See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/8074162

Working Conditions and Three Types of Well-Being: A

Longitudinal Study With Self-Report and Rating Data

Article in Journal of Occupational Health Psychology · January 2005

DOI: 10.1037/1076-8998.10.1.31 · Source: PubMed

CITATIONS READS
169 1,958

3 authors:

Simone Grebner Norbert Semmer

Universität Bern Universität Bern
63 PUBLICATIONS 1,797 CITATIONS 255 PUBLICATIONS 15,047 CITATIONS

SEE PROFILE SEE PROFILE

Achim Elfering
Universität Bern
264 PUBLICATIONS 11,143 CITATIONS

SEE PROFILE

All content following this page was uploaded by Norbert Semmer on 20 May 2014.

The user has requested enhancement of the downloaded file.

Journal of Occupational Health Psychology Copyright 2005 by the Educational Publishing Foundation
2005, Vol. 10, No. 1, 31– 43 1076-8998/05/$12.00 DOI: 10.1037/1076-8998.10.1.31

Working Conditions and Three Types of Well-Being:

A Longitudinal Study With Self-Report and Rating Data
Simone Grebner, Norbert K. Semmer, and Achim Elfering
University of Bern

The mean of self-report and observer ratings of working conditions was used to predict 3 types
of well-being in 52 young workers: general well-being, job-related well-being, and spillover from
work to nonwork domains. Longitudinally, job control predicted spillover. There was no strong
evidence for reverse causation. Synchronously, Time 2 job stressors predicted all types of
well-being, and job control predicted general well-being. Because dependent variables at Time 1
are controlled for, this indicates short-term effects. Results for stressors are in line with a stress
reaction model, indicating a rather quick symptom development but reversibility. The effect of
control on spillover, however, suggests a sleeper effect model, with symptoms appearing with
delay.

Stressors at work are a possible source of impaired In studies of occupational stress, it is useful to
well-being, health, and performance (e.g., Ganster & distinguish between general, or context-free, well-
Schaubroeck, 1991; Kahn & Byosiere, 1992; Mar- being (such as psychosomatic complaints) and job-
mot, Siegrist, Theorell, & Feeney, 1999; Sonnentag related well-being (such as job satisfaction; Warr,
& Frese, 2003; Theorell & Karasek, 1996). Impaired 1999). Furthermore, the literature on “unwinding”
well-being implies suffering for the individual, but it suggests that a third type of well-being may be of
is also is a source of costs for organizations and special importance, namely the spillover of work
society, such as health care costs (Ganster, Fox, & problems into private life, as indicated by the inabil-
Dwyer, 2001; Manning, Jackson, & Fusilier, 1996). ity to relax after work or by exhaustion after work
Moreover, impaired well-being can be considered as (Frankenhaeuser, 1991; McEwen, 1998; Meijman &
an early warning system for health risks; it may be a Mulder, 1998; Mohr, 1991). While stressors are as-
preliminary stage of more severe mental problems sociated with impaired well-being and health, the
(Mohr, 1991) and a predictor of somatic illness opposite applies to job control, that is, the possibility
(Hemingway & Marmot, 1999; Kiecolt-Glaser, of influencing one’s own work (e.g., Marmot,
McGuire, Tobles, & Glaser, 2002; Marmot et al., Bosma, Hemingway, Brunner, & Stansfeld, 1997;
1999; Räikkönen, Matthews, Flory, Owens, & Semmer, 1998; Terry & Jimmieson, 1999).
Gump, 1999; Rozanski, Blumenthal, & Kaplan, Despite some recent progress, occupational stress
1999; Steptoe, 2001). research is hampered by a number of methodological
problems. Thus, the majority of studies have used
cross-sectional designs (Zapf, Dormann, & Frese,
Simone Grebner, Norbert K. Semmer, and Achim Elfer- 1996). This implies that the alternative explanation of
ing, Department of Psychology, University of Bern, Bern, reverse causation often cannot be ruled out. There-
Switzerland. fore, stressor–strain relationships may be overesti-
This research was funded by the Swiss National Science mated because healthier people get better jobs in the
Foundation, SPP “Demain la Suisse” for the project “Work-
ing Experience and Quality of Life in Switzerland” (an long run (the drift hypothesis; see Zapf et al., 1996),
in-depth study of “Stressful Experiences at Work”; principal or they may be underestimated because people who
investigator: Norbert K. Semmer). have developed symptoms of ill health might quit
We are indebted to Nadine Vogel, Daniel Birkhofer, their job or be assigned to less stressful activities (the
Claudia Kaiser-Probst, Marie-Louise Schlapbach, Nicola
Jacobshagen, Tanja Perrot, Manuela Füglister, Sascha Na- healthy worker effect or refuge effect; see Garst,
varra, Beatrice Schenker, and Pascale Huguenin, who col- Frese, & Molenaar, 2000; Zapf et al., 1996).
lected the data. We thank in particular Nadine Vogel for her Furthermore, most of the field studies of occupa-
valuable assistance throughout this project. tional stress use a monomethod approach, usually
Correspondence concerning this article should be ad-
dressed to Simone Grebner, Department of Psychology,
based on self-reports (see Kahn & Byosiere, 1992).
University of Bern, Muesmattstrasse 45, CH-3000 Bern 9, This has raised concern about inflated correlations
Switzerland. E-mail: [email protected] due to common method variance (e.g., Kasl, 1986;

