A Meta-Analytic Examination of the Correlates of the Three Dimensions of Job Burnout

Abstract

This meta-analysis examined how demand and resource correlates and behavioral and attitudinal correlates were related to each of the 3 dimensions of job burnout. Both the demand and resource correlates were more strongly related to emotional exhaustion than to either depersonalization or personal accomplishment. Consistent with the conservation of resources theory of stress, emotional exhaustion was more strongly related to the demand correlates than to the resource correlates, suggesting that workers might have been sensitive to the possibility of resource loss. The 3 burnout dimensions were differentially related to turnover intentions, organizational commitment, and control coping. Implications for research and the amelioration of burnout are discussed.

Full text

Journal
of
Applied
Psychology
1996,
Vol.
SI.
No.
2,
123-133
Copyright
1996
by
the
American
Psychological
Association, Inc.
002I-90IO/96/S3.00
A
Meta-Analytic
Examination
of the
Correlates
of the
Three Dimensions
of Job
Burnout
Raymond
T. Lee
University
of
Manitoba
Blake
E.
Ashforth
Concordia
University
This
meta-analysis
examined
how
demand
and
resource
correlates
and
behavioral
and
attitudinal
correlates
were
related
to
each
of the 3
dimensions
of job
burnout.
Both
the
demand
and
resource
correlates
were
more
strongly
related
to
emotional
exhaustion
than
to
either
depersonalization
or
personal
accomplishment.
Consistent
with
the
conserva-
tion
of
resources
theory
of
stress,
emotional
exhaustion
was
more
strongly
related
to the
demand
correlates
than
to the
resource
correlates,
suggesting
that
workers
might
have
been
sensitive
to the
possibility
of
resource
loss.
The 3
burnout
dimensions
were
differ-
entially
related
to
turnover
intentions,
organizational
commitment,
and
control
coping.
Implications
for
research
and the
amelioration
of
burnout
arc
discussed.
Over
the
past dozen years,
the
phenomenon
of job
burnout
has
been investigated
in a
variety
of
service
oc-
cupations
and
settings.
The
construct
has
been linked
to
job
stress
and is
thought
to
represent
a
unique
response
to
frequent
and
intense
client-patient
interactions
(Cordes&Dougherty,
1993;
Maslach,
1982).
According
to
Maslach
(1982),
burnout consists
of
three dimensions: emotional exhaustion, depersonaliza-
tion
(felt
distance
from
others),
and
diminished personal
accomplishment. Recent
efforts
have
been made toward
understanding
how
various correlates
are
associated with
each
of the
three dimensions (Burke
&
Richardsen,
1993;
Cordes&
Dougherty,
1993;Shirom,
1989).
The
conservation
of
resources theory
of
stress
(Hobfoll,
1989;
Hobfoll
&
Freedy, 1993) provides
a
framework
for
understanding
how
such
correlates
are re-
Raymond
T.
Lee,
Department
of
Business
Administration.
University
of
Manitoba,
Winnipeg,
Manitoba,
Canada;
Blake
E.
Ashforth,
Department
of
Management,
Concordia
Univer-
sity,
Montreal.
Quebec,
Canada.
Portions
of
this
study
were
presented
at the
13th
World
Con-
gress
of
Sociology:
Sociology
of
Mental
Health
Working
Group,
Bielefeld,
Germany,
July
1994.
This
study
was
funded
in
part
by a
1993-1994
University
of
Manitoba,
Faculty
of
Management
research
support
grant.
We
are
indebted
to
Monica
Filz
for
assistance
with
the
data
collection
and
Stephen
Hladkyj
for
assistance
with
the
meta-
analysis.
Correspondence
concerning
this
article
should
be
addressed
to
Raymond
T.
Lee,
Department
of
Business
Administration,
Faculty
of
Management,
University
of
Manitoba,
Winnipeg,
Manitoba,
Canada
R3T
2N2.
Electronic
mail
may be
sent
via
Bitnet
to
lee@ccm.umanitoba.ca.
lated
to
burnout.
The
theory suggests that burnout
oc-
curs
when certain valued resources
are
lost,
are
inade-
quate
to
meet demands,
or do not
yield
the
anticipated
returns.
The
major
demands
of
work include role ambi-
guity,
role conflict,
stressful
events,
heavy
workload,
and
pressure.
The
major resources include social
support
from
various sources;
job
enhancement opportunities,
such
as
control, participation
in
decision
making,
and au-
tonomy;
and
reinforcement contingencies
(Burke
& Ri-
chardsen,
1993;
Cordes&
Dougherty, 1993).
The
theory
also
states that certain behavioral
and
attitudinal out-
comes
arc
likely
to
occur
as a
result
of
resource loss
and
burnout.
The
major outcomes include behavioral coping
responses;
turnover intentions;
and
erosion
of
organiza-
tional commitment,
job
involvement,
and job
satisfac-
tion
(Burke
&
Richardsen, 1993; Kahili, 1988).
Using
the
conservation
of
resources
framework, recent
studies have investigated
how
certain demand
and re-
source predictors
are
associated with each
of
the
burnout
dimensions
(Lee
&
Ashforth,
1993a;
Leiter,
1993;Leiter
&
Maslach, 1988).
Leiter
and
Maslach suggested that
work
demands
and
resources
are
potential sources
of
stress
in the
stress-strain-coping-self-evaluation
process
(Cummings
&
Cooper, 1979; Lazarus
&
Folkman,
1984).
Emotional exhaustion
as a
form
of
strain
is di-
rectly
affected
by
these correlates, whereas depersonaliza-
tion
as a
form
of
defensive
coping
and
personal accom-
plishment
as a
form
of
self-evaluation
are
indirectly
affected
through emotional exhaustion.
Leiter
and
Maslach's
(1988)
model
has
received par-
tial
support.
Lee and
Ashforth (1993a,
1993b)
found
several
of the
correlates
to be
associated with emotional
exhaustion
and
depersonalization,
but the
correlates
123
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

124
LEE
AND
ASHFORTH
were
only weakly
associated
with
personal
accomplish-
ment. Moreover,
Golembiewski's
(1989)
model
and Lee
and
Ashforth's
(1993b)
data have raised some doubts
re-
garding
the
sequential ordering among
the
burnout
di-
mensions
as
posited
by
Leiter
and
Maslach.
In
light
of
these
findings,
Leiter
{1993)
proposed
a re-
vised
model
in
which
the
demand
and
resource correlates
are
posited
to be
differentially
associated with
the
three
dimensions.
Specifically,
the
demands
are
posited
to be
more
strongly
related
to
emotional exhaustion, whereas
the
resources
are
posited
to be
more strongly
related
to
either
depersonalization
or
personal accomplishment.
As
Hobfoll
and
Freedy
(1993)
stated,
demands trigger strain
in
the
form
of
physical
and
emotional
exhaustion,
whereas
resources
help
to
overcome
the
need
for
defen-
sive
coping
and
enhance
one's
self-efficacy.
Leiter
(1991
b,
1993)
also posited that outcomes
reflecting
withdrawal
tendencies (e.g.,
escape
coping,
low
job
involvement,
and
desire
to
quit)
are
more
related
to
either
emotional
ex-
haustion
or
depersonalization, whereas outcomes
re-
flecting
positive
self-efficacy
(e.g.,
control coping
and fa-
vorable work attitudes)
are
more related
to
personal
accomplishment.
Leiter's
(1993)
revised model also
posits
how the
burn-
out
dimensions themselves
may be
interrelated (cf.
Go-
lembiewski,
1989;
Leiter,
1989).
Specifically,
it is now
believed
that
both emotional exhaustion
and
personal
ac-
complishment develop
in
parallel with
(rather
than
fol-
lowing
from)
each other. Hence, both
of
these dimensions
are
posited
to be
"reactions
to
different
aspects
of the
work
environment that
pose
difficulties
for
human ser-
vice
workers"
(Leiter,
1993,
p.
246).
Although
the
recent models
are
based
in
part
on
Hob-
foil's
(1989)
conservation
of
resources theory,
the
empir-
ical
evidence
has
been
piecemeal
and
limited
in
scope.
