditions. Journal of Occupational Health Psychol- as predictors of SF-36 health functioning: the Whitehall II
ogy, 1, 27–41. study. Psychosomatic Medicine, 60(3), 247–255.
Siegrist, J. (1998). Adverse health effects of effort–reward Stansfeld, S. A., Fuhrer, R., Shipley, M., & Marmot, M. (1999).
imbalance at work: theory, empirical support, and implications Work characteristics predict psychiatric disorder: prospec-
for prevention. In C. L. Cooper (Ed.). Theories of organiza- tive results from the Whitehall II study. Occupational
tional stress (pp. 190–204). Oxford: Oxford University Press. Environment and Medicine, 56, 302–307.
Siegrist, J. (1999). Occupational health and public health in Taris, T. W., & Kompier, M. (2003). Challenges in longitudinal
Germany. In P. M. Le Blanc, M. C. W. Peeters, A. Büssing, designs in occupational health psychology. Scandinavian
& W. B. Schaufeli (Eds.). Organizational psychology and Journal of Work, Environment and Health, 29(1), 1–4.
healthcare: European contributions (pp. 35–44). München: Tsutsumi, A., Kayaba, K., Theorell, T., & Siegrist, J. (2001).
Rainer Hampp Verlag. Association between job stress and depression among
Siegrist, J. (2002). Effort-reward imbalance at work and health. Japanese employees threatened by job loss in a comparison
In P. L. Perrewé, & D. C. Ganster (Eds.). Historical and between two complementary job-stress models. Scandina-
current perspectives on stress and health (pp. 261–291). vian Journal of Work, Environment and Health, 27(2),
Amsterdam: JAI Elsevier. 146–153.
 ARTICLE IN PRESS
N. van Vegchel et al. / Social Science & Medicine 60 (2005) 1117–1131 1131

van Vegchel, N., de Jonge, J., Bakker, A. B., & Schaufeli, W. B. Vrijkotte, T. G. M., van Doornen, L. J. P., & de Geus, E. J. C.
(2002). Testing global and specific indicators of rewards in (1999). Work stress and metabolic and hemostatic risk
the Effort–Reward Imbalance Model: does it make any factors. Psychosomatic Medicine, 61, 796–805.
difference? European Journal of Work and Organizational Vrijkotte, T. G. M., van Doornen, L. J. P., & de Geus, E. J. C.
Psychology, 11(4), 403–421. (2000). Effects of work stress on ambulatory blood pressure,
van Vegchel, N., de Jonge, J., Meijer, T., & Hamers, J. P. H. heart rate, and heart variability. Hypertension, 35, 880–886.
(2001). Different effort constructs and effort–reward Warr, P. (1987). Work, unemployment and mental health.
imbalance: effects on employee well-being in ancillary Oxford: Clarendon Press.
health care workers. Journal of Advanced Nursing, 34(1), Weiner, H. (1992). Perturbing the organism: the biology of
128–136. stressful experience. Chicago: University of Chicago Press.