31
32 GREBNER, SEMMER, AND ELFERING

Kessler, 1987; Zapf et al., 1996). This effect is by no may be periods of heightened vulnerability that pro-
means automatic, and underestimations of true asso- vide good opportunities to start longitudinal projects.
ciations may occur as well (Spector & Brannick, Transition into work often is regarded as such a
1995; Williams & Brown, 1994). In any case, com- critical period (see Barling & Kelloway, 1999; Frese,
mon method variance may distort true associations 1982; Frone, 2000; Loughlin & Barling, 2001; Saks
(Williams & Brown, 1994). & Ashforth, 2000). This period seems to be associ-
There are some, but not many, studies that use ated with less disruption and more continuity than
measures other than self-report, such as observer one might expect (Arnold, 1997; Arnold & Reicherts,
ratings (e.g., Greiner, Krause, Ragland, & Fisher, 2000), and it may well imply positive developments
1998; Semmer, Zapf, & Greif, 1996; Spector, 1992; (e.g., increasing job control and higher status; Kälin
Spector, Zapf, Chen, & Frese, 2000; Spector & Jex, et al., 2000). Nevertheless, this transition still yields
1998). Note, however, that such measures should not a good opportunity to capture early symptom devel-
be regarded as alternatives that can replace self- opment. Because there is no long history of work, it
report, based on the assumption that they are auto- is less likely that, owing to work experiences, symp-
matically better or more “objective” (Semmer, Greb- toms will have developed that have already become
ner, & Elfering, 2004; Spector, 1992; Spector et al., chronic.
2000). Like all measures, they are subject to mea- The present study investigated young people at the
surement error. Because measurement error in exter- beginning of their working life. We used a multi-
nal ratings is less likely to be correlated with the error method approach, in that working conditions—task-
in outcome measures, observer ratings often lead to related stressors and control—are assessed both by
an underestimation of relations (Semmer et al., 1996; self-report and by ratings of trained observers. We
Spector, 1992; Zapf, 1989). Typically, therefore, as- examined relations between stressors and job control,
sociations between well-being and work characteris- on the one hand, and several different indicators of
tics tend to be lower when work characteristics are well-being, including context-free, job-related, and
assessed by observer ratings as compared with self- spillover-oriented measures of well-being, on the
report, although there are some exceptions. The pat- other. The design was longitudinal, enabling us to
tern of associations, however, tends to be similar examine both longitudinal and synchronous effects
(Frese, 1985; Semmer, Zapf, & Dunckel, 1995; Spec- and to test for both directions of causality.
tor, 1992; Zapf, 1989).
A combination of methods therefore seems most
adequate (Bailey & Bhagat, 1987). Multitrait, multi-
Method
method models based on such combinations have Sample
shown that despite modest correlations between self-
report and observer ratings, both contain valuable The analyses are based on a two-wave field study of 93
information, and the relation to well-being is still male and female participants from five occupations (nurses,
substantial after removing method variance (Semmer cooks, sales persons, bank clerks, and electronic techni-
et al., 1996; Spector, 1992; Zapf, 1989). cians), representing a subsample of the German-speaking
participants of the project AEQUAS (Working Experiences
One blind spot in occupational stress research con- and Quality of Life in Switzerland; see Kälin et al., 2000).
cerns the question as to when one can expect which The first measurement (Time 1) took place in the first
outcome to develop as a consequence of stress (see months of the participants’ professional life after finishing
Frese & Zapf, 1988; McGrath & Beehr, 1990). This vocational training. The second wave (Time 2) was con-
ducted 1 year later. At Time 1, the 93 participants worked in
relates to the optimal time to start such an investiga- 71 low-, medium- and large-size organizations located in
tion. Ideally, a study would begin before work-re- rural and urban areas of the German-speaking part of Swit-
lated stress symptoms have developed. However, zerland. Questionnaires were filled in by 86 (Time 1) and 65
most studies that concentrate on adults start when this (Time 2) participants. The questionnaire sample of Time 1
development may already have set in (see Frone, contains 39 nurses, 12 cooks, 10 sales assistants, 12 bank
clerks, and 13 electronic technicians. Seventy-four (Time 1)
2000). To the extent that symptoms develop some and 54 (Time 2) of them could be observed at work. Data
stability of their own (Frese & Zapf, 1988), this will from both methods are available for 69 (Time 1) and 52
not allow detecting associations longitudinally, even participants (Time 2). Therefore, the longitudinal sample
if working conditions have played a role in symptom consists of 52 participants. Mean age was 22 years (SD ⫽
3.3) at Time 1. Fifty-eight women and 28 men took part at
development. Although such an argument would im- Time 1, and 46 women and 19 men at Time 2. Informed
ply that our studies should start very early and last consent of participants (and supervisors in the case of ob-
very long, from a more pragmatic viewpoint there servation at work) was obtained.
 WORKING CONDITIONS AND WELL-BEING 33

Measures instrument. Correlations between ratings of two indepen-

dent observers reported by Semmer et al. (1995) are mostly
Working conditions. Working conditions were mea- in the range of .55 (concentration and time pressure) to .87
sured through self-report and through ratings by trained (control); only one scale had a coefficient below .50 (un-
observers. A short self-report version of the Instrument for certainty: r ⫽ .44).
Stress Oriented Task Analysis (ISTA) was used, along with Well-being. Well-being was assessed in terms of (a)
the unabridged rating version (Semmer et al., 1995). All general or context-free well-being, (b) job-related well-
items have a 5-point Likert format, reflecting either intensity being, and (c) spillover from work into nonwork domains.
or frequency. Job control captures aspects of method con- Of the two indicators for context-free well-being, one was a
trol (e.g., independently plan and organize one’s own work) measure of psychosomatic complaints , developed by Mohr
and time control (e.g., influence on work pace and sched- (1991; see also Mohr, 2000) based on Fahrenberg (1975). It
ule). Five task stressors, or “regulation problems” (Frese & asks for frequency of headaches, stomachaches, nervous-
Zapf, 1994), were assessed: time pressure, concentration ness, and so on during the preceding year. It is comparable
demands, problems of organization of work (e.g., having to with many similar scales used in this type of research (e.g.,
work with inadequate devices or obsolete information), the Physical Symptoms Inventory by Spector & Jex, 1998).
uncertainty (e.g., unclear instructions or decisions based on It has been used in a variety of studies on stress at work in
insufficient information), and work interruptions by super- German-speaking countries (e.g., Büssing, 1999; Frese,
visors, colleagues, or clients. To enhance clarity of the 1985; Garst et al., 2000). The second indicator of general
presentation, and to avoid the danger of obtaining statisti- well-being was irritated reactions (e.g., “I am easily an-
cally significant effects by chance due to a large number of noyed”), measured with a subscale of Mohr’s (1991) scale
statistical tests on a rather small sample, we combined the on Irritation/Strain (see Garst et al., 2000).
five stressors into a single index in the analyses (see Frese, Job-related well-being was measured by job satisfaction.
1985, for a similar procedure). The scale contains three items developed by Oegerli (1984)
Ratings were based on an initial observation period of 60 plus a Kunin Faces scale. It has been shown to be a good
min, plus three additional time samples of 15 min each. predictor of turnover (Baillod & Semmer, 1994; Semmer,
Observation was complemented by an interview with the Baillod, Stadler, & Gail, 1996). A second indicator is re-
jobholder (30 – 45 min) as well as questions to supervisor(s) signed attitude toward one’s job. It is based on the concept
and colleague(s), and consultation of organizational docu- of resigned job satisfaction (Bruggemann, 1974; for an
ments, if necessary. The interviews were semistructured, English description, see Büssing, 1992). Items are again
based on an interview guideline containing key questions from Oegerli (1984) and ask how often one has thoughts
for each dimension. Their purpose was to obtain an over- such as “My job is not ideal, but it could be worse,”
view of activities involved in a given job, to enhance suggesting a defensive or resentful adaptation to working
comprehension of what is observed, to verify impressions of conditions that are not ideal (Semmer, 2003; see also Kälin
the observers, and to help interpret actions with components et al., 2000).
that are not directly observable. While interviews with job- Spillover from work into nonwork domains was again
holders are often unavoidable (see Morsh, 1964), observers assessed by two indicators. The first is a second subset of
were trained not to accept statements made by jobholders at Mohr’s (1991) Irritation/Strain scale. It refers to rumination
face value but rather to use them as one source of informa-
about work problems and the inability to switch one’s mind
tion. Together with their own observations and, in cases of
off of job-related matters after work (e.g., “It is hard for me
ambiguities, information given by superiors, colleagues, or
to switch off my mind after work”). Garst et al. (2000)
documents, they were to come to conclusions of their own.
called this subscale worrying. We prefer the more specific
The rating version is basically identical to the self-report
term inability to switch off. The second indicator refers to
version. It contains almost all items from the self-report
well-being in leisure time. It was assessed by a one-item
version, except for items asking for information that is
assessable to jobholders only (e.g., an item concerning time measure of exhaustion in the evening of a rest day (“How
control: “To what extent can you decide how long you work exhausted do you feel today?”; see Grebner, 2001). Partic-
on a task?”), and therefore rule out an independent judg- ipants were asked to rest on a day off and to avoid strenuous
ment. Item wordings of the rating version are identical mental and physical work, such as homework or gardening.
except for the formulation concerning the focal person (e.g., Recommended activities were nonstrenuous activities like
“To what extent are your tasks . . .” vs. “To what extent are listening to music, non-work-related reading, watching tele-
the tasks of the jobholder . . .”). vision, watching a movie, and taking a leisurely walk. To
Raters were trained intensively, including (a) study of the check compliance, we interviewed participants in the late
literature and instructions concerning job analysis, stress- evening of the rest day or on the day after with respect to
related job analysis (particularly ISTA), the methodology of activities that were undertaken. These interviews revealed
the observation interview, observation biases, and general good compliance with our instructions. Main activities were
information about occupations; (b) three tutorial exercise sleeping late, watching TV, visiting friends or relatives,
observer ratings at different workplaces with subsequent reading, shopping, taking a walk, and visiting restaurants.
discussion of problems encountered; and (c) a comprehen- Some did unavoidable homework like preparing a meal or
sion and memory test of dimensions of the instrument. All ironing, but only for short periods. Therefore, none of the
in all the training lasted 71⁄2 days (for details, see Grebner, participants had to be excluded from data analyses, and we
2001). are fairly confident that this measure does, indeed, indicate
Unfortunately, it was not possible to have two observers insufficient recovery at the end of 1 day off. It was assessed
rating independently each workplace. Interrater reliabilities by diary in a subgroup of our sample (Time 1 N ⫽ 61, Time
have been reported for previous studies with the ISTA 2 N ⫽ 49; see Grebner, 2001). Therefore, for analyses based
34 GREBNER, SEMMER, AND ELFERING