Thus,
our
meta-analysis
assessed
the
associations
be-
tween
various correlates
and
each
of the
burnout dimen-
sions. Only studies using
the
Maslach Burnout Inventory
(MBI;
Maslach
&
Jackson, 1981, 1986)
were
included
because
the
scale
is the
most
widely
used
operationaliza-
tion
of
burnout.
As a
multidimensional measure,
the
MBI
makes
it
possible
to
determine whether
the
symp-
toms
of
burnout
are
empirically distinct
from
each other
and
whether they
are
differentially
associated
with other
variables. Therefore,
our
study provides
a
summary
of
the
bivariate associations reported
in the
literature
and
has
implications
for the
diagnosis
and the
amelioration
of
burnout.
Method
Literature
Search
We
undertook both computer
and
manual searches
for
pub-
lished
articles that used
the
MBI.
Using
the key
phrase
"Mas-
lach Burnout Inventory,"
the
computer search scanned
the
ab-
stract
databases
of
clinical psychology,
education,
management,
nursing,
psychology,
social
work,
and
sociology
from
1982
to
1994.
The
manual search scanned
the
reference
lists
of
Burke
and
Richardsen
(1993),
Cordes
and
Dougherty
(1993),
Kahili
(1988),
and
Shirom
(1989)
for
published sources
not
found
in
the
aforementioned
abstracts.
The
search
yielded
77
sources,
of
which
58
provided
suffi-
cient
information
to
conduct
the
meta-analysis.
Of the
remain-
ing
19
sources,
11
examined
a
correlate unique
to
that study,
and 8 did not
provide information
on
reliability
estimates,
in-
tercorrelations among
the
variables,
or
both.
A
letter requesting
the
missing
information
was
sent
to
one
of
the
authors
of
each
of
these
8
articles. This request yielded
3
additional sources,
for
a
total
of
61.1
The
remaining
5 and the
11
with
unique
corre-
lates were excluded
from
the
analysis.
The
61
studies used
in the
meta-analysis covered
33
corre-
lates
based
on 56
independent samples.
The
sample sizes ranged
from
34 to 906
(M
=
198.07,
SD =
172.08).
Approximately
80%
of the
studies sampled human service providers
(e.g.,
teachers, nurses, counselors,
police
officers,
and
social
workers),
whereas
most
of the
remaining
studies sampled supervisors
and
managers
of
such
service
providers. Because
the
symptoms
of
burnout appear
to be
fundamentally
similar across hierarchical
levels
(Lee
&
Ashforth,
1993a, 1993b;
Maslach
&
Jackson,
1986),
we did not
differentiate
the
results
on
these
bases.
Each
study
was
coded
on
whether
the MBI
used
the
frequency
response format,
the
intensity response
format,
or
both
(see
Maslach
&
Jackson,
1981);
the
reliability estimates
of all
mea-
sures;
and the
correlations among
the
burnout dimensions
and
correlates.
Mela-Analytic
Procedure
We
followed
the
procedure
given
by
Hunter
and
Schmidt
(1990).
This procedure determines
the
extent
to
which
corre-
lation
differences
across
studies
can be
attributed
to
certain
sta-
tistical artifacts.
It is
only
after
such artifacts
are
taken into
ac-
count
that
the
corrected
correlations
between variables,
as
well
as the
variance
in
correlations
across
studies,
can be
estimated.
First,
we
calculated
the
uncorrected weighted mean correla-
tion
(/•)
by
using
the
zero-order correlations. Second, because
reliability
estimates might have varied
widely
across studies,
we
corrected
each
within-study
zero-order
correlation
for
mea-
surement
unreliability (Hunter
&
Schmidt, 1990). Where
reli-
ability
estimates
were
not
reported,
we
substituted
the
weighted
mean reliability
(see
Table
1)
into
the
correction
formula.
Then,
we
calculated
the
individually
corrected
weighted mean
correlation
(rc).
Because
so
few
studies reported
scale
ranges,
means,
and
standard deviations,
it was not
possible
to
correct
for
range restrictions.
Third,
we
computed
the
variance
of the
individually
cor-
rected weighted mean
correlation
after
removing
the
variances
1
The
correlations were provided upon request
for the
articles
by
Constable
and
Russell
(1986);
Ross,
Altmaier,
and
Russell
(1989);
and
Russell, Altmaier,
and Van
Velzen
(1987).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

META-ANALYSIS
OF JOB
BURNOUT
125
Table
1
Weighted
Mean Reliability
Estimates
oj
All
Variables
Correlate
Job
stressors
Role
ambiguity
Role
clarity
Role
conflict
Role
stress
Stressful
events
Workload
Work pressure
Physical
comfort
Support
Social support
Supervisor support
Coworker
support
Work
friends
Community
bond
Family
resources
Peer
cohesion
Team
cohesion
Job
enhancement opportunities
Autonomy
Innovation
Participation
Skill
utilization
Task
orientation"
k
6
6
10
3
4
5
5
4
6
12
13
2
4
2
5
2
11
4
6
6
4
N
1,518
929
2,612
703
903
1,238
681
433
1,010
3,377
3,340
153
433
244
657
156
2,177
433
1,763
1.489
433
«xx
.78
.85
.79
.76
.87
.76
.79
.80
.69
.85
.86
.80
.80
.70
.74
.86
.79
.80
.83
.80
.75
Correlate
Reinforcement
contingencies
Unmet expectations
Contingent rewards
Noncontingent rewards
Contingent punishments
Noncontingent punishments
Frequency
response format
Emotional exhaustion
Depersonalization
Personal accomplishment
Intensity
response format
Emotional
exhaustion
Depersonalization
Personal accomplishment
Behavioral
and
attitudinal
outcomes
Control coping
Preventive
coping
Turnover
intentions
Organizational commitment
Job
involvement
Job
satisfaction
Attitudes toward workplace climate
k
2
2
2
2
2
47
47
47
9
9
9
3
2
7
6
8
16
2
N
248
496
496
496
496
9,668
9,668
9,668
1,637
1,637
1,637
425
470
1,231
1,866
1,077
3,788
227
Rxx
.59
.96
.69
.84
.80
.86
.76
.77
.82
.78
.79
.83
.64
.78
.87
.72
.79
.80
Note,
k
=
number
of
samples providing
reliability
information;
N =
total number
of
individuals
in the k
samples;
7?Xx
=
weighted mean
reliability
estimate
of
each variable across
the k
samples.
"
Extent
to
which
the
work environment emphasizes
efficiency
and
good planning.
due to
sampling error
and
measurement unreliability.
We
also
computed
the
percentage
of
variance
due to
uncorrected
arti-
facts.
Fourth,
we
computed both
the
credibility
and
confidence
intervals
(Hunter
&
Schmidt, 1990).
The
credibility interval
was
based
on the
standard
deviation
of the
distribution
of
effect
sizes
after
correcting
for
statistical artifacts
and
provided
infor-
mation
on
whether moderators were
likely
to be
present.2
The
confidence
interval
was
based
on the
standard error
of the
mean
effect
size
and
provided information
on the
accuracy
of the in-
dividually
corrected weighted mean correlation estimate. Last,
we
conducted
a
test
for the
homogeneity
of
effect
sizes across
samples
by
using Hunter
and
Schmidt's
(1990)
Q
statistic.
Be-
cause
of
disagreements over
the
appropriate
ratio
of
artifacts
to
sample variance
to
infer
that
the
effect
sizes
are
homogeneous
(cf.
Hunter
&
Schmidt, 1990),
we
used
both
the
credibility
in-
terval
and the Q
statistic
to
detect moderators (Whitener,
1990).
Where
the Q
statistic could
not be
calculated,
we
used
only
the
credibility interval.
To
determine
if the
effect
sizes
are
homogeneous, Whitener
(1990,
p.
317)
recommended that
the
credibility interval
be
checked
to see if it is
"sufficiently
large"
or
includes zero. How-
ever,
it
is
unclear what
qualifies
as a
sufficiently
large interval.
Thus,
the
effect
sizes were considered homogeneous
if the
cred-
ibility
interval
did not
include zero
and the Q
statistic
was not
significant.
We
computed
the
confidence interval
for
homoge-
neous
effect
sizes
by
using
the
formula provided
by
Whitener
(1990,
p.
316).
The
effect
sizes were considered heterogeneous
if
the
credibility interval
did
include zero
or if the Q
statistic
was
significant
(either
case indicated that
a
moderator
was
probably
operating).