on this measure, sample sizes are smaller than for the other in the model, which further reduces power in an already
study variables. rather small sample. Nevertheless, we feel that this control
is absolutely necessary. Working conditions in the various
occupations are quite different (see Kälin et al., 2000). In
Data Analysis addition, training time ranges from 2 (salespeople) to 4
(nurses) years, reflecting differences in complexity of the
To test causal effects while controlling for common task requirements involved. Furthermore, gender distribu-
method variance, combining information from different tion is far from equal across occupations. Not controlling for
sources in a structural equation model would be most ap- occupation and gender might, therefore, substantially bias
propriate (cf. Semmer et al., 1996; Zapf et al., 1996). results, and models with and without these controls have
Unfortunately, our sample size is not sufficient for such yielded different results in many cases. We have therefore
analyses. With regard to analyses, we therefore used biva- included these control variables in all our analyses with this
riate correlation and multiple regression analyses instead. data set, both with regard to the project at large (Kälin et al.,
With regard to the combination of self-report versus ob- 2000) and to in-depth studies with subsamples, such as the
server data, we decided to use the mean of the two sources present one.
in the regression analyses. The reasoning behind this deci- For each dependent variable, two regression analyses
sion is that both will contain method variance but that the were performed. One concerns the longitudinal prediction
method variance contained in self-report is likely to be of well-being by working conditions at Time 1. The second
uncorrelated with that contained in ratings. Adding the two analysis concerns synchronous effects at Time 2. Well-
sources up should, therefore, attenuate bias stemming from being at Time 1 is controlled for in these models, thus going
method variance, as the two sources of error should, at least beyond mere cross-sectional analyses (see Zapf et al.,
to some degree, cancel each other out. This combined 1996).
measure should, therefore, lead to a more accurate estimate
of the true relationship between work characteristics and
well-being (cf. Grebner, 2001; Semmer et al., 2004). Results
In all regression models, gender and occupation (dummy
coding) were entered in a first step, together with the de-
pendent variable at Time 1. Stressors and control were Table 1 presents descriptive statistics and reliabilities
entered next. Controlling for the dependent variable at Time of all study variables. Self-reports of working condi-
1 also controls for possible confounding effects of stable tions (S) showed higher means than observer ratings (R)
personality characteristics, such as negative affectivity, be- for both stressors and control. Internal consistencies
cause these should, by definition, already be contained in
the Time 1 value of well-being (Spector et al., 2000). We
(coefficient ␣) of the stressor scales (time pressure,
realize that controlling for gender and occupation is not uncertainty, organizational problems, concentration,
without costs, because it increases the number of variables and work interruptions) vary for self-reports between

Table 1
Means, Standard Deviations, and Internal Consistency (Cronbach’s Alpha) of All Study Variables
Time 1 Time 2
Variable No. of items M SD N ␣ M SD N ␣
Working conditions
Job control
Sa 5 3.53 0.73 86 .76 3.56 0.75 65 .80
Ra 12 2.52 0.68 74 .90 2.63 0.67 54 .91
SR 2 scales 3.04 0.54 69 —c 3.13 0.62 52 —c
Job stressors
Sa 5 scales 3.19 0.61 85 —c 3.20 0.65 65 —c
Ra 5 scales 2.75 0.54 74 —c 2.88 0.51 54 —c
SR 2 scales 3.01 0.53 68 —c 3.08 0.53 52 —c
Well-being
Psychosomatic complaintsa 20 2.11 0.61 88 .90 2.11 0.66 68 .92
Irritated reactionsb 4 2.72 1.32 85 .87 2.68 1.29 65 .90
Job satisfactionb 4 4.70 1.23 82 .83 4.52 1.31 62 .83
Resigned attitude toward one’s jobb 4 2.48 1.00 86 .72 2.37 1.02 66 .72
Inability to switch off after workb 4 3.01 1.28 84 .78 2.93 1.23 65 .84
Exhaustion in the evening of a rest daya 1 2.55 1.06 58 —d 2.51 1.03 41 —d
Note. S ⫽ self-report; R ⫽ observer rating; SR ⫽ mean of self-report and observer rating.
a
Scored from 1 to 5. b Scored from 1 to 7. c Mean of several scales, no internal consistency calculated. d
Single item,
no internal consistency estimate possible.
 WORKING CONDITIONS AND WELL-BEING 35

.73 and .85 and for ratings between .58 and .83. Internal Spillover. Job Control ⫻ SR (Time 1) strongly
consistencies of the control scales (method and time predicts both inability to switch off and exhaustion
control) range between .61 and .69 for self-reports and on rest day in the longitudinal analysis. Stressors
between .82 and .91 for ratings. Means demonstrate a predict the inability to switch off in the synchronous
remarkable stability. Of all variables, only job satisfac- model (Table 6).
tion shows a statistically reliable difference between
Time 1 and Time 2, with Time 2 values being lower. Drift Hypothesis
For more detailed information, see Grebner (2001).
Intercorrelations between variables are shown in Ta- To examine the drift hypothesis, we estimated
ble 2 (cross-sectionally) and Table 3 (longitudinally). regression models predicting working conditions
Both working conditions and well-being demonstrate (i.e., Job Control ⫻ SR and Job Stressors ⫻ SR,
fairly high stabilities, with job satisfaction and resigned respectively) from well-being. A regression analysis
attitude toward one’s job being the least stable. In line was run for each of the well-being indicators, forcing
with theoretical expectations, job stressors showed a them into the model after control variables (sex and
pattern of associations with inability to switch off, ten- occupation) and the dependent variable at Time 1.
dency to react in an irritated way, psychosomatic com- For control, there was one effect that was marginally
plaints, resigned attitude toward one’s job, and job significant. It concerned job satisfaction, which pre-
satisfaction. For job control, the pattern was similar in dicted job control at Time 2 (B ⫽ .11, SE B ⫽ .06,
kind but in the opposite direction. This pattern is fairly ␤ ⫽ .21, p ⫽ .08). None of the other variables
consistent across cross-sectional and longitudinal anal- yielded a ␤ coefficient that reached .20; the largest
yses, with values being weakest for the association coefficient was –.15 (psychosomatic complaints pre-
between stressors Time 1 and well-being Time 2. It can dicting job control, p ⫽ .22), whereas all others were
be seen both in the self-report (S) and the rating (R) between .02 and .08, in absolute terms. For stressors,
measures, as well as in their combination (SR). The none of the variables came even close to what might be
combined measure tends to show equal or stronger considered a substantial effect even by the most liberal
associations than either S or R alone. Convergence of standards (coefficients ranged between –.04 and .07).
self-reported and observed working conditions was To guard against Type II error with regard to
modest for control but rather high for stressors. It should reversed causation, we conducted additional analy-
be noted that analyses based on the individual scales ses. First, we combined the indicators for each do-
contained in the stressor index (not shown) yield a very main (general, job related, and spillover) after stan-
similar pattern (see Grebner, 2001). (Data can be ob- dardizing them. Second, we combined all of the
tained on request.) indicators into one single index. In case any effect of
well-being on working conditions was due not to a
Working Conditions Predicting Well-Being single aspect of well-being but to a common feature
of either domain-specific or general well-being, these
General well-being. Table 4 shows the analyses effects should have a better chance to be detected in
for the two indicators of general well-being, regressed this type of analysis. Only one of the coefficients
on stressors and control, using the combined self-report obtained came close to being significant (prediction
and rating measures. There is no longitudinal prediction of control by job-related well-being; ␤ ⫽ –.22, p ⫽
for psychosomatic complaints, which is also the well- .09), and the magnitude of the other coefficients in
being indicator with the greatest stability. For irritated absolute terms was very small, ranging from .00 to
reactions, however, there is an effect of Job Control ⫻ –.12. Thus, if there is reversed causation, our data
SR (Time 1), which is marginally significant. indicate that this effect is not very strong, and in any
For synchronous effects, a different picture case smaller than the effect of working conditions on
emerges. Stressors do predict psychosomatic com- well-being. The only exception is an effect of job sat-
plaints, and control predicts irritated reactions. isfaction that is classified as job-related well-being.
Job-related well-being. Table 5 shows the results
for the job-related attitudes. Job Stressors ⫻ SR Dropouts
yielded a marginally significant effect on resigned
attitude toward one’s job, both in the longitudinal and Finally, we compared means of working condi-
the synchronous analysis, and a rather strong and tions and indicators of well-being of those who par-
statistically significant effect on job satisfaction in ticipated only in Time 1 with those who participated
the synchronous model. in both waves. The two groups did not differ signif-
 36

Table 2
Cross-Sectional Intercorrelations of Working Conditions and Well-Being (Time 1 and Time 2)
Variable 1 2 3 4 5 6 7 8 9 10 11 12
Job control
1. JC–S — .45** .87*** ⫺.20 ⫺.39** ⫺.38** ⫺.20 ⫺.32** .33** ⫺.14 ⫺.22† ⫺.07
2. JC–R .32** — .83*** ⫺.33* ⫺.43** ⫺.44** ⫺.18 ⫺.25† .16 ⫺.28* ⫺.27* ⫺.22
3. JC–SR .83*** .79*** — ⫺.39** ⫺.50*** ⫺.48*** ⫺.23 ⫺.33* .34* ⫺.24† ⫺.33* ⫺.16
Job stressors
4. JST–S ⫺.28* ⫺.15 ⫺.31* — .70*** .94*** .44*** .15 ⫺.40** .00 .50*** .04
5. JST–R ⫺.37** ⫺.06 ⫺.21† .65*** — .90*** .25† .27† ⫺.38** .17 .42** ⫺.08
6. JST–SR ⫺.39** ⫺.07 ⫺.30* .92*** .89*** — .34* .31* ⫺.44** .14 .57*** ⫺.01
Well-being
7. PSY ⫺.20† ⫺.16 ⫺.28* .22* .03 .18 — .54*** ⫺.60*** .31* .54*** .42**
8. IRREA ⫺.22* ⫺.28* ⫺.34** .26* .26* .34** .51*** — ⫺.39** .35** .61*** .39*
9. SAT .36** .29* .43*** ⫺.17 ⫺.08 ⫺.17 ⫺.48*** ⫺.38** — ⫺.44*** ⫺.42** ⫺.23
10. RES ⫺.38*** ⫺.33** ⫺.47*** .18 .11 .20 .31** .34** ⫺.61*** — .40** .34*
11. INSO ⫺.28** ⫺.18 ⫺.34** .56*** .26* .50*** .56*** .60*** ⫺.36** .19† — .27†
12. EXHR ⫺.28* ⫺.08 ⫺.23† .27* .02 .15 .43** .42** ⫺.27* .16 .47*** —
GREBNER, SEMMER, AND ELFERING

Note. Intercorrelations for Time 1 are presented below the diagonal, for Time 2 above the diagonal. Method convergence in italics. JC ⫽ job control; S ⫽ self-report; R ⫽ observer
rating; SR ⫽ mean of self-report and observer rating; JST ⫽ job stressors; PSY ⫽ psychosomatic complaints; IRREA ⫽ irritated reactions; SAT ⫽ job satisfaction; RES ⫽ resigned
attitude toward one’s job; INSO ⫽ inability to switch off after work; EXHR ⫽ exhaustion in the evening of a rest day. Sample sizes: Time 1: N(S) ⫽ 85, N(R) ⫽ 74, N(SR) ⫽
69; Time 2: N(S) ⫽ 65, N(R) ⫽ 54, N(SR) ⫽ 51. Sample size for Exhaustion Time 1: N ⫽ 58; Exhaustion Time 2: N ⫽ 41.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
Table 3
Longitudinal Intercorrelations of Working Conditions and Well-Being
Time 2
Time 1 1 2 3 4 5 6 7 8 9 10 11 12
Job control
1. JC–S .53*** .35* .65*** ⫺.25† ⫺.29* ⫺.34* ⫺.22† ⫺.27* .25* ⫺.22† ⫺.45*** ⫺.29†
2. JC–R .26† .60*** .45** ⫺.34* ⫺.20 ⫺.34* ⫺.33* ⫺.30* .08 ⫺.29* ⫺.31* ⫺.21
3. JC–SR .58*** .57*** .70*** ⫺.40** ⫺.30* ⫺.41** ⫺.37** ⫺.38** .27* ⫺.35* ⫺.50*** ⫺.39*
Job stressors
4. JST–S ⫺.09 ⫺.24 ⫺.26† .71*** .56*** .70*** .25* .08 ⫺.19 .06 .38** ⫺.03
5. JST–R ⫺.30* ⫺.35* ⫺.45** .50*** .71*** .65*** .02 .14 ⫺.13 ⫺.03 .21 ⫺.20
6. JST–SR ⫺.22 ⫺.34* ⫺.40** .69*** .70*** .75*** .16 .12 ⫺.21 .06 .35* ⫺.14
Well-being
7. PSY ⫺.24† ⫺.26† ⫺.30* .32* .21 .27† .83*** .51*** ⫺.45*** .21† .47*** .56***
8. IRREA ⫺.09 ⫺.24† ⫺.21 .27* .10 .18 .47*** .67*** ⫺.28* .13 .45*** .12
9. SAT .31* .34* .42** ⫺.11 ⫺.17 ⫺.22 ⫺.37** ⫺.50*** .44** ⫺.21 ⫺.43** ⫺.29†
10. RES ⫺.16 ⫺.33* ⫺.29* .00 .00 .06 .25* .37** ⫺.27* .41** .22† .37*
11. INSO ⫺.12 ⫺.20 ⫺.25† .54*** .37** .44** .45*** .36** ⫺.28* .16 .56*** .20
12. EXHR ⫺.02 .08 .03 .14 ⫺.13 ⫺.02 .35* .37* ⫺.25† .11 .26† .51**
WORKING CONDITIONS AND WELL-BEING