We
computed
the
confidence interval
for
heteroge-
neous
effect
sizes
by
using
the
formula provided
by
Whitener
(1990,
p.
317).
Results
Table
1
reports
the
number
of
studies,
the
total sample
size,
and the
weighted mean reliability
estimate
for
each
variable. Tables
2-5
report
the
uncorrected weighted
mean correlation,
the
individually corrected weighted
mean correlation,
the
variance,
the
lower
and
upper
range
of the
uncorrected weighted mean correlation
based
on the 95%
confidence interval,
and the
results
of
the Q
test
for
each correlate.
Mela-Correlations
Among
the
Burnout
Dimensions
Of
the 61
studies included
in the
meta-analysis,
33
used
the
frequency
response format,
11
used
the
intensity
2
A
credibility interval that includes zero indicates
the
prob-
ability
of a
disordinal
moderator, where
the
relationship
be-
tween
two
variables
is
positive
in
certain samples
and
negative
in
others. However,
a
moderator
is
still possible even when
the
interval
does
not
include zero
but is
large enough
to
suggest that
the
effect
sizes
are
heterogeneous
(Whitener,
1990).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

126
LEE
AND
ASHFORTH
response format,
and the
remaining
17
used both,
as
Maslach
and
Jackson
(1981)
did
originally.
Of the 17
studies
that used both formats,
9
provided
intradimen-
sional
correlations between response formats. Across
all
three
dimensions,
the
individually
corrected
weighted
mean
correlations were greater than .90, suggesting that
the two
formats
are
largely
redundant
and
that only
one
is
necessary (Maslach
&
Jackson, 1986;
Schaufeli,
Enz-
mann,
&
Girault, 1993).
Table
2
shows
the
meta-correlations
among
the
three
burnout dimensions.
For the
frequency
response
format,
emotional
exhaustion
was
strongly related
to
depersonal-
ization
(rc
=
.64),
whereas both dimensions were moder-
ately
negatively related
to
personal accomplishment
(rcs
=
—.33
for
emotional exhaustion
and
—.36
for
depersonalization).
The
individually
corrected
weighted
mean
correlations among
the
three dimensions
for the
intensity
response format were similar
in
magnitudes.
Given
the
high
correlations between
the two
response for-
mats
and
that only
a few
studies used
the
intensity
re-
sponse format,
all
subsequent
meta-analyses
examined
the
three dimensions using only
the
frequency
response
format.
Homogeneity
of
Effect
Sizes
Before
examining
the
individually corrected weighted
mean
correlations between
the
correlates
and
the
burn-
out
dimensions,
we
determined
the
number
of
homoge-
neous
effect
sizes.
For
emotional exhaustion,
the
effect
sizes
were homogeneous
for
13
of 32
correlates;
for de-
personalization,
the
effect
sizes were homogeneous
for 16
of
33
correlates;
and for
personal accomplishment,
the
effect
sizes were
horn
ogeneous
for
10
of
3
2
correlates.
Be-
cause
of the
small number
of
studies
per
variable,
it was
not
possible
to
conduct
any
moderator
analysis
on
corre-
lates
with heterogeneous
effect
sizes. Their individually
corrected weighted mean correlations should thus
be in-
terpreted with caution, especially
where
both
the
credi-
bility
and
confidence
intervals include zero
(Whitener,
1990).
Mela-Correlations
Between
the
Correlates
and the
Burnout
Dimensions
For the
demand correlates, Table
3
shows that
the
indi-
vidually
corrected weighted mean correlations between
emotional
exhaustion
and
role
conflict,
role stress, stress-
ful
events, workload,
and
work pressure were
all
greater
than
or
equal
to
.50. Table
4
shows that
the
individually
corrected weighted mean correlations between deperson-
alization
and
role ambiguity, role
conflict,
role stress,
stressful
events,
and
workload were
all
greater than
or
equal
to
.34. Table
5
shows
that
the
individually
corrected
weighted
mean
correlations
between personal accom-
plishment
and the
demand correlates were generally
weak,
with
the
highest being .22.
For the
resource correlates, Table
3
shows that
the in-
dividually
corrected
weighted mean correlations between
emotional exhaustion
and
social
support,
supervisor
sup-
port,
community
bond,
innovation,
participation,
unmet
expectations,
and
noncontingent punishment were
all
greater than
or
equal
to |
.301.
Table
4
shows that
the in-
dividually
corrected weighted mean correlations between
Table
2
Mela-Correlations
Among
the Job
Burnout
Dimensions
95%
CI
Burnout
dimension-burnout
dimension
k
N
r
rc
VAR
Lower
Upper
%
due
to
artifacts
Q
statistic
Frequency
response
format
Emotional
exhaustion-depersonalization
Emotional
exhaustion-personal
accomplishment
Depersonalization-personal
accomplishment
47
47
47
9,668
8,977
8,977
.52
-.27
-.29
.64
-.33
-.36
.029
.024
.053
.54
-.46
-.52
.75
-.21
-.26
88.76
55.69
88.00
36.40
10.44
49.01*
Intensity
response
format
Emotional
exhaustion-depersonalization
Emotional
exhaustion-personal
accomplishment
Depersonalization-personal
accomplishment
9
9
9
1,637
1,637
1.637
.53
-.18
-.27
.65
-.22
-.34
.050
.082
.085
.53
-.37
-.54
.77
-.08
-.15
86.67
86.69
87.48
69.53**
67.70**
71.92**
Note.
The
boldface
r^s
indicate
that
the
95%
credibility
interval
did not
include
zero.
In
cases
where
the
corrected
variances
were
less
than
zero,
the
Q
statistic
could
not be
calculated.
When
credibility
intervals
did not
include
zero
and the Q
statistic
was not
significant,
the
confidence
interval
(CI)
for
homogeneous
effect
sizes
was
constructed.
Otherwise,
the CI for
heterogeneous
effect
sizes
was
constructed,
k =
number
of
studies;
N =
combined
sample
size;
r =
uncorrccted
weighted
mean
correlation;
rc
=
weighted
mean
correlation
after
correcting
for
within-study
measurement
unreliability;
VAR
=
variance
of
^
after
correcting
for
sampling
error
and
measurement
unreliability;
%
due
to
artifacts
=
percentage
of
variance
due
to
artifacts
other
than
sampling
error
and
measurement
unreliability.
V
:£.()].
**/><.001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

META-ANALYSIS
OF JOB
BURNOUT
127
Table
3
Meta-Correlations
With
Emotional Exhaustion
(Frequency)
95%
CI
Correlate
Job
stressors (demands)
Role ambiguity
Role clarity
Role
conflict
Role
stress
Stressful
events
Workload
Work
pressure
Physical
comfort
Support (resources)
Social support
Supervisor
support
Coworker support
Work
friends
Community
bond
Family resources
Peer
cohesion
Team
cohesion
Job
enhancement opportunities (resources)
Innovation
Autonomy
Participation
Skill
utilization
Task
orientation'
Reinforcement
contingencies (resources)
Unmet
expectations
Contingent
rewards
Noncontingent
rewards
Contingent
punishments
Noncontingent
punishments
Behavioral
and
attitudinal outcomes
Control coping
Turnover intentions
Organizational
commitment
Job
involvement
Job
satisfaction
Attitudes toward workplace climate
k
6
6
11
3
4
6
5
4
6
13
14
2
4
2
5
2
4
11
6
7
4
2
2
2
2
2
3
7
7
8
17
2
N
1,518
929
2,824
703
903
1,450
681
433
1,010
3,589
3,552
153
433
244
657
156
433
2,177
1,763
1,701
433
248
496
496
496
496
425
1,231
2,078
1,077
4,000
227
r
.16
-.30
.42
.50
.45
.52
.41
-.09
-.26
-.31
-.18
,05
-.38
-.16
-.08
-.20
-.24
-.13
-.26
-.20
-.20
.38
-.24
.12
.09
.27
-.26
.37
-.38
-.03
-.26
-.11
rc
.21
-.35
.53
.62
.52
.65
.50
-.11
-32
-.37
-.22
.05
-.48
-.17
-.10
-.22
-.30
-.15
.31
-.28
-.26
.53
-.26
.15
.10
.32
-.30
.44
-.43
-.04
-.31
-.13
VAR
.215
.012
.032
.002
.039
.009
.006
<0
.018
.018
.019
.044
<0
.002
.207
.011
.008
.131
.002
.005
.002
.007
<0
<0
<0
<0
<0
.001
.002
.356
.264
<0
Lower
-.11
-.51
.41
.41
.33
.55
.35
-.28
-.48
-.49
-.35
-.28
-.64
-.35
-.45
-.44
-.48
-.35
.41
-.40
-.44
.36
-.38
.06
-.02
.23
-.46
.31
-.54
-.21
-.44
-.26
Upper
.54
-.19
.65
.72
.71
.75
.65
.05
-.16
-.25
-.08
.39
-.31
.00
.26
-.01
-.12
.10
-.20
-.16
-.07
.69
-.14
,24
.23
.40
-.14
.57
-.33
.13
-.19
.00
%
due to
artifact
97.51
64.01
89.01
41.18
92.36
69.70
48.62
65.94
82.64
79.88
71.10
12.77
94.48
42.51
39.13
94.78
41.43
49.35
11.90
35.04
62.25
39.75
96.65
98.02
Q
statistic
244.74**
16.67*
129.78**
7.93
52.88**
34.94**
9.84
19.72*
75.02**
69.58**
6.92*
2.34
90.77**
3.48
6.57
211.15**
10.31
14.33
4.55
11.30**
20.14*
11.66
238.80**
879.82**
Note.