Note. JC ⫽ job control; S ⫽ self-report; R ⫽ observer rating; SR ⫽ mean of self-report and observer rating; JST ⫽ job stressors; PSY ⫽ psychosomatic complaints; IRREA ⫽
irritated reactions; SAT ⫽ job satisfaction; RES ⫽ resigned attitude toward one’s job; INSO ⫽ inability to switch off after work; EXHR ⫽ exhaustion in the evening of a rest day.
Longitudinal sample size: N(S) ⫽ 64, N(R) ⫽ 53, N(SR) ⫽ 52. Sample size EXHR: N ⫽ 48.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
37
38 GREBNER, SEMMER, AND ELFERING

Table 4
Prediction of Well-Being: General
PSY Time 2 IRREA Time 2
Step and variable B SE B ␤ ⌬R2 B SE B ␤ ⌬R2
Longitudinal analyses

1 Dependent Time 1 0.94 .09 .80*** .73*** 0.61 .12 .62***45***

2 Job Stressors ⫻ SR 0.00 .15 .00 ⫺0.25 .50 ⫺.09
Job Control ⫻ SR ⫺0.03 .12 ⫺.03 .00 ⫺0.63 .34 ⫺.26† .04
Synchronous analyses

1 Dependent Time 1 0.90 .09 .83*** .71*** 0.68 .10 .73*** .60***
2 Job Stressors ⫻ SR 0.34 .15 .28* 0.54 .33 .22
Job Control ⫻ SR 0.06 .10 .06 .03† ⫺0.57 .25 ⫺.27* .08*
Note. PSY ⫽ psychosomatic complaints; IRREA ⫽ irritated reactions; SR ⫽ mean of
self-report and observer rating. Sex and occupation were included in Step 1 but are not shown.
Longitudinal analyses: PSY N ⫽ 52; IRREA N ⫽ 52; PSY adjusted R2 ⫽ .69 for Step 1;
IRREA adjusted R2 ⫽ .38 for Step 1. Synchronous analyses: PSY N ⫽ 51. IRREA N ⫽ 50.
PSY adjusted R2 ⫽ .67 for Step 1. IRREA adjusted R2 ⫽ .54 for Step 1.
† p ⬍ .10. * p ⬍ .05. *** p ⬍ .001.

icantly in working conditions. Only two effects were conditions that is less subject to common method
found for well-being; they concern lower job satis- variance than the self-report measures typically used.
faction (p ⬍ .10) and more exhaustion (p ⬍ .05) for Method convergence was quite high (see Spector &
the participants who dropped out at Time 2. Jex, 1998). Bivariate associations of working condi-
tions and well-being were somewhat higher than typ-
Discussion ically reported in the literature (cf. Semmer et al.,
1996; Zapf et al., 1996).
This study shows associations between working Longitudinal effects were found only for leisure-
conditions and well-being with a measure of working time well-being. Controlling for the dependent vari-

Table 5
Prediction of Well-Being: Job-Related
SAT Time 2 RES Time 2
Step and variable B SE B ␤ ⌬R2 B SE B ␤ ⌬R2
Longitudinal analyses

1 Dependent Time 1 0.59 .17 .50** .25* 0.35 .17 .31* .28*
2 Job Stressors ⫻ SR ⫺0.51 .53 ⫺.19 0.76 .38 .35†
Job Control ⫻ SR 0.43 .43 .18 .04 ⫺0.32 .30 ⫺.17 .09†
Synchronous analyses

1 Dependent Time 1 0.56 .18 .48** .23† 0.31 .14 .32* .43***
2 Job Stressors ⫻ SR ⫺1.46 .45 ⫺.56** 0.55 .30 .31†
Job Control ⫻ SR 0.55 .40 .24 .21** ⫺0.11 .22 ⫺.07 .05
Note. SAT ⫽ job satisfaction; RES ⫽ resigned attitude toward one’s job; SR ⫽ mean of
self-report and observer rating. Sex and occupation were included in Step 1 but are not shown.
Longitudinal analyses: SAT N ⫽ 51; RES N ⫽ 52; SAT adjusted R2 ⫽ .15 for Step 1; RES
adjusted R2 ⫽ .19 for Step 1. Synchronous analyses: SAT N ⫽ 48; RES N ⫽ 50; SAT adjusted
R2 ⫽ .11 for Step 1; RES adjusted R2 ⫽ .35 for Step 1.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
 WORKING CONDITIONS AND WELL-BEING 39

Table 6
Prediction of Well-Being: Spillover
INSO Time 2 EXHR Time 2
Step and variable B SE B ␤ ⌬R2 B SE B ␤ ⌬R2
Longitudinal analyses

1 Dependent Time 1 0.59 .13 .62*** .36** ⫺0.52 .19 .44* .33†
2 Job Stressors ⫻ SR 0.70 .46 .28 ⫺0.13 .38 ⫺.07
Job Control ⫻ SR ⫺1.04 .34 ⫺.45** .16** ⫺1.33 .46 ⫺.66** .17*
Synchronous analyses

1 Dependent Time 1 0.61 .12 .62*** .49*** 0.52 .17 .48** .41*
2 Job Stressors ⫻ SR 1.33 .36 .55** 0.13 .39 .07
Job Control ⫻ SR ⫺0.41 .25 ⫺.20 .16*** ⫺0.60 .36 ⫺.35 .06
Note. INSO ⫽ inability to switch off after work; EXHR ⫽ exhaustion in the evening of a
rest day; SR ⫽ mean of self-report and observer rating. Sex and occupation were included in
Step 1 but are not shown. Longitudinal analyses: INSO N ⫽ 51; EXHR N ⫽ 34; INSO
adjusted R2 ⫽ .27 for Step 1; EXHR adjusted R2 ⫽ .27 for Step 1. Synchronous analyses:
INSO N ⫽ 49; EXHR N ⫽ 32; INSO adjusted R2 ⫽ .42 for Step 1; EXHR adjusted R2 ⫽ .27
for Step 1.
† p ⬍ .10. * p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

able at Time 1, job control at Time 1 predicted in terms of the absolute size of the coefficients. To
inability to switch off at Time 2, as well as exhaus- minimize the risk of a Type II error, we ran several
tion in the evening of a rest day. analyses, using well-being variables alone, and as
Of course, the lack of significant findings with indexes combining domain-specific indicators as well
regard to general and job-related well-being could be as an overall index of well-being. The results support
due to lack of power because of our small sample the conclusion that the prediction of working condi-
size. Concentrating on statistical significance only tions by well-being is, if at all present, smaller than
might therefore be associated with a high probability the prediction of well-being by working conditions.
of a Type II error. Schmidt (1996) therefore recom- The only exception refers to job satisfaction, which
mended not concluding prematurely that there is no predicts control. It is tempting to speculate that a vari-
effect but rather to put more emphasis on the actual able that strongly reflects an attitudinal component
coefficients obtained, which, after all, are the best might be more visible to employers and supervisors
estimates of the actual relationship. The final conclu- than symptoms such as difficulties relaxing after work
sion then has to be reserved for meta-analyses that or stomachaches, which might be more hidden, at least
can be based on coefficients thus obtained. In our as long as they do not reach very high levels. But this is
results, only one standardized regression coefficient not more than a speculation, and its main purpose is to
has an absolute value that is above .20 for psycho- encourage researchers to look for such specific effects
somatic complaints, whereas this is the case for three in their data. Note that even the presence of stronger
out of four coefficients for spillover. Results for effects in terms of reverse causation would not invali-
work-related well-being are in between, with three date the prediction of well-being by working conditions,
out of four coefficients being close to .20, the fourth as one may plausibly assume that both might reinforce
one higher and marginally significant. Thus, with each other reciprocally (Edwards, 1998).
regard to the 1-year lag, we are quite confident about Synchronous effects were more widespread than
work characteristics predicting spillover, and not longitudinal ones, yielding associations between
strongly predicting general well-being. With regard stressors and psychosomatic complaints, job satisfac-
to job satisfaction and job-related resignation, we tion, resigned attitude toward one’s job, and inability
cannot be confident that there is an effect but we to switch off, and between control and irritated reac-
would not claim that there is no effect either. tions. In addition, there are some coefficients that are
We did not find strong indications for reverse not significant but .20 or higher in absolute value.
causation, either in terms of statistical significance or This applies for stressors in the prediction of irritated
40 GREBNER, SEMMER, AND ELFERING

reaction and for control in the prediction of job 1991; McEwen, 1998; Meijman & Mulder, 1998).
satisfaction and both indicators of spillover. Kasl (1992) regarded spillover into other life do-
Of course, synchronous models are weaker than mains as a key condition for adverse long-term ef-
the longitudinal ones in terms of causality. Neverthe- fects of working conditions. Our data suggest that
less, they go well beyond simple cross-sectional anal- such variables should routinely be included in re-
yses, as Time 1 values of the dependent variable are search on stress at work.
controlled for. Therefore, current working conditions Stressors seem to exert their effects more quickly
predict residual well-being, which is the deviation of and retain some reversibility. Note, however, that this
the value that would be expected on the basis of its refers to a very young sample. It may well be that this
initial value. Thus, selection effects based on a stable reversibility becomes less and less as people are
well-being cannot account for these associations. For exposed to stressful conditions over extended periods
that to be the case, one would have to argue that of time.
relative changes in well-being— changes that deviate Apart from the small sample size, one reason for
from what can be expected from someone with a not finding more longitudinal effects may lie in the
given initial value—would cause a deterioration in time interval of 1 year. It seems plausible to assume
working conditions. Employers would have to notice that for many indicators of well-being, especially the
deterioration in well-being and react to it by assign- more stable ones, enduring changes are not easily
ing more stressful working conditions to these em- achieved but require an exposure over a longer time
ployees. Given that their basic well-being cannot be interval. To detect those, one would need repeated
responsible for such a decision (as well-being Time 1 assessments over rather long intervals, so that people
is controlled for), this does not seem very likely. For can be identified who are subject to unfavorable
employees whose basic well-being was not consid- conditions for an extended period of time. By con-
ered a reason for transferring them to less favorable trast, more transient changes that are due to currently,
conditions, it seems more likely that employers or recently, encountered working conditions can be
would, if anything, try to protect them from excessive detected within shorter intervals. If conditions get
stress, so that they might regain their initial well- better, these changes may be reversed, if not, they
being (corresponding to the refuge model; Garst et may gradually become chronic (see Sonnentag &
al., 2000). It therefore seems more likely that the Frese, 2003, for a similar conclusion).
synchronous results reflect short-term effects of
working conditions on well-being (see Zapf et al., Strengths and Limitations
1996).
If this interpretation is correct, it implies that for This study has several limitations. Participants are
our indicators of general and of job-related well- rather young, and generalization of results to older
being, fluctuations around the baseline value occur people remains open. Occupations are not equally
rather quickly in reaction to working conditions but distributed in our sample. Also, although most mea-
are reversible—at least in such a young population. sures indicate satisfying levels of internal consis-
By contrast, the spillover indicators may reflect long- tency, for some scales it was below .70. The greatest
term effects of undesirable working conditions. limitation, however, concerns sample size. Given this
Note also that of the four effects that are at least sample, we may be quite confident about effects that
marginally significant in the longitudinal analyses, are statistically significant. For nonsignificant coeffi-
three concern control, including both significant cients, however, there is the risk of a Type II error.
ones. Conversely, four out of five predictions that are We therefore emphasize that some coefficients,
at least marginally significant in the synchronous mostly with regard to work-related well-being, are
analyses involve stressors. not significant yet substantial in absolute value, so we
Taken together, these results suggest that delayed certainly do not claim that there is no effect in this
effects of job control may build up rather early in case.
one’s working career, underscoring the importance of Nevertheless, there are several assets of this study.
control at work (Marmot et al., 1997; Semmer, 1998; First, there is the use of rating data as well as self-
Terry & Jimmieson, 1999). Furthermore, the effect of reports for assessment of working conditions, which
job control seems to be especially strong on variables at least attenuates problems arising from common
that reflect insufficient recovery from work. Such method variance. Another asset is the longitudinal
insufficient “unwinding” is regarded as especially design, which allows us to partial out initial values of
important by many researchers (e.g., Frankenhaeuser, the dependent variable, and which also allows for
 WORKING CONDITIONS AND WELL-BEING 41