The
boldface
rcs
indicate that
the
95%
credibility interval
did not
include zero.
In
cases
where
the
corrected
variances were less than zero,
the
Q
statistic could
not be
calculated. When
the
credibility interval
did not
include
zero
and the Q
statistic
was not
significant,
the
confidence interval
(CI)
for
homogeneous
effect
sizes
was
constructed. Otherwise,
the CI for
heterogeneous
effect
sizes
was
constructed,
k
=
number
of
studies;
N
=
combined
sample size;
r =
uncorrected weighted mean correlations;
rc
-
weighted mean correlation
after
correcting
for
within-study
measurement
unreliability;
VAR
-
variance
of
rc
after
correcting
for
sampling
error
and
measurement
unreliability;
% due to
artifacts
=
percentage
of
variance
due
to
artifacts other than sampling
error
and
measurement unreliability.
"
Extent
to
which
the
work environment emphasizes
efficiency
and
good
planning.
depersonalization
and
community
bond,
team
cohesion,
and
skill
utilization
were
all
greater
than
or
equal
to
|
-.341.
Table
5
shows
that
the
individually
corrected
weighted
mean
correlations
between
personal
accom-
plishment
and
work
friends,
participation,
and
team
ori-
entation
were
greater
than
or
equal
to
.
301.
In
summary,
13
of the 26
demand
and
resource
corre-
lates
had
individually
corrected
weighted
mean
correlations
greater
than
or
equal
to |
.301
with
emotional
exhaustion
(Table
3),
8 of the 26
demand
and
resource
correlates
had
individually
corrected
weighted
mean
correlations
greater
than
or
equal
to
].34|
with
depersonalization
(Table
4),
and 3 of the 26
demand
and
resource
correlates
had
indi-
vidually
corrected
weighted
mean
correlations
greater
than
or
equal
to
.
301
with
personal
accomplishment
(Table
5).
For
the
behavioral
and
attitudinal
correlates,
Tables
3-
5
show
that
the
burnout
dimensions
were
differentially
associated
with
three
of the
outcomes.
Emotional
ex-
haustion
was
positively
associated
with
turnover
inten-
tions
(rc
=
.44)
and
negatively
associated
with
organiza-
tional
commitment
(rc
=
-.43),
depersonalization
was
negatively
associated
with
organizational
commitment
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

128
LEE
AND
ASHFORTH
Table
4
Meta-Correlations
With
Depersonalization
(Frequency)
95%
Cl
Correlate
Job
stressors
(demands)
Role ambiguity
Role clarity
Role
conflict
Role
stress
Stressful
events
Workload
Work pressure
Physical
comfort
Support (resources)
Social support
Supervisor support
Coworker
support
Work
friends
Community
bond
Family
resources
Peer cohesion
Team cohesion
Job
enhancement opportunities (resources)
Innovation
Autonomy
Participation
Skill
utilization
Task
orientation"
Reinforcement
contingencies (resources)
Unmet expectations
Contingent rewards
Noncontingent
rewards
Contingent
punishments
Noncontingent
punishments
Behavioral
and
attitudinal outcomes
Control coping
Preventive
coping
Turnover
intentions
Organizational commitment
Job
involvement
Job
satisfaction
Attitudes
toward workplace climate
k
6
6
11
2
4
5
5
4
7
12
13
2
4
2
5
2
4
II
5
6
4
2
2
2
2
2
3
2
4
7
6
12
2
A"
1,518
929
2,824
286
903
1,033
681
433
1,176
3,172
3,135
153
433
244
657
156
433
2,177
1,346
1,663
433
496
496
496
496
496
425
470
564
2.078
749
2,102
227
/•
.25
-.28
.27
.41
.39
.24
.18
.04
-.16
-.19
-.17
-.07
-.36
-.15
-.14
-.28
-.22
-.10
-.14
-.29
-
.23
.13
-.10
.18
.08
.20
-.24
-.25
.24
-.33
-.09
-.33
-.12
re
.34
-.27
.37
.54
.50
.34
.22
.06
-.21
-.24
-.22
-.08
-.46
-.21
-.20
-.34
-.28
-.13
-.17
-.39
-.29
.19
-.11
.25
.10
.25
-.28
-.37
.31
-.42
-.11
-.44
-.15
VAR
.030
.004
.056
<0
.051
.045
<0
.002
.070
.004
.015
.019
.016
.011
.189
.024
.007
.068
.010
<0
.016
.007
<0
<0
<0
<0
<0
<0
<0
.008
.163
.136
<0
Lower
.20
-
.41
22
.30
.28
.17
.06
-.14
-.41
-.36
-.34
-.35
-.62
-.39
.35
-.71
-.46
-.31
-.29
-.50
-.47
.07
-.19
.13
-.03
.17
-.43
-.46
.18
-.53
-.29
-.57
-.28
Upper
.49
.11
.52
.68
.70
.52
.39
.25
.00
-.12
-.09
-.08
-
.29
-.04
.30
-.12
-.10
.05
-.05
-.28
-.11
.31
-.02
.37
.22
.34
-.12
-.28
.44
-.32
.06
-.31
-.03
%
due to
artifact
84.32
34.04
91.57
91.62
83.95
11.27
87.49
39.16
71.65
47.87
63.32
40.91
92.90
61.88
37.22
89.56
64.12
55.77
44.32
59.94
92.32
95.29
Q
statistic
39.02"
9.23
130.82"
58.45**
34.39**
4.51
57.43**
20.52*
47.19**
3.84
10.91
3.38
70.45**
5.30
6.37
105.37**
14.16*
9.10
3.81
25.55**
78.34**
262.89**
Note.
The
boldface
r^
indicate that
the 95%
credibility
interval
did not
include zero.
In
cases where
the
corrected
variances were
less
than
zero,
the
Q
statistic could
not
be
calculated. When
the
credibility interval
did not
include zero
and the Q
statistic
was not
significant,
the
confidence
interval
(CI)
for
homogeneous
effect
sizes
was
constructed.
Otherwise,
the
CI
for
heterogeneous
effect
sizes
was
constructed,
k =
number
of
studies;
N -
combined
sample size;
r =
uncorrected weighted mean correlation;
rc
=
weighted mean correlation
after
correcting
for
within-study
measurement
unreliability;
VAR =
variance
of
rc
after correcting
for
sampling error
and
measurement
unreliability;
% due to
artifacts
=
percentage
of
variance
due
to
artifacts
other
than
sampling error
and
measurement
unreliability.
*
Extent
to
which
the
work environment emphasizes
efficiency
and
good
planning.
(rc
=
-.42)andjobsatisfaction(rc
=
—.44),
and
personal
accomplishment
was
positively
associated
with control
coping
(rc
=
.52).
Discussion
Major
Findings
The
meta-correlations
found
among
the
three
burnout
dimensions
are
similar
to
those
reported
in
Maslach
and
Jackson's
(1986)
manual
and in
other recent reviews
(Cordes&
Dougherty,
1993;
Schaufelietal.,
1993).