testing reverse causation. Finally, our focus on dif- Arnold, J., & Reicherts, M. (2000). The transition into
ferent types of well-being allows us to analyze dif- work: An editorial commentary. Swiss Journal of Psy-
chology, 59, 221–226.
ferential associations. These analyses show that spill-
Bailey, J. M., & Bhagat, R. S. (1987). Meaning and mea-
over from work into private life seems to be a surement of stressors in the work environment: An eval-
promising area for further study. uation. In S. V. Kasl & C. L. Cooper (Eds.), Stress and
Note that there is a trade-off between some of our health: Issues in research methodology (pp. 207–229).
assets and our major limitation. Many authors have Chichester, UK: Wiley.
Baillod, J., & Semmer, N. (1994). Fluktuation und Berufs-
called for more longitudinal studies (Zapf et al., verläufe bei Computerfachleuten [Turnover and career
1996) and for measures other than self-report (e.g., paths among computer specialists]. Zeitschrift für Ar-
Ganster & Schaubroeck, 1991). The problem is that beits- und Organisationspsychologie, 38, 152–163.
following these recommendations makes it more dif- Barling, J., & Kelloway, K. E. (1999). Introduction. In J.
ficult to get large sample sizes. Longitudinal studies Barling (Ed.), Young workers: Varieties of experience
(pp. 3–15). Washington, DC: American Psychological
inevitably involve some dropout, and sending ob- Association.
servers to people’s workplace limits the sample size Bruggemann, A. (1974). Zur Unterscheidung verschiedener
by increasing the research effort (including costs, Formen von ‘Arbeitszufriedenheit’ [Different forms of
which makes it more difficult to get funded). Further- job satisfaction]. Arbeit und Leistung, 28, 281–284.
more, it requires the consent of yet another group of Büssing, A. (1992). A dynamic view of job satisfaction in
psychiatric nurses in Germany. Work & Stress, 6, 239 –259.
people, that is, management representatives of the Büssing, A. (1999). Can control at work and social support
companies involved. We nevertheless opted for this moderate psychological consequences of job insecurity?
study design, because we are convinced that, in the Results from a quasi-experimental study in the steel
long run, this type of data will advance the field. industry. European Journal of Work and Organizational
Psychology, 8, 219 –242.
Schmidt (1996) strongly urged researchers not to
Edwards, J. R. (1998). Cybernetic theory of stress, coping,
decide against conducting such studies. He convinc- and well-being. In C. L. Cooper (Ed.), Theories of orga-
ingly argued that a large number of small-scale stud- nizational stress (pp. 122–152). Oxford, UK: Oxford
ies would, in the end, allow firm conclusions based University Press.
on meta-analysis. This implies that the confidence we Fahrenberg, J. (1975). Die Freiburger Beschwerdeliste FBL
[The Freiburg Bodily Complaints Inventory]. Zeitschrift
can have in our particular results is limited. It also für Klinische Psychologie, 4, 49 –100.
implies, however, that without such studies, progress Frankenhaeuser, M. (1991). The psychophysiology of sex
will be slower altogether. differences as related to occupational status. In M. Fran-
kenhaeuser, U. Lundberg, & M. Chesney (Eds.), Women,
work, and health: Stress and opportunities (pp. 39 – 61).
Conclusions New York: Plenum Press.
Frese, M. (1982). Occupational socialization and psycho-
Our data indicate rather short-term, and thus re- logical development: An underemphasized research per-
versible, effects of stressors on different kinds of spective in industrial psychology. Journal of Occupa-
tional Psychology, 55, 209 –224.
well-being. This is in line with many other findings Frese, M. (1985). Stress at work and psychosomatic com-
(Sonnentag & Frese, 2003) and therefore adds an- plaints: A causal interpretation. Journal of Applied Psy-
other piece of evidence to this literature. Our data chology, 70, 314 –328.
also indicate that more lasting effects over time may Frese, M., & Zapf, D. (1988). Methodological issues in the
develop even within an interval of 1 year in a sample study of work stress: Objective vs. subjective measurement
and the question of longitudinal studies. In C. L. Cooper &
of rather young people. These results, of course, R. Payne (Eds.), Causes, coping and consequences of stress
require replication. They do suggest, however, that at work (pp. 375– 411). Chichester, UK: Wiley.
variables that concern spillover from work into pri- Frese, M., & Zapf, D. (1994). Action as the core of work
vate life deserve more attention in such studies. Fur- psychology: A German approach. In M. D. Dunnette,
L. M. Hough, & H. C. Triandis (Eds.), Handbook of indus-
thermore, if such results are confirmed in future stud-
trial and organizational psychology (Vol. 4, pp. 271–340).
ies, they imply that prevention efforts in terms of job Palo Alto, CA: Consulting Psychologists Press.
design may be most promising if they start very early. Frone, M. R. (2000). Interpersonal conflict at work and
psychological outcomes: Testing a model among young
workers. Journal of Occupational Health Psychology, 5,
References 246 –255.
Ganster, D. C., Fox, M. L., & Dwyer, D. J. (2001). Explain-
Arnold, J. (1997). The psychology of careers in organiza- ing employees’ health care costs: A prospective exami-
tions. In C. L. Cooper & I. T. Robertson (Eds.), Interna- nation of stressful job demands, personal control, and
tional review of industrial and organizational psychology physiological reaction. Journal of Applied Psychology,
(Vol. 12, pp. 1–37). Chichester, UK: Wiley. 86, 954 –964.
42 GREBNER, SEMMER, AND ELFERING