The
results
are
also consistent
with
Leiter's
f
1993)
belief that
personal accomplishment develops largely independently
of
emotional exhaustion
and
depersonalization.
The
patterns
of
associations
between
the
demand
and
resource correlates
and the
three dimensions
appear
to
be
consistent with
the
conservation
of
resources explana-
tion
of
burnout
(Hobfoll
&
Freedy, 1993; Leiter, 1993).
Five
of the
eight demand
correlates
were strongly
(rcs
>
.50)
associated with emotional exhaustion, whereas only
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

META-ANALYSIS
OF JOB
BURNOUT
129
Table
5
Meta-Correlations
With
Personal
Accomplishment
(Frequency)
95%
CI
Correlate
Job
stressors
(demands)
Role
ambiguity
Role
clarity
Role
conflict
Role
stress
Stressful
events
Workload
Work
pressure
Physical
comfort
Support (resources)
Social
support
Supervisor
support
Coworker
support
Work
friends
Community
bond
Family
resources
Peer cohesion
Team cohesion
Job
enhancement opportunities (resources)
Innovation
Autonomy
Participation
Skill
utilization
Task
orientation8
Reinforcement
contingencies (resources)
Unmet
expectations
Contingent
rewards
Noncontingent
rewards
Contingent
punishments
Noncontingent
punishments
Behavioral
and
attitudinal outcomes
Control coping
Turnover
intentions
Organizational
commitment
Job
involvement
Job
satisfaction
Attitudes
toward workplace climate
k
6
6
10
3
4
7
5
4
7
13
15
2
4
2
5
2
4
11
6
6
4
•}
2
•y
2
2
3
2
7
6
1 ]
2
N
1,131
929
2,271
703
903
1,616
681
433
1,176
3,589
3,614
153
433
244
657
156
433
2,177
1,763
1,663
433
496
496
496
496
496
425
286
2,078
749
2,241
227
r
.09
-.03
-.15
.00
.18
-.07
-.07
-.06
.16
.11
.10
.37
-.21
-.09
-.17
.09
-.21
.06
.23
.13
-.26
-.13
.13
-.13
.02
-.13
.42
-.13
-.02
-.10
.22
.23
rc
.11
-.03
-.21
.00
.22
-.09
-.09
-.07
.20
.14
.13
.49
-.26
-.12
-.22
.12
-.26
.07
.30
.24
-.33
.19
.14
-.17
.02
-.16
.52
-.16
-.02
-.11
.29
.28
VAR
.374
.066
.105
.040
.160
.068
.007
<0
.114
.073
.083
<0
.010
<0
.078
.345
.094
.103
.017
.178
.003
.002
<0
<0
<0
<0
<0
<0
.132
.182
.394
<0
Lower
-.29
-.26
-.41
-.21
-.13
-.29
-.27
-.26
-.04
-.03
-.04
.36
-.44
-.27
-.47
-.56
-.56
-.13
.16
-.04
-.51
-.31
.02
-.29
-.09
-.25
.42
-.05
-.27
-.29
-.16
.16
Upper
.51
.20
-.02
.21
.57
.09
.09
.11
.44
.30
.30
.62
-.07
.02
.03
.79
.03
.28
.43
.52
-.16
-.07
.27
-.04
.15
.07
.62
-.27
.22
.21
.42
.40
%
due to
artifact
97.71
87.26
93.94
84.27
96.56
89.79
40.37
92.01
93.01
93.11
46.16
86.53
94.41
88.40
92.91
78.95
97.02
19.96
24.56
96.26
92.19
98.28
Q
statistic
26
1
.52**
47.10**
165.13**
19.07**
116.88**
68.60**
8.39
90.11**
187.79**
219.22**
7.43
37.15**
35.92**
34.18**
155.18"
29.42**
202.69**
5.04
2.90
187.24**
76.96**
641.61**
Note.
The
boldface
rcs
indicate that
the 95%
credibility interval
did not
include zero.
In
cases where
the
corrected variances were
less
than zero,
the
Q
statistic could
not be
calculated.
When
the
credibility
interval
did
not
include
zero
and the Q
statistic
was not
significant,
the
confidence interval
(CI)
for
homogeneous
effect
sizes
was
constructed. Otherwise,
the CI for
heterogeneous
effect
sizes
was
constructed,
k =
number
of
studies;
A'
=
combined
sample size;
r
=
uncorrectcd weighted mean
correlation;
rc
=
weighted mean correlation
after
correcting
for
within-study
measurement
unreliability;
VAR =
variance
of
rc
after
correcting
for
sampling error
and
measurement
unreliability;
% due to
artifacts
=
percentage
of
variance
due to
artifacts
other than sampling error
and
measurement unreliability.
'
Extent
to
which
the
work
environment emphasizes
efficiency
and
good
planning.
"ps.OOI.
unmet expectations
as a
resource correlate
had the
same
magnitude
of
association with emotional exhaustion
(rc
=
.53).3
Consistent with
the
primacy
of
resource loss
ar-
gument made
by
Hobfoll
and
Freedy,
the
results suggest
that individuals
may be
more sensitive
to
demands
placed
on
them than
to the
resources received.
Work
demands
are
generally
perceived
to be
losses
be-
cause meeting such demands requires
the
investment
of
valued
resources,
viewed
as
gains
(Hobfoll
&
Freedy,
1993).
According
to
prospect
theory,
to the
extent
that
individuals
overweigh
the
consequences
of
losses,
the re-
sources
expended
to
prevent
further
loss
are
greater than
the
threat
of
losses
presented
by the
demands
(Tversky
&
Kahneman,
1981).
This overcompensation
may
partly
explain
why
service workers seem
to be
more adversely
3
Expectations
can be
perceived
as
either
resources
or de-
mands,
depending
on the
degree
to
which
they
are
met.
Unmet
expectations
would
likely
be
viewed
as
demands
because
they
threaten
one's
work
adjustment,
whereas
met
expectations
would
likely
be
viewed
as
resources
because
they
confirm
or
support
one's
adjustment.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

130
LEE
AND
ASHFORTH
affected
by the
stressors.
For
many service providers,
the
key
decision
is the
amount
of one or
more
resources
they
need
to
invest
to
meet
the
demands
and to
protect them-
selves
from
further
resource
depletion. Strain occurs
when
the
workers
feel
they
no
longer
have
sufficient
emotional
resources
to
handle
the
interpersonal stressors (cf. Hob-
foil,
1989; Maslach, 1982). Instead, many
adopt
the de-
fensive
strategy
of
withdrawal
(rather
than engagement)
through depersonalization (Lee
&
Ashforth,
1993a).
Thus,
it is not
surprising
to
find
that depersonalization
was
also
strongly associated
with
role stress
(rc
=
.54)
and
stressful
events
(rc
=
.50).
A
surprising
finding was the
weak
associations
(i.e.,
rcs
< ±
.30)
between personal accomplishment
and
most
of
the
resource correlates.
The two
exceptions were work
friends
(rc
=
.49)
and
participation
(rc
=
.30).
To the
extent that work acquaintances provide individuals with
a
sense
of
competence through support,
self-efficacy
and
self-appraisal
may be
heightened
(Jayaratne
&
Chess,
1986).
Similarly,
to the
extent that participation leads
to
greater
felt
mastery
and
control
(i.e.,
empowerment)
over
the
work environment,
self-efficacy
and
self-appraisal
may
also
be
enhanced.
Although
gains
in
resources
are
viewed
by
most work-
ers as
less salient than
the
prevention
of
loss, gains
are not
inconsequential (Hobfoll
&
Freedy,
1993). Resources
may
directly compensate
for
certain losses,
but
they
may
also
indirectly
help workers cope
with
work
demands.
For
example,
with weak social
support,
interpersonal stres-
sors
will
likely
lead
to
emotional strain,
but
with strong
support, their relationship
will
likely
be
attenuated
(Jayaratne
&
Chess,
1986;
Russell, Altmaier,
& Van
Vel-
zen,
1987).
This proposition
is
consistent
with
interac-
tionist
accounts
of
organizational behavior (see Mitchell
&
James, 1989).
Consistent with Leiter's
(1993)
model,
our findings re-
veal
that
the
three dimensions were
differentially
associ-
ated
with
several
of the
behavioral
and
atliludinal corre-
lates.
Emotional exhaustion
and
depersonalization were
strongly
associated with turnover intentions
and
organi-
zational
commitment
but
were weakly associated with
control coping.
As a
caveat,
it is
unclear whether
the
weak
associations reflect
the
underuse
of
such
a
coping behav-
ior or its
lack
of
effectiveness
and
subsequent abandon-
ment.
In any
case,
the
lack
of an
effective
coping response
might
have served
to
reinforce
subsequent
feelings
of
helplessness
and
futility
(Lee
&
Ashforth,
1993a).
In
contrast, personal accomplishment
was
strongly related
to
control coping, suggesting that
a
problem-focused
re-
sponse
and a
positive self-appraisal
may be
mutually
re-
inforcing
(Lazarus
&Folkman,
1984;cf.
Leiter,
1991b).
Our findings
suggest that outcomes that stem
from
emotional exhaustion
and
reflect
the
desire
to
withdraw
may
be
offset
by
outcomes that stem
from
personal
ac-
complishment
and
reflect
the
desire
to
seek control.
Whichever response (either withdrawing
or
seeking
control)
predominates
may
thus depend
on the
relative
strengths
of
emotional
exhaustion
and
personal
accom-
plishment
as
each develops independently over
time
(Lee
&
Ashforth,
1993a; Leiter, 1993).
Study
Limitations
One
limitation
is
that
the
meta-correlations
show only
the
strength
of the
relations among
the
burnout dimen-
sions
and by no
means deal with
the
controversy regard-
ing
their sequential ordering (Golembiewski, 1989;
Leiter,
1989).
The
arguments over
the
correct sequence
are
based
on
circumstantial evidence
of how
strongly
each
dimension
is
related
to
certain antecedent
and
out-
come variables
(Cordes
&
Dougherty, 1993; e.g.,
see Lee
&
Ashforth,
1993b).
A
second limitation
is
that
the
strong
association
found
between
emotional exhaustion
and five of the
demand
stressors
may be
partly attributed
to a
measurement
arti-
fact.
Three
of
Maslach
and
Jackson's
(1981,
1986)
emo-
tional exhaustion
scale
items (i.e.,
"I
feel
emotionally
drained
from
my
work," "Working with people directly
puts
too
much stress
on
me,"
and
"Working with people
all
day is
really
a
strain
for
me")
are
direct
measures
of
individual
stress reactions. These items correspond
closely
to
items
that
tap
perceived stress
found
in
fre-
quently
used scales
of
role overload, role
conflict,
role
ambiguity,
stressful
events,
and
work pressure.
A
third limitation
is
that only certain demands
and re-
sources were examined here.
The
fact
that emotional
ex-
haustion
was
strongly
associated
with several
of the
cor-
relates does
not
necessarily imply
the
absence
of
other
correlates that
may be
more strongly
associated
with
the
other
two
dimensions.
Two
sets
of
correlates
not
examined here
are
demo-
graphic
(e.g., sex, age,
and
education)
and
dispositional
attributes (e.g., Type
A
personality, psychological hardi-
ness,
and
locus
of
control). Although these attributes
have
been linked
to
burnout
(Cordes
&
Dougherty,
1993;
Mclntyre,
1984;Nowack,
1986),
their
potential
as
mod-
erators requires
further
clarification
and
investigation.
Implications
for
Future
Research
Although
the
demand
and
resource correlates
of job
burnout
have
been
the
focus
of
much attention, their
links
to
stress-coping
and
adjustment require
further
study.
According
to the
primacy
of
loss hypothesis, indi-
viduals
are
more sensitive
to
demands
and
thus overcom-
pensate
in the
amount
of
resources expended
to
prevent
further
loss
(Hobfoll
&
Freedy,
1993).
To the
extent that
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

META-ANALYS1S
OF JOB
BURNOUT
131
burnout develops because
of
this
overcompensation,
two
questions
are
worth
investigating:
(a) To
what extent
can
individuals
deal
with
or
even overcome burnout
by
reap-
praising their demands
as
potential
gains
(challenges)
rather
than certain losses (obstacles; Lazarus
&
Folkman,
1984;
Tversky
&
Kahneman,
1981)
and (b) if
demands
can
indeed
be
reframed
as
potential gains,
how can re-
sources
be
used
to
change
the way
stressors
are
appraised?
For
example, with
the
right kind
of
supervisory support,
workers
may
come
to
perceive ambiguous role expecta-
tions
as
opportunities
to
carry
out
their
own
initiatives
(potential gains) rather than
as
restrictions
on
their
ac-
tions (certain
losses).
A
broader question
that
needs
to be
addressed
is the
extent
to
which
the
burnout phenomenon
is
generaliz-
able
to
occupations other than human services (Shirom,
1989).
Perhaps
the
most distinguishing aspect
of
burn-
out is not so
much
its
psychological
and
behavioral symp-
toms
but
rather
the
demands
and the
resources perceived
to be the
most salient
in
service
settings
(cf.
Cordes
&
Dougherty,
1993).
Service providers
may
work
in
situa-
tions
with
demands
and
resources that
are
different
from
those experienced
by,
say, production workers
in
manu-
facturing
settings.
If
true, this suggests that stress
re-
searchers should
identify
and
include demands
and
re-
sources perceived
by
their study participants
to be the
most relevant
for
their
work
context (Maslach,
1982).
Moreover,
a
comparison
of the
perceived importance
of a
common core
of
demands
and
resources between
service
and
nonservice workers
may
reveal
how and why
stressful
situations
are
dealt with
differently
between occupations.
Last, longitudinal designs
may
show
that
how
workers
handle
burnout
is
guided
by the
type
and the
frequency
of
feedback
on the
usefulness
of
their initial coping
be-
haviors
(Leiter,
1991
b).
This longitudinal approach sug-
gests several forms
of
reciprocal relationships
involving
burnout
and
adjustment
responses (cf. Cummings
&
Cooper,
1979).
Multiple
data
points over time
may
thus
help determine
the
kinds
of
feedback
required
to
amelio-
rate
or
prevent burnout. Such longitudinal
data
collec-
tion should lead
to a
better
understanding
of how
burnout
develops
and
contribute
to the
broader research
on
stress-coping.
References
References
marked with
an
asterisk indicate studies included
in
the
meta-analysis.
*Anderson,
M. B. G., &
Iwanicki,
E. F.
(1984).
Teacher moti-
vation
and its
relationship
to
burnout.
Educational
Admin-
istration
Quarterly,
20,
109-132.
*Barrick,
R.
K.
(1989).
Burnout
and job
satisfaction
of
voca-
tional
supervisors. Journal
of
Agricultural Education,
30, 35-
41.
•Bloom,
K.
D.,
Buhrke,
R.
A.,
&
Scott,
T. B.
(1988).
Burnout
and job
expectations
of
state agency rehabilitation counsel-
lors
in
North Dakota.
Journal
of
Applied
Rehabilitation
Counseling,
19,
32-36.
*Brookings,
J.
B.,
Bolton,
B.,
Brown,
C.
E.,
&
McEvoy,
A.
(1985).
Self-reported
job
burnout among
female
human ser-
vice
professionals. Journal
of
Occupational Behaviour,
6,
143-150.
Burke,
R.
J.,
&
Richardscn,
A. M.
(1993).
Psychological burn-
out
in
organizations. Unpublished manuscript,
York
Univer-
sity,
Toronto,
Ontario, Canada.
•Clanton,
L.
D.,
Rude,
S. S., &
Taylor,
C.
(1992).
Learned
re-
sourcefulness
as a
moderator
of
burnout
in a
sample
of
reha-
bilitation providers. Rehabilitation
Psychology,
37,
131
-140.
*Constable,
J.
F.,
&
Russell,
D. W.
(1986).
The
effect
of
social
support
and the
work environment upon burnout among
nurses.
Journal
of
Human
Stress,
11,
20-26.
Cordes,
C.
L.,
&
Dougherty,
T. W.
(1993).
A
review
and
inte-
gration
of
research
on job
burnout. Academy
of-Management
Review,
18,
621-656.
*Corrigan,
P.
W.,
Holmes.