Ganster, D. C., & Schaubroeck, J. (1991). Work stress and of workload. In P. J. D. Drenth, H. Thierry, & C. J. de
employee health. Journal of Management, 17, 235–271. Wolff (Eds.), Handbook of work and organizational psy-
Garst, H., Frese, M., & Molenaar, P. C. M. (2000). The chology: Vol. 2. Work psychology (2nd ed., pp. 5–33).
temporal factor of change in stressor–strain relationships: Hove, UK: Psychology Press.
A growth curve model on a longitudinal study in East Mohr, G. (1991). Fünf Subkonstrukte psychischer Befin-
Germany. Journal of Applied Psychology, 85, 417– 438. densbeeinträchtigungen bei Industriearbeitern: Auswahl
Grebner, S. (2001). Stress at work, well-being, blood pres- und Entwicklung [Five subconstructs of psychological
sure, and cortisol: Two field studies. Unpublished doc- strain in industrial personnel: Selection and development].
toral dissertation, University of Bern, Bern, Switzerland. In S. Greif, E. Bamberg, & N. Semmer (Eds.), Psychischer
Greiner, B., Krause, N., Ragland, D. R., & Fisher, J. M. Stress am Arbeitsplatz [Psychological stress in the work-
(1998). Objective stress factors, accidents, and absentee- place] (pp. 91–119). Göttingen, Germany: Hogrefe.
ism in transit operators: A theoretical framework and Mohr, G. B. (2000). The changing significance of different
empirical evidence. Journal of Occupational Health Psy- stressors after the announcement of bankruptcy: A lon-
chology, 3, 130 –146. gitudinal investigation with special emphasis on job inse-
Hemingway, H., & Marmot, M. (1999). Psychosocial fac- curity. Journal of Organizational Behavior, 21, 337–359.
tors in the aetiology and prognosis of coronary heart Morsh, J. E. (1964). Job analysis in the United States Air
disease: Systematic review of prospective cohort studies. Force. Personnel Psychology, 17, 7–17.
British Medical Journal, 318, 1460 –1467. Oegerli, K. (1984). Arbeitszufriedenheit: Versuch einer
Kahn, R. L., & Byosiere, P. (1992). Stress in organizations. In qualitativen Bestimmung [Job satisfaction: A qualitative
M. D. Dunnette & L. M. Hough (Eds.), Handbook of analysis]. Unpublished doctoral dissertation, University
industrial and organizational psychology (2nd ed., pp. 571– of Bern, Bern, Switzerland.
650). Palo Alto, CA: Consulting Psychologists Press. Räikkönen, K., Matthews, K. A., Flory, J. D., Owens, J. F.,
Kälin, W., Semmer, N. K., Elfering, A., Tschan, F., Dauwalder, J. & Gump, B. B. (1999). Effects of optimism, pessimism,
-P., Heunert, S., & Crettaz de Roten, F. (2000). Work charac- and trait anxiety on ambulatory blood pressure and mood
teristics and well-being of Swiss apprentices entering the during everyday life. Journal of Personality and Social
labor market. Swiss Journal of Psychology, 59, 272–290. Psychology, 76, 104 –113.
Kasl, S. V. (1986). Stress and disease in the workplace: A meth- Rozanski, A., Blumenthal, J. A., & Kaplan, J. (1999). Im-
odological commentary on the accumulated evidence. In M. F. pact of psychological factors on the pathogenesis of
Cataldo & T. J. Coates (Eds.), Health and industry: A behav- cardiovascular disease and implications for therapy. Cir-
ioral medicine perspective (pp. 52–85). New York: Wiley. culation, 99, 2192–2217.
Kasl, S. V. (1992). Surveillance of psychological disorders in Saks, A. M., & Ashforth, B. E. (2000). The role of dispo-
the workplace. In G. P. Keita & S. L. Sauter (Eds.), Work sitions, entry stressors, and behavioral plasticity theory in
and well-being: An agenda for the 1990s (pp. 73–95). predicting newcomers’ adjustment to work. Journal of
Washington, DC: American Psychological Association. Organizational Behavior, 21, 43– 62.
Kessler, R. C. (1987). The interplay of research design Schmidt, F. L. (1996). Statistical significance testing and cu-
strategies and data analysis procedures in evaluating the mulative knowledge in psychology: Implications for train-
effects of stress on health. In S. V. Kasl & C. L. Cooper ing of researchers. Psychological Methods, 1, 115–129.
(Eds.), Research methods in stress and health psychology Semmer, N. (1998). Control at work: Issues of specificity,
(pp. 113–140). Chichester, UK: Wiley. generality, and legitimacy. In W. J. Perrig & A. Grob
Kiecolt-Glaser, J. K., McGuire, L. Tobles, T. F., & Glaser, (Eds.), Control of human behavior, mental processes, and
R. (2002). Emotions, morbidity, and mortality: New per- consciousness: Essays in honor of the 60th birthday of
spectives from psychoneuroimmunology. Annual Review August Flammer (pp. 714 –741). Mahwah, NJ: Erlbaum.
of Psychology, 53, 83–107. Semmer, N. (2003). Individual differences, work stress, and
Loughlin, C., & Barling, J. (2001). Young workers’ work health. In M. J. Schabracq, J. A. M. Winnubst, & C. L.
values, attitudes, and behaviours. Journal of Occupa- Cooper (Eds.), Handbook of work and health psychology
tional and Organizational Psychology, 74, 543–558. (2nd ed., pp. 83–120). Chichester, UK: Wiley.
Manning, M. R., Jackson, C. N., & Fusilier, M. R. (1996). Semmer, N., Baillod, J., Stadler, R., & Gail, K. (1996).
Occupational stress and health care use. Journal of Oc- Fluktuation bei Computerfachleuten: Eine follow-up
cupational Health Psychology, 1, 100 –109. Studie [Turnover among computer specialists: A fol-
Marmot, M., Bosma, H., Hemingway, H., Brunner, E. J., & low-up study]. Zeitschrift für Arbeits- und Organisations-
Stansfeld, S. A. (1997). Contribution of job control and psychologie, 40, 190 –199.
other risk factors to social variations in coronary heart Semmer, N. K., Grebner, S., & Elfering, A. (2004). Beyond
disease incidence. Lancet, 30, 235–239. self-report: Using observational, physiological, and
Marmot, M., Siegrist, J., Theorell, T., & Feeney, A. (1999). event-based measures in research on occupational stress.
Health and the psychosocial environment at work. In M. In P. L. Perrewé & D. C. Ganster (Eds.), Research in
Marmot & R. G. Wilkinson (Eds.), Social determinants of occupational stress and well-being: Vol. 3. Emotional
health (pp. 105–131). Oxford, UK: Oxford University Press. and physiological processes and positive intervention
McEwen, B. S. (1998). Protective and damaging effects of strategies (pp. 205–263). Amsterdam: Elsevier.
stress mediators. New England Journal of Medicine, 338, Semmer, N., Zapf, D., & Dunckel, H. (1995). Assessing
171–179. stress at work: A framework and an instrument. In O.
McGrath, J. E., & Beehr, T. A. (1990). Time and the stress Svane & C. Johansen (Eds.), Work and health: Scientific
process: Some temporal issues in the conceptualisation basis of progress in the working environment (pp. 105–
and measurement of stress. Stress Medicine, 6, 93–104. 113). Luxembourg: Office for Official Publications of the
Meijman, T. F., & Mulder, G. (1998). Psychological aspects European Communities.
 WORKING CONDITIONS AND WELL-BEING 43

Semmer, N., Zapf, D., & Greif, S. (1996). “Shared job ogy: Vol. 8. Comprehensive clinical psychology (pp. 39 –
strain”: A new approach for assessing the validity of job 78). Amsterdam: Elsevier.
stress measurements. Journal of Occupational and Or- Terry, D. J., & Jimmieson, N. L. (1999). Work control and
ganizational Psychology, 69, 293–310. employee well-being: A decade review. In C. L. Cooper
Sonnentag, S., & Frese, M. (2003). Stress in organizations. & I. T. Robertson (Eds.), International review of indus-
In W. C. Borman, D. R. Ilgen, & J. R. Klimoski (Eds.), trial and organizational psychology (Vol. 14, pp. 95–
Comprehensive handbook of psychology: Industrial and 148). Chichester, UK: Wiley.
organizational psychology (Vol. 12, pp. 453– 491). New Theorell, T., & Karasek, R. A. (1996). Current issues relating to
York: Wiley. psychosocial job strain and cardiovascular disease research.
Spector, P. E. (1992). A consideration of the validity and Journal of Occupational Health Psychology, 1, 9–26.
meaning of self-report measures of job conditions. In Warr, P. (1999). Well-being and the workplace. In D. Kah-
C. L. Cooper & I. T. Robertson (Eds.), International neman, E. Diener, & N. Schwarz (Eds.), Well-being: The
review of industrial and organizational psychology (Vol. foundations of hedonic psychology (pp. 393– 412). New
7, pp. 123–151). Chichester, UK: Wiley. York: Russell Sage Foundation.
Spector, P. E., & Brannick, M. T. (1995). The nature and Williams, L. J., & Brown, B. K. (1994). Method variance in
effects of method variance in organizational research. In organizational behavior and human resources research:
C. L. Cooper & I. T. Robertson (Eds.), International Effects on correlations, path coefficients, and hypothesis
review of industrial and organizational psychology (Vol. testing. Organizational Behavior and Human Decision
10, pp. 249 –274). Chichester, UK: Wiley. Processes, 57, 185–209.
Zapf, D. (1989). Selbst- und Fremdbeobachtung in der
Spector, P. E., & Jex, S. M. (1998). Development of four
psychologischen Arbeitsanalyse [Self-report and obser-
self-report measures of job stressors and strain: Interper-
vation rating in psychological job analysis]. Göttingen,
sonal Conflict at Work Scale, Organizational Constraints
Germany: Hogrefe.
Scale, Quantitative Workload Inventory, and Physical
Zapf, D., Dormann, C., & Frese, M. (1996). Longitudinal
Symptoms Inventory. Journal of Occupational Health
studies in organizational stress research: A review of the
Psychology, 3, 356 –367. literature with reference to methodological issues. Jour-
Spector, P. E., Zapf, D., Chen, P. Y., & Frese, M. (2000). nal of Occupational Health Psychology, 1, 145–169.
Why negative affectivity should not be controlled in job
stress research: Don’t throw out the baby with the bath
water. Journal of Organizational Behavior, 21, 79 –95. Received July 11, 2003
Steptoe, A. (2001). Psychophysiological bases of disease. In Revision received February 1, 2004
D. W. Johnston & M. Johnston (Eds.), Health psychol- Accepted February 16, 2004 y

View publication stats