E. P.,
Luchins,
D.,
Buican,
B.,
Basil,
A.,
&
Parks,
J. J.
(1994).
Staff
burnout
in a
psychiatric hos-
pital:
A
cross-lagged panel design. Journal
of
Organizational
Behaviour,
15,
65-74.
Cummings,
T.
G.,
&
Cooper,
C. L.
(1979).
A
cybernetic
frame-
work
for
studying occupational stress. Human Relations,
32,
395-418.
*Davis-Sacks,
M.
L.,
Jayaratne,
S.,
&
Chess,
W. A.
(1985).
A
comparison
of the
effects
of
social support
on
the
incidence
of
burnout. Social
Work,
30,
40-44.
*Dignam,
J. T,
Barrera,
M.,
Jr.,
&
West,
S. G.
(1986).
Occupa-
tional stress, social support,
and
burnout among correctional
officers.
American Journal
of
Community
Psychology,
14,
177-193.
*Dolan,
S.
L.,
&
Renaud,
S.
(1992).
Individual, organizational
and
social determinants
of
managerial burnout:
A
multivari-
ate
approach.
Journal
oj'Social
Behaviour
and
Personality.
7,
95-110.
*Eisenstat,
R.
A.,
&
Felner,
R. D.
(1984).
Toward
a
differenti-
ated
view
of
burnout:
Personal satisfaction
and
organizational
mediators
of job
satisfaction
and
stress.
American
Journal
of
Community
Psychology,
12.
411-430.
*Erera,
I.
(1992).
Supervisors
can
burn-out too. Clinical
Su-
pervisor,
9,
131-148.
*Fong,
C. M.
(1993).
A
longitudinal study
of the
relationships
between overload, social
support,
and
burnout among nursing
educators.
Journal
of
Nursing
Education,
32,
24-29.
*Friesen,
D.,
&
Sarros,
J. C.
(1989).
Sources
of
burnout among
educators. Journal
of
Organizational Behaviour,
10,
179-
188.
*Gaines,
J.,
&
Jermier,
J. M.
(1983).
Emotional exhaustion
in
a
high
stress
organization. Academy
oj
Management
Journal,
26,
567-586.
Golembiewski,
R. T.
(1989).
A
note
on
Leiter's
study:
High-
lighting
two
models
of
burnout.
Group
and
Organization
Studies,
14,
5-13.
Hobfoll,
S. E.
(1989).
Conservation
of
resources:
A new at-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

132
LEE
AND
ASHFORTH
tempt
at
conceptualizing stress. American
Psychologist,
44,
513-524.
Hobfoll,
S. E., &
Freedy,
J.
(1993).
Conservation
of
resources:
A
general stress theory applied
to
burnout.
In W. B.
Schaufeli,
C.
Maslach,
& T.
Marek
(Eds.),
Professional
burnout: Recent
developments
in
theory
and
research (pp.
115-129).
Wash-
ington,
DC:
Taylor
&
Francis.
*Huebner,
E. S.
(1992).
Burnout among
school
psychologists:
An
exploratory investigation into
its
nature, extent,
and
cor-
relates.
School
Psychology
Quarterly.
7,
129-136.
Hunter,
J.
E.,
&
Schmidt,
F.
L.
(1990).
Methods
ofmeta-analy-
sis:
Correcting error
and
bias
in
research
findings.
Newbury
Park,
CA:
Sage.
"Jackson,
S.
E.,
Schwab,
R.
L.,
&
Schuler,
R. S.
(1986).
Toward
an
understanding
of the
burnout phenomenon. Journal
oj
Ap-
plied
Psychology,
71,
630-640.
"Jackson,
S.
E.,
Turner,
J.
A.,
&
Brief,
A. P.
(1987).
Correlates
of
burnout among public service lawyers. Journal
of
Occupa-
tional Behaviour,
8,
339-349.
"Jayaratne,
S.,
&
Chess,
W. A.
(1984).
The
effects
of
emotional
support
on
perceived
job
stress
and
strain.
Journal
of
Applied
Behavioral Science,
20,
141-153.
"Jayaratne,
S.,
&
Chess,
W. A.
(1986).
Job
stress,
job
deficit,
emotional
support,
and
competence:
Their relationship
to
burnout. Journal
of
Applied
Social Sciences,
10,
135-151.
•Jorgenson,
D.
M.
(1986).
Burnout
in
nurse
faculty:
Relation-
ships
with
management
style,
collegia!
support,
and
work
load
in
collegiate programs. Journal
of
Professional
Nursing,
2,
252-260.
"Jorgenson,
D. M.
(1992). Burnout
in
doctorally
prepared
nurse
family.
Journal
of
Nursing Education,
31,
341-346.
Kahili,
S.
(1988).
Symptoms
of
professional
burnout:
A
review
of the
empirical evidence. Canadian
Psychology,
29,
284-
297.
*Koeske,
G.
F.,
&
Koeske,
R. D.
(1989).
Construct validity
of
the
Maslach Burnout Inventory:
A
critical review
and
recon-
ceptualization. Journal
of
Applied
Behavioral Science,
25,
131-132.
"Kruger.
L.
J.,
Botman,
H.
I.,
&
Goodenow,
C.
(1991).
An in-
vestigation
of
social support
and
burnout among residential
counsellors. Child
and
Youth
Care Forum,
20,
335-352.
Lazarus,
R.
L.,
&
Folkman,
S.
(1984).
Stress,
appraisal,
and
coping.
New
\brk:
Springer.
*Lee,
R. T, &
Ashforth,
B. E.
(1990).
On the
meaning
of
Mas-
lach's
three
dimensions
of
burnout. Journal
of
Applied
Psy-
chology,
75,
743-747.
*Lee,
R. T, &
Ashforth,
B. E.
(1992).
Work-unit structure
and
processes
and
job-related stressors
as
predictors
of
manage-
rial
burnout. Journal
of
Applied
Social
Psychology,
21,
1831-
1847.
"Lee,
R. T, &
Ashforth,
B. E.
(1993a).
A
further examination
of
managerial burnout: Toward
an
integrated
model. Journal
of
Organizational Behavior,
14,
3-20.
"Lee,
R.
T.,
&
Ashforth,
B. E.
(1993b).
A
longitudinal study
of
burnout among supervisors
and
managers: Comparisons
between
the
Leiter
and
Maslach
(1988)
and
Golembiewski
et
al.
(1986)
models.
Organizational
Behaviour
and
Human
Decision
Processes.
54,
369-398.
"Leiter,
M. P.
(1988a).
Burnout
as a
function
of
communica-
tion
patterns. Group
and
Organization Studies,
13,
111
-128.
"Leiter,
M. P.
{1988b).
Commitment
as a
function
of
stress
re-
actions among nurses:
A
model
of
psychological evaluations
of
work settings. Canadian Journal
of
Community Menial
Health,
7,
115-132.
Leiter,
M. P.
(1989).
Conceptual implications
of two
models
of
burnout:
A
response
to
Golembiewski. Group
and
Organiza-
tion
Studies,
14,
15-22.
"Leiter,
M. P.
(1990).
The
impact
of
family
resources,
control
coping,
and
skill
utilization
on the
development
of
burnout:
A
longitudinal
study.
Human Relations,
11,
1067-1083.
"Leiter,
M. P.
(1991a).
Burnout
as a
crisis
in
professional role
structures:
Measurement
and
conceptual issues. Anxiety,
Stress,
and
Coping,
5,
79-93.
"Leiter,
M. P.
(199
Ib).
Coping
patterns
as
predictors
of
burn-
out:
The
function
of
control
and
escapist coping patterns.
Journal
of
Organizational Behaviour,
12,
123-144.
Leiter,
M. P.
(1993).
Burnout
as a
developmental
process:
Con-
sideration
of
models.
In W. B.
Schaufeli,
C.
Maslach,
& T.
Marek
(Eds.),
Professional
burnout: Recent developments
in
theory
and
research (pp.
237-250).
Washington,
DC:
Taylor
&
Francis.
"Leiter,
M. P., &
Maslach,
C.
(1988).
The
impact
of
interper-
sonal
environment
on
burnout
and
organizational commit-
ment. Journal
of
Organizational Behavior,
9,
297-308.
"Lewis,
S.
L.,
Becktell,
P.
J.,
Bonner,
P. N.,
Campbell,
M.
A.,
Cooper,
C.
L.,
&
Hunt,
W. C.
(1992).
Work stress, burnout,
and
sense
of
coherence among dialysis nurses. American
Ne-
phrology
Nurse's
Association
Journal,
19,
545-553.
Maslach,
C.
(1982).
Burnout:
The
cost
of
caring. Englewood
Cliffs,
NJ:
Prentice
Hall.
"Maslach,
C.,
&
Florian,
V.
(1988).
Burnout,
job
setting,
and
self-evaluation
among rehabilitation counsellors. Rehabilita-
tion
Psychology,
33,
85-93.
Maslach,
C.,
&
Jackson,
S. E.
(1981).
The
measurement
of
ex-
perienced
burnout. Journal
of
Occupational
Behavior,
2, 99-
113.
Maslach,
C.,
&
Jackson,
S. E.
(1986).
Maslach Burnout Inven-
tory
manual (2nd ed.).
Palo
Alto,
CA:
Consulting Psycholo-
gists
Press.
Mclntyre,
T. C.
(1984).
The
relationship between locus
of
con-
trol
and
teacher burnout.
British
Journal
of
Educational Psy-
chology,
54,
235-238.
"McMullen,
M.
B.,
&
Krantz,
M.
(1988).
Burnout
in day
care
workers:
The
effects
of
learned helplessness
and
self-esteem.
Child
&
Youth
Care
Quarterly,
17,
275-280.
"Miller,
K.
I.,
Ellis,
B.
H.,
Zook,
E.
G.,
&
Lyles,
J. S.
(1990).
An
integrated model
of
communication,
stress,
and
burnout
in
the
workplace. Communication Research,
17,
300-326.
"Miller,
K. L,
Stiff,
J.
B.,
&
Ellis,
B. H.
(1988).
Communication
and
empathy
as
precursors
to
burnout among human service
workers.
Communication
Monographs,
55,
250-265.
Mitchell,
T.
R.,
&
James,
L. R.
(1989).
Theory development
forum:
Situational
versus
dispositional factors: Competing
explanations
of
behavior; Conclusions
and
future
directions.
Academy
of
Management
Review,
14,
331-332,401-407.
"Nagy,
S.
(1985).
Burnout
and
selected variables
as
compo-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

META-ANALYSIS
OF JOB
BURNOUT
133
nents
of
occupational stress.
Psychological
Reports,
56,
195-
200.
Nowack,
K..
M.
(1986).
Type
A,
hardiness,
and
psychological
distress. Journal
of
Behavior
Medicine,
9,
537-547.
*Nowack,
K.
M.,
&
Hanson,
A.
L.
(1983).
The
relationship
between
stress,
job
performance,
and
burnout
in
college stu-
dent
resident assistants. Journal
of
College
Student
Personnel,
23,
545-550.
*Oehler,
J.
M.,
&
Davidson,
M. G.
(1992).
Job
stress
and
burn-
out
in
acute
and
nonacute
pediatric
nurses.
American Journal
of
Critical
Care,
7,81-90.
*Oehler,
J.
M.,
Davidson,
M.
G.,
Starr,
L. E., &
Lee,
D. A.
(1991).
Burnout,
job
stress, anxiety,
and
perceived social
sup-
port
in
neonatal nurses. Heart
and
Lung,
20,
500-505.
*Oktay,
J. S.
(1992).
Burnout
in
hospital social
workers
who
work
with
AIDS patients. Social
Work,
37,
432-438.
•Penn,
M.,
Romano,
J.
L.,
&
Foat.
D.
(1988).
The
relationship
between
job
satisfaction
and
burnout:
A
study
of
human ser-
vice
professionals.
Administration
in
Mental
Health,
15,
157-
165.
*Randall,
M.,
&
Scott,
W. A.
(1988).
Burnout,
job
satisfaction,
and
job
performance.
Australian
Psychologist,
23,
335-347.
*Richardsen,
A.
M.,
Burke,
R.
J.,
&
Leiter,
M. P.
(1992).
Oc-
cupational demands, psychological burnout
and
anxiety
among hospital personnel
in
Norway. Anxiety, Stress,
and
Coping,
5,
55-68.
*Rimmerman,
A.
(1989).
Burnout
among beginning rehabili-
tation workers
in
Israel
and its
relationship
to
social support,
supervision,
and job
satisfaction.
Rehabilitation Counseling
Bulletin,
32,
243-247.
*Ross,
R.
R.,
Altmaier,
E.,
&
Russell,
D. W.
(1989).
Job
stress,
social support,
and
burnout among counseling center
staff.
Journal
a/Counseling
Psychology,
36,
464-470.
*Russell,
D.
W.,
Altmaier,
E.,
& Van
Vclzen,
D.
(1987).
Job-
related
stress, social support,
and
burnout among
classroom
teachers. Journal
of
Applied
Psychology,
72,
269-274.
*Sarros,
J.
C,
&
Friesen,
D.
(1987).
The
etiology
of
admin-
istrator burnout.
Alberta
Journal
of
Educational
Research,
33,
163-179.
*Savicki,
V.,
&
Cooley,
E.
(1987).
The
relationship
of
work
en-
vironment
and
client contact
to
burnout
in
mental health
professionals.
Journal
of
Counseling
and
Development,
65,
249-252.
*Saxton,
M.
J.,
Phillips,
J.
S.,
&
Blakeney,
R. N.
(1991).
Ante-
cedents
and
consequences
of
emotional exhaustion
in the
air-
line
reservations service sector. Human Relations,
44,
583-
595.
Schaufeli,
W. B.,
Enzmann,
D.,
&Girault,
N.
(1993).
Measure-
ment
of
burnout:
A
review.
In W. B.
Schaufeli,
C.
Maslach,
&
T.
Marek
(Eds.),
Professional
burnout: Recent
developments
in
theory
and
research
(pp.
199-215).
Washington,
DC:
Tay-
lor &
Francis.
*Schwab,
R.
L.,
Jackson,
S.
E.,
&
Schuler,
R. S.
(1986).
Educa-
tor
burnout: Sources
and
consequences. Educational
Re-
search
Quarterly,
10,
14-29.
Shirom,
A.
(1989).
Burnout
in
work organizations.
In
C. L.
Cooper
& I.
Robertson
(Eds.),
International review
of
indus-
trial
and
organizational
psychology
(pp.
25-48).
New
\brk:
Wiley.
•Starnaman,
S.
M.,
&
Miller,
K. I.
(1992).
A
test
of a
causal
model
of
communications
and
burnout
in the
teaching pro-
fession.
Communication
Education,
41,
40-53.
*Stevens,
G.
B.,
&
O'Neill,
P.
(1983).
Expectation
and
burnout
in
the
developmental
disabilities
field.
American Journal
of
Community
Psychology,
11,
615-627.
Tversky,
A.,
&
Kahneman,
D.
(1981,
January
30).
The
framing
of
decision
and the
psychology
of
choice. Science,
24,
453-
458.
•Vealey.
R.
S.,
Udry,
E.
M.,
Zimmerman,
V.,
&
Soliday,
J.
(1992).
Intrapersonal
and
situational
predictors
of
coaching
burnout. Journal
of
Sport
and
Exercise
Psychology,
14, 40-
58.
*Whitehead,
J. T.
(1987).
Probation
officer
job
burnout:
A
test
of
two
theories. Journal
of
Criminal Justice,
15,
1-16.
Whitener,
E. M.
(1990).
Confusion
of
confidence intervals
and
credibility intervals
in
meta-analysis.
Journal
of
Applied
Psy-
chology,
73,315-321.
*Wilcoxon,
S. A.
(1989).
Leadership behaviour
and
therapist
burnout:
A
study
of
rural agency
settings.
Journal
of
Rural
Community
Psychology,
10,
3-14.
•Williams,
C. A.
(1989).
Empathy
and
burnout
in
male
and
female
helping professionals. Research
in
Nursing
and
Health,
12,
169-178.
*Wolpin,
J.,
Burke,
R.
J.,
&
Greenglass,
E. R.
(1991).
Is job
satisfaction
an
antecedent
or a
consequence
of
psychological
burnout? Human Relations,
44,
193-209.
Received
December
20,
1994
Revision
received
August
3,
1995
Accepted
October
27,
1995
•
